repo_name
stringlengths
6
97
path
stringlengths
3
341
text
stringlengths
8
1.02M
Alqueraf/lms-app
foosball/adb_foos.py
import CHIP_IO.GPIO as GPIO import time import os import urllib2 import socket import logging from subprocess import call TIMESTR = time.strftime("%Y%m%d-%H%M%S") logging.basicConfig(level=logging.DEBUG, filename='foos_' + TIMESTR + '.log') TABLE_ID = "5FLOOR" BTNRESET = "GPIO2" BTN1 = "GPIO3" BTN2 = "GPIO4" LED1 = "GPIO5" LED2 = "GPIO6" OFF = GPIO.HIGH ON = GPIO.LOW FLASH_TIME = 0.2 FLASH_COUNT = 10 GPIO.setup(LED1, GPIO.OUT) GPIO.setup(LED2, GPIO.OUT) GPIO.output(LED1, OFF) GPIO.output(LED2, OFF) GPIO.setup(BTN1, GPIO.IN) GPIO.setup(BTN2, GPIO.IN) GPIO.setup(BTNRESET, GPIO.IN) def printlog(message): print(message) logging.info(message) def blink(speed, count, *leds): while count > 0: count -= 1 for led in leds: GPIO.output(led, ON) time.sleep(speed) for led in leds: GPIO.output(led, OFF) time.sleep(speed) def command(command): call(command, shell=True) # Initialize GPIO.output(LED1, ON) GPIO.output(LED2, ON) #Init adb printlog("Initilizing ADB") command('adb devices') blink(FLASH_TIME, FLASH_COUNT, LED1, LED2) flashing = False def updateScore(side): try: if side == 0: command('adb shell am broadcast -a action_goal --es scoringSide "SIDE_1"') else: command('adb shell am broadcast -a action_goal --es scoringSide "SIDE_2"') return True except: logging.exception("Update Score failure:") return False def buttoncallback(button): global flashing if flashing: return flashing = True if button == BTN1: printlog("TEAM 1 GOAL!") GPIO.output(LED1, ON) if updateScore(0): blink(FLASH_TIME, FLASH_COUNT, LED1) else: GPIO.output(LED1, OFF) else: printlog("TEAM 2 GOAL!") GPIO.output(LED2, ON) if updateScore(1): blink(FLASH_TIME, FLASH_COUNT, LED2) else: GPIO.output(LED2, OFF) flashing = False GPIO.add_event_detect(BTN1, GPIO.FALLING, buttoncallback) GPIO.add_event_detect(BTN2, GPIO.FALLING, buttoncallback) printlog ("Awaiting Button press") try: GPIO.wait_for_edge(BTNRESET, GPIO.FALLING) printlog("Reset button pressed") blink(0.1, 20, LED1, LED2) os.system("reboot") except: logging.exception("Exiting reset button loop:") GPIO.cleanup() printlog("Cleaning up") GPIO.cleanup()
PhoeniXuzoo/NU-Projects
EE475/Ch6P16.py
<filename>EE475/Ch6P16.py<gh_stars>0 import numpy as np import matplotlib.pyplot as plt def readData(csvname): data = np.loadtxt(csvname, delimiter=',') x = data[:-1, :] y = data[-1:, :] return x, y fot = lambda x : np.exp(-x) / (1 + np.exp(-x)) sot = lambda x : ( 1 / (1 + np.exp(x))) * (1 - ( 1 / (1 + np.exp(x)))) # power is y_p * x_p.T * w # firstOrderTerm is e^-power / (1 + e^-power) def first_order(x, y, w, class_weight, power): total = np.zeros(w.shape) firstOrderTerm = fot(power) for i in range(np.size(y)): total += class_weight[i] * firstOrderTerm[:,i] * y[:,i] * x[:,[i]] return (-1) * (total / np.size(y)) def second_order(x, y, w, class_weight, power): total = np.zeros([x.shape[0], x.shape[0]]) secondOrderTerm = sot(power) for i in range(np.size(y)): total += class_weight[i] * secondOrderTerm[:, i] * x[:, [i]] * x[:, [i]].T return total / np.size(y) def newton_method(x, y, w, class_weight): power = y * np.transpose(np.dot(x.T, w)) firstOrder = first_order(x, y, w, class_weight, power) secondOrder = second_order(x, y, w, class_weight, power) return w - np.dot(np.linalg.inv(secondOrder), firstOrder) def costFunc(x, y, w, class_weight): temp = np.log(1 + np.exp(-y*np.transpose(np.dot(np.transpose(x), w)))) cost = 0 for i in range(np.size(y)): cost += temp[0][i] * class_weight[i] return cost / float(np.size(y)) if __name__ == "__main__": csvname = '3d_classification_data_v2_mbalanced.csv' x, y = readData(csvname) w = np.ones([x.shape[0] + 1, 1]) x = np.insert(x, 0, values=np.ones([1, x.shape[1]]), axis=0) positiveOneWeight = 7/11 negativeOneWeight = 4/11 class_weight = [] for i in range(np.size(y)): if (y[:, i] > 0): class_weight.append(positiveOneWeight) else: class_weight.append(negativeOneWeight) position = x[[1, 2]] positiveOneXList = [] positiveOneYList = [] negativeOneXList = [] negativeOneYList = [] for i in range(position.shape[1]): if (y[0][i] > 0): positiveOneXList.append(position[0][i]) positiveOneYList.append(position[1][i]) else: negativeOneXList.append(position[0][i]) negativeOneYList.append(position[1][i]) plt.scatter(positiveOneXList, positiveOneYList, color='red') plt.scatter(negativeOneXList, negativeOneYList, color='blue') for i in range(5): w = newton_method(x, y, w, class_weight) a = -(w[1][0]/w[2][0]) b = -(w[0][0]/w[2][0]) foo = lambda x : a * x + b i = -0.1 xList = [] yList = [] while (i < 1.1): xList.append(i) yList.append(foo(i)) i += 0.1 plt.plot(xList, yList) plt.show()
PhoeniXuzoo/NU-Projects
EE475/Ch6P13.py
import numpy as np import matplotlib.pyplot as plt ## softmax: 0.1 600 ## perceptron: 0.05 550 def readData(csvname): data = np.loadtxt(csvname, delimiter=',') x = data[:-1, :] y = data[-1:, :] return x, y def softmaxCostFunc(x, y, w): cost = np.sum(np.log(1 + np.exp(-y*np.transpose(np.dot(np.transpose(x), w))))) return cost / float(np.size(y)) def gradientDescentOneStepForSoftmax(x, y, w, alpha=0.1): total = np.zeros([9,1]) for i in range(np.size(y)): power = np.exp(-y[:,i] * np.dot(x[:,i], w)) term = power / (1 + power) total += term * y[:,i] * x[:,[i]] w = w + alpha * (1/np.size(y)) * total return w def perceptronCostFunc(x, y, w): cost = 0 a = (-y*np.transpose(np.dot(np.transpose(x), w)))[0] for i in range(len(a)): cost += a[i] if (a[i] > 0) else 0 return cost / float(np.size(y)) def gradientDescentOneStepForPerceptron(x, y, w, alpha=0.05): total = np.zeros([9,1]) for i in range(np.size(y)): term = -y[:,i] * np.dot(x[:,[i]].T, w) total += 0 if term <= 0 else -y[:,i] * x[:,[i]] w = w - alpha * (1/np.size(y)) * total return w if __name__ == "__main__": csvname = 'breast_cancer_data.csv' x, y = readData(csvname) w = np.ones([x.shape[0] + 1, 1]) x = np.insert(x, 0, values=np.ones([1, x.shape[1]]), axis=0) xSoftList = [0] ySoftList = [softmaxCostFunc(x, y, w)] for i in range(600): w = gradientDescentOneStepForSoftmax(x, y, w) xSoftList.append(i+1) ySoftList.append(softmaxCostFunc(x, y, w)) yPredic = np.transpose(np.dot(np.transpose(x), w)) wrong = 0 for i in range(np.size(yPredic)): if ((yPredic[0][i] > 0) != (y[0][i] > 0)): wrong += 1 print("Softmax Wrong Prediction: ", wrong) w = np.ones([x.shape[0], 1]) xPerceptronList = [0] yPerceptronList = [perceptronCostFunc(x, y, w)] for i in range(550): w = gradientDescentOneStepForPerceptron(x, y, w) xPerceptronList.append(i+1) yPerceptronList.append(perceptronCostFunc(x, y, w)) plt.plot(xSoftList, ySoftList, label="Softmax Cost Function",color="#F08080") plt.plot(xPerceptronList, yPerceptronList, label="Perceptro Cost Function") plt.legend(loc="upper right") plt.show() plt.close() yPredic = np.transpose(np.dot(np.transpose(x), w)) wrong = 0 for i in range(np.size(yPredic)): if ((yPredic[0][i] > 0) != (y[0][i] > 0)): wrong += 1 print("Perceptron Wrong Prediction: ", wrong)
PhoeniXuzoo/NU-Projects
EE475/9.2.py
<filename>EE475/9.2.py import ssl from sklearn.datasets import fetch_openml import autograd.numpy as np from autograd import grad from autograd.misc.flatten import flatten_func import matplotlib.pyplot as plt import random # import MNIST x, y = fetch_openml('mnist_784', version=1, return_X_y=True) # re-shape input/output data x = x.T y = np.array([int(v) for v in y])[np.newaxis,:] for i in range(len(x)): if np.average(x[i]) == 0.0: continue x[i] = (x[i] - np.mean(x[i])) / np.std(x[i]) num_sample = 50000 inds = np.random.permutation(y.shape[1])[:num_sample] x = x[:,inds] y = y[:,inds] lam=0.00001 print(np.shape(x)) print(np.shape(y)) x_1=x y_1=y import data_transformer x_sample_edgebased_features = data_transformer.edge_transformer(x) x_2=x_sample_edgebased_features print('shape of transformed input ', x_sample_edgebased_features.shape) def model(x,w): a = w[0] + np.dot(x.T, w[1:]) return a.T def multiclass_softmax(w): all_evals = model(x, w) a = np.log(np.sum(np.exp(all_evals), axis=0)) b = all_evals[y.astype(int).flatten(), np.arange(np.size(y))] cost = np.sum(a - b) cost = cost + lam * np.linalg.norm(w[1:, :], 'fro') ** 2 return cost / float(np.size(y)) def gradientDescent(w, mp_flat): gradient = grad(mp_flat) grad_eval = gradient(w) return w - 0.01 * grad_eval #w = 0.1*np.ones([x.shape[0] + 1, 10]) #ms_flat, unflatten_func, w = flatten_func(multiclass_softmax, w) def training(x_1): w = 0.01 * np.ones([x_1.shape[0] + 1, 10]) ms_flat, unflatten_func, w = flatten_func(multiclass_softmax, w) xList = [0] yList = [ms_flat(w)] for epoch in range(20): for i in range(250): x = x_1[:, i*200:i*200 + 200] y = y_1[:, i*200:i*200 + 200] w = gradientDescent(w, ms_flat) xList.append(epoch + 1) yList.append(ms_flat(w)) plt.plot(xList, yList) training(x_1) print("Succuss!") x=x_2 training(x_2) print("Finish!") plt.show()
PhoeniXuzoo/NU-Projects
EE475/edgebased_feature_extractor.py
<gh_stars>0 from autograd import numpy as np class tensor_conv_layer: # convolution function def conv_function(self, tensor_window): tensor_window = np.reshape(tensor_window, ( np.shape(tensor_window)[0], np.shape(tensor_window)[1] * np.shape(tensor_window)[2])) t = np.dot(self.kernels, tensor_window.T) return t # pooling / downsampling parameters def pool_function(self, tensor_window): t = np.max(tensor_window, axis=(1, 2)) return t # activation def activation(self, tensor_window): return np.maximum(0, tensor_window) # pad image with appropriate number of zeros for convolution def pad_tensor(self, tensor, kernel_size): odd_nums = np.array([int(2 * n + 1) for n in range(100)]) pad_val = np.argwhere(odd_nums == kernel_size)[0][0] tensor_padded = np.zeros( (np.shape(tensor)[0], np.shape(tensor)[1] + 2 * pad_val, np.shape(tensor)[2] + 2 * pad_val)) tensor_padded[:, pad_val:-pad_val, pad_val:-pad_val] = tensor return tensor_padded # sliding window for image augmentation def sliding_window_tensor(self, tensor, window_size, stride, func): # grab image size, set container for results image_size = np.shape(tensor)[1] results = [] # slide window over input image with given window size / stride and function for i in np.arange(0, image_size - window_size + 1, stride): for j in np.arange(0, image_size - window_size + 1, stride): # take a window of input tensor tensor_window = tensor[:, i:i + window_size, j:j + window_size] # now process entire windowed tensor at once tensor_window = np.array(tensor_window) yo = func(tensor_window) # store weight results.append(yo) # re-shape properly results = np.array(results) results = results.swapaxes(0, 1) if func == self.conv_function: results = results.swapaxes(1, 2) return results # make feature map def make_feature_tensor(self, tensor): # create feature map via convolution --> returns flattened convolution calculations conv_stride = 1 feature_tensor = self.sliding_window_tensor(tensor, self.kernel_size, conv_stride, self.conv_function) # re-shape convolution output ---> to square of same size as original input num_filters = np.shape(feature_tensor)[0] num_images = np.shape(feature_tensor)[1] square_dim = int((np.shape(feature_tensor)[2]) ** (0.5)) feature_tensor = np.reshape(feature_tensor, (num_filters, num_images, square_dim, square_dim)) # shove feature map through nonlinearity feature_tensor = self.activation(feature_tensor) # pool feature map --- i.e., downsample it pool_stride = 2 pool_window_size = 3 downsampled_feature_map = [] for t in range(np.shape(feature_tensor)[0]): temp_tens = feature_tensor[t, :, :, :] d = self.sliding_window_tensor(temp_tens, pool_window_size, pool_stride, self.pool_function) downsampled_feature_map.append(d) downsampled_feature_map = np.array(downsampled_feature_map) # return downsampled feature map --> flattened return downsampled_feature_map # our normalization function def normalize(self, data, data_mean, data_std): normalized_data = (data - data_mean) / (data_std + 10 ** (-5)) return normalized_data # convolution layer def conv_layer(self, tensor, kernels): #### prep input tensor ##### # pluck out dimensions for image-tensor reshape num_images = np.shape(tensor)[0] num_kernels = np.shape(kernels)[0] # create tensor out of input images (assumed to be stacked vertically as columns) tensor = np.reshape(tensor, ( np.shape(tensor)[0], int((np.shape(tensor)[1]) ** (0.5)), int((np.shape(tensor)[1]) ** (0.5))), order='F') # pad tensor kernel = kernels[0] self.kernel_size = np.shape(kernel)[0] padded_tensor = self.pad_tensor(tensor, self.kernel_size) #### prep kernels - reshape into array for more effecient computation #### self.kernels = np.reshape(kernels, (np.shape(kernels)[0], np.shape(kernels)[1] * np.shape(kernels)[2])) #### compute convolution feature maps / downsample via pooling one map at a time over entire tensor ##### # compute feature map for current image using current convolution kernel feature_tensor = self.make_feature_tensor(padded_tensor) feature_tensor = feature_tensor.swapaxes(0, 1) feature_tensor = np.reshape(feature_tensor, ( np.shape(feature_tensor)[0], np.shape(feature_tensor)[1] * np.shape(feature_tensor)[2]), order='F') return feature_tensor
PhoeniXuzoo/NU-Projects
EE475/Ch7P3.py
<gh_stars>0 import autograd.numpy as np import matplotlib.pyplot as plt from autograd import grad from autograd.misc.flatten import flatten_func def readData(csvname): data = np.loadtxt(csvname, delimiter=',') x = data[:-1, :] y = data[-1:, :] return x, y def model(x, w): a = w[0] + np.dot(x.T, w[1:]) return a.T def multiclass_perceptron (w): all_evals = model(x, w) a = np.max(all_evals, axis = 0) b = all_evals[y.astype(int).flatten(), np.arange(np.size(y))] cost = np.sum(a - b) cost = cost + lam*np.linalg.norm(w[1: ,:], 'fro')**2 return cost / float(np.size(y)) def gradientDescent(w, mp_flat): gradient = grad(mp_flat) grad_eval = gradient(w) return w - 0.1 * grad_eval def statistic(x, y, w): prediction = np.argmax(model(x,w),axis=0) wrong = 0 for i in range(np.size(y)): if prediction[i] != y[0][i]: wrong += 1 return wrong # global variable lam = 10 ** -5 csvname = '3class_data.csv' x, y = readData(csvname) lam = 10 ** -5 if __name__ == "__main__": w = np.ones([x.shape[0] + 1, 3]) w += 1 mp_flat, unflatten_func, w = flatten_func(multiclass_perceptron, w) xList = [0] yList = [mp_flat(w)] for i in range(100): w = gradientDescent(w, mp_flat) xList.append(i+1) yList.append(mp_flat(w)) w = w.reshape(3, 3) plt.plot(xList, yList) plt.show() print("misclassification: ", statistic(x, y, w))
PhoeniXuzoo/NU-Projects
EE475/Ch3P5.py
import numpy as np import matplotlib.pyplot as plt from matplotlib.backends.backend_pdf import PdfPages def gradientDescentEveryStep(derivative, startPoint, stepLength): return startPoint - stepLength * derivative(startPoint) if __name__ == "__main__": # g(w) = (1/50) * ( w^4 + w^2 + w) with PdfPages('Chapter3Practice5.pdf') as pdf: polynomial = np.poly1d([1/50, 0, 1/50, 1/5, 0]) # derivative of g(w) derivative = polynomial.deriv() stepLength = [1, 0.1, 0.01] for alpha in stepLength: startPoint = 2 fig = plt.figure() plt.title("Exercise 3.5: steplength is " + str(alpha)) xList = [0] yList = [polynomial(startPoint)] for i in range(1000): startPoint = gradientDescentEveryStep(derivative, startPoint, alpha) xList.append(i+1) yList.append(polynomial(startPoint)) plt.plot(xList, yList) plt.show() pdf.savefig(fig) plt.close()
PhoeniXuzoo/NU-Projects
EE475/Ch3P6.py
<reponame>PhoeniXuzoo/NU-Projects import numpy as np import matplotlib.pyplot as plt from matplotlib.backends.backend_pdf import PdfPages def gradientDescentEveryStep(derivative, startPoint, stepLength): return startPoint - stepLength * derivative(startPoint) if __name__ == "__main__": # g(w) = |x| with PdfPages('Chapter3Practice6.pdf') as pdf: costFunc = lambda x : abs(x) derivative = lambda x: 1 if (x >= 0) else -1 # fixed steplength startPoint = 1.75 fig = plt.figure() plt.title("Exercise 3.6: fixed steplength is 0.5") xList = [0] yList = [costFunc(startPoint)] for i in range(20): startPoint = gradientDescentEveryStep(derivative, startPoint, 0.5) xList.append(i+1) yList.append(costFunc(startPoint)) plt.plot(xList, yList) plt.show() pdf.savefig(fig) plt.close() # diminishing step length startPoint = 1.75 fig = plt.figure() plt.title("Exercise 3.6: diminishing steplength") xList = [0] yList = [costFunc(startPoint)] for i in range(20): startPoint = gradientDescentEveryStep(derivative, startPoint, 1/(i+1)) xList.append(i + 1) yList.append(costFunc(startPoint)) plt.plot(xList, yList) plt.show() pdf.savefig(fig) plt.close()
PhoeniXuzoo/NU-Projects
EE475/Ch4P5c.py
from autograd import grad from autograd import hessian import autograd.numpy as np import matplotlib.pyplot as plt def newtons_method(g, max_its, w): gradient = grad(g) hess = hessian(g) epsilon = 10 ** (-7) weight_history = [w] cost_history = [g(w)] for k in range(max_its): grad_eval = gradient(w) hess_eval = hess(w) hess_eval.shape = (int((np.size(hess_eval)) ** (0.5)), int((np.size(hess_eval)) ** (0.5))) A = hess_eval + epsilon * np.eye(w.size) b = grad_eval w = np.linalg.solve(A, np.dot(A, w) - b) weight_history.append(w) cost_history.append(g(w)) return weight_history, cost_history def g(w): return np.log(1 + np.exp(np.dot(np.transpose(w), w))) if __name__ == "__main__": w = 1 * np.ones([2, 1]) max_its = 10 weight_history, cost_history = newtons_method(g, max_its, w) print(weight_history) print(cost_history) xList = [] yList = [] for i in range(max_its): xList.append(i+1) yList.append(cost_history[i][0][0]) fig = plt.figure() plt.title("Exercise 4.5 (c)") plt.plot(xList, yList) plt.show() plt.close()
PhoeniXuzoo/NU-Projects
EE475/Ch3P8.py
import numpy as np import matplotlib.pyplot as plt from matplotlib.backends.backend_pdf import PdfPages def gradientDescentEveryStep(derivative, startPoint, stepLength): return startPoint - stepLength * derivative(startPoint) def constFunc(w): total = 0 for i in w: total += i*i return total if __name__ == "__main__" : # g(w) = w^T * w w is a 10*1 vector with PdfPages('Chapter3Practice8.pdf') as pdf: derivate = lambda w: 2 * w stepLength = [0.001, 0.1, 1] for alpha in stepLength: startPoint = np.ones(10) * 10 fig = plt.figure() plt.title("Exercise 3.8: steplength is " + str(alpha)) xList = [0] yList = [constFunc(startPoint)] for i in range(100): startPoint = gradientDescentEveryStep(derivate, startPoint, alpha) xList.append(i+1) yList.append(constFunc(startPoint)) plt.plot(xList, yList) plt.show() pdf.savefig(fig) plt.close()
PhoeniXuzoo/NU-Projects
EE475/Ch9P2.py
<reponame>PhoeniXuzoo/NU-Projects import ssl from sklearn.datasets import fetch_openml import autograd.numpy as np import matplotlib.pyplot as plt from matplotlib.pyplot import MultipleLocator from autograd import grad from autograd.misc.flatten import flatten_func import copy ssl._create_default_https_context = ssl._create_unverified_context class tensor_conv_layer: def conv_function(self, tensor_window): tensor_window = np.reshape(tensor_window, ( np.shape(tensor_window)[0], np.shape(tensor_window)[1] * np.shape(tensor_window)[2])) t = np.dot(self.kernels, tensor_window.T) return t def pool_function(self, tensor_window): t = np.max(tensor_window, axis=(1, 2)) return t def activation(self, tensor_window): return np.maximum(0, tensor_window) def pad_tensor(self, tensor, kernel_size): odd_nums = np.array([int(2 * n + 1) for n in range(100)]) pad_val = np.argwhere(odd_nums == kernel_size)[0][0] tensor_padded = np.zeros( (np.shape(tensor)[0], np.shape(tensor)[1] + 2 * pad_val, np.shape(tensor)[2] + 2 * pad_val)) tensor_padded[:, pad_val:-pad_val, pad_val:-pad_val] = tensor return tensor_padded def sliding_window_tensor(self, tensor, window_size, stride, func): image_size = np.shape(tensor)[1] results = [] for i in np.arange(0, image_size - window_size + 1, stride): for j in np.arange(0, image_size - window_size + 1, stride): tensor_window = tensor[:, i:i + window_size, j:j + window_size] tensor_window = np.array(tensor_window) yo = func(tensor_window) results.append(yo) results = np.array(results) results = results.swapaxes(0, 1) if func == self.conv_function: results = results.swapaxes(1, 2) return results def make_feature_tensor(self, tensor): conv_stride = 1 feature_tensor = self.sliding_window_tensor(tensor, self.kernel_size, conv_stride, self.conv_function) num_filters = np.shape(feature_tensor)[0] num_images = np.shape(feature_tensor)[1] square_dim = int((np.shape(feature_tensor)[2]) ** (0.5)) feature_tensor = np.reshape(feature_tensor, (num_filters, num_images, square_dim, square_dim)) feature_tensor = self.activation(feature_tensor) pool_stride = 2 pool_window_size = 3 downsampled_feature_map = [] for t in range(np.shape(feature_tensor)[0]): temp_tens = feature_tensor[t, :, :, :] d = self.sliding_window_tensor(temp_tens, pool_window_size, pool_stride, self.pool_function) downsampled_feature_map.append(d) downsampled_feature_map = np.array(downsampled_feature_map) return downsampled_feature_map def normalize(self, data, data_mean, data_std): normalized_data = (data - data_mean) / (data_std + 10 ** (-5)) return normalized_data def conv_layer(self, tensor, kernels): num_images = np.shape(tensor)[0] num_kernels = np.shape(kernels)[0] tensor = np.reshape(tensor, ( np.shape(tensor)[0], int((np.shape(tensor)[1]) ** (0.5)), int((np.shape(tensor)[1]) ** (0.5))), order='F') kernel = kernels[0] self.kernel_size = np.shape(kernel)[0] padded_tensor = self.pad_tensor(tensor, self.kernel_size) self.kernels = np.reshape(kernels, (np.shape(kernels)[0], np.shape(kernels)[1] * np.shape(kernels)[2])) feature_tensor = self.make_feature_tensor(padded_tensor) feature_tensor = feature_tensor.swapaxes(0, 1) feature_tensor = np.reshape(feature_tensor, ( np.shape(feature_tensor)[0], np.shape(feature_tensor)[1] * np.shape(feature_tensor)[2]), order='F') return feature_tensor def readData(): x, y = fetch_openml('mnist_784', version=1, return_X_y=True) x = x.T y = np.array([int(v) for v in y])[np.newaxis,:] for i in range(len(x)): if np.average(x[i]) == 0.0: continue x[i] = (x[i] - np.mean(x[i])) / np.std(x[i]) return x[:, 0:50000], y[:, 0:50000] def edge_transformer(x): kernels = np.array([ [[-1, -1, -1], [ 0, 0, 0], [ 1, 1, 1]], [[-1, -1, 0], [-1, 0, 1], [ 0, 1, 1]], [[-1, 0, 1], [-1, 0, 1], [-1, 0, 1]], [[ 0, 1, 1], [-1, 0, 1], [-1, -1, 0]], [[ 1, 0, -1], [ 1, 0, -1], [ 1, 0, -1]], [[ 0, -1, -1], [ 1, 0, -1], [ 1, 1, 0]], [[ 1, 1, 1], [ 0, 0, 0], [-1, -1, -1]], [[ 1, 1, 0], [ 1, 0, -1], [ 0, -1, -1]]]) demo = tensor_conv_layer() x_transformed = demo.conv_layer(x.T,kernels).T return x_transformed #global variable x_batch, y_batch = readData() x = x_batch y = y_batch x_edgebased_features = edge_transformer(x) lam = 10**(-5) def model(vector,w): a = w[0] + np.dot(vector.T, w[1:]) return a.T def multiclass_softmax(w): all_evals = model(x, w) a = np.log(np.sum(np.exp(all_evals), axis=0)) b = all_evals[y.astype(int).flatten(), np.arange(np.size(y))] cost = np.sum(a - b) cost = cost + lam * np.linalg.norm(w[1:, :], 'fro') ** 2 return cost / float(np.size(y)) def gradientDescent(w, mp_flat): gradient = grad(mp_flat) grad_eval = gradient(w) return w - 0.01 * grad_eval if __name__ == "__main__": plt.figure() w = np.ones([x_batch.shape[0] + 1, 10]) w *= 0.01 y1wrong = [] prediction = np.argmax(model(x_batch, w), axis=0) wrong = 0 for i in range(np.size(y_batch)): if prediction[i] != y_batch[0][i]: wrong += 1 y1wrong.append(wrong) ms_flat, unflatten_func, w = flatten_func(multiclass_softmax, w) xList = [0] x = x_batch[:, 0:200] y = y_batch[:, 0:200] yList = [ms_flat(w)] for epoch in range(20): for i in range(250): x = x_batch[:, i * 200: i * 200 + 200] y = y_batch[:, i * 200: i * 200 + 200] w = gradientDescent(w, ms_flat) xList.append(epoch + 1) yList.append(ms_flat(w)) w_temp = copy.deepcopy(w) w_temp = w_temp.reshape(x_batch.shape[0] + 1, 10) prediction = np.argmax(model(x_batch, w_temp), axis=0) wrong = 0 for i in range(np.size(y_batch)): if prediction[i] != y_batch[0][i]: wrong += 1 y1wrong.append(wrong) plt.plot(xList, yList, label="raw data point", color='blue') w = np.ones([x_edgebased_features.shape[0] + 1, 10]) w *= 0.01 y2wrong = [] prediction = np.argmax(model(x_edgebased_features, w), axis=0) wrong = 0 for i in range(np.size(y_batch)): if prediction[i] != y_batch[0][i]: wrong += 1 y2wrong.append(wrong) ms_flat, unflatten_func, w = flatten_func(multiclass_softmax, w) xList = [0] x = x_edgebased_features[:, 0:200] y = y_batch[:, 0:200] yList = [ms_flat(w)] for epoch in range(20): for i in range(250): x = x_edgebased_features[:, i * 200: i * 200 + 200] y = y_batch[:, i * 200: i * 200 + 200] w = gradientDescent(w, ms_flat) xList.append(epoch + 1) yList.append(ms_flat(w)) w_temp = copy.deepcopy(w) w_temp = w_temp.reshape(x_edgebased_features.shape[0] + 1, 10) prediction = np.argmax(model(x_edgebased_features, w_temp), axis=0) wrong = 0 for i in range(np.size(y_batch)): if prediction[i] != y_batch[0][i]: wrong += 1 y2wrong.append(wrong) plt.plot(xList, yList, label="edge histogram based", color="red") plt.legend(loc="upper right") xmajorLocator = MultipleLocator(5) ax = plt.gca() ax.xaxis.set_major_locator(xmajorLocator) plt.show() plt.close() plt.figure() plt.plot(xList, y1wrong, label="raw data point", color="blue") plt.plot(xList, y2wrong, label="edge histogram based", color="red") plt.legend(loc="upper right") ymajorLocator = MultipleLocator(1000) ax = plt.gca() ax.xaxis.set_major_locator(xmajorLocator) ax.yaxis.set_major_locator(ymajorLocator) plt.ylim(0, 10000) plt.show() plt.close()
PhoeniXuzoo/NU-Projects
EE475/Ch5P9b.py
<reponame>PhoeniXuzoo/NU-Projects<gh_stars>0 import numpy as np def readData(csvname): data = np.loadtxt(csvname, delimiter=',') x = data[:-1, :] y = data[-1:, :] for i in range(len(x)): x[i] = (x[i] - np.mean(x[i])) / np.std(x[i]) return x, y def gradientDescentEachStep(w, x, y): w_T = np.transpose(w) new_W = [] for i in range(len(w)): total = 0 for j in range(len(y[0])): total += (np.dot(w_T, np.transpose(np.array([x[:,j]]))) - y[0][j]) * x[i][j] new_W.append(w[i][0] - 0.1 * (1/len(y[0])) * total) for i in range(len(w)): w[i][0] = new_W[i] return w if __name__ == "__main__": csvname = 'auto_data.csv' x, y = readData(csvname) w = np.ones([x.shape[0] + 1,1]) x = np.insert(x, 0, values=np.ones([1, x.shape[1]]), axis=0) for i in range(1000): w = gradientDescentEachStep(w, x, y) totalForRMSE = 0 totalForMAD = 0 w_T = np.transpose(w) for i in range(len(y[0])): temp = (np.dot(w_T, np.transpose(np.array([x[:,i]]))) - y[0][i]) totalForRMSE += temp ** 2 totalForMAD += abs(temp) MSE = totalForRMSE / len(y[0]) RMSE = MSE ** 0.5 MAD = totalForMAD / len(y[0]) print("RMSE is ", RMSE) print("MAD is ", MAD)
PhoeniXuzoo/NU-Projects
EE475/Ch5P2.py
import csv import numpy as np import math def readData(): rawData = csv.reader(open("kleibers_law_data.csv", 'r')) xList = [] yList = [] i = 0 for raw in rawData: if i == 0: xList = raw i += 1 else: yList = raw for i in range(len(xList)): xList[i] = math.log(float(xList[i])) yList[i] = math.log(float(yList[i])) return xList, yList if __name__ == "__main__": xList, yList = readData() xBar = np.mean(xList) yBar = np.mean(yList) a = 0 b = 0 for i in range(len(xList)): a += (xList[i] - xBar) * (yList[i] - yBar) b += (xList[i] - xBar) ** 2 w1 = a / b w2 = yBar - w1 * xBar print(w1, w2)
PhoeniXuzoo/NU-Projects
EE475/data_transformer.py
from autograd import numpy as np import edgebased_feature_extractor # edge-based feature extraction def edge_transformer(x): # edge-based directions for kernel-based feature extraction kernels = np.array([ [[-1, -1, -1], [ 0, 0, 0], [ 1, 1, 1]], [[-1, -1, 0], [-1, 0, 1], [ 0, 1, 1]], [[-1, 0, 1], [-1, 0, 1], [-1, 0, 1]], [[ 0, 1, 1], [-1, 0, 1], [-1, -1, 0]], [[ 1, 0, -1], [ 1, 0, -1], [ 1, 0, -1]], [[ 0, -1, -1], [ 1, 0, -1], [ 1, 1, 0]], [[ 1, 1, 1], [ 0, 0, 0], [-1, -1, -1]], [[ 1, 1, 0], [ 1, 0, -1], [ 0, -1, -1]]]) # compute edge-based features demo = edgebased_feature_extractor.tensor_conv_layer() x_transformed = demo.conv_layer(x.T,kernels).T return x_transformed
PhoeniXuzoo/NU-Projects
EE475/Ch7P2.py
<reponame>PhoeniXuzoo/NU-Projects import numpy as np import matplotlib.pyplot as plt import copy def readData(csvname): data = np.loadtxt(csvname, delimiter=',') x = data[:-1, :] y = data[-1:, :] return x, y def drawpoint(x, y): position = x[[1,2]] zeroXList = [] zeroYList = [] oneXList = [] oneYList = [] twoXList = [] twoYList = [] threeXList = [] threeYList = [] for i in range(position.shape[1]): if (int(y[0][i]) == 0): zeroXList.append(position[0][i]) zeroYList.append(position[1][i]) elif (int(y[0][i]) == 1): oneXList.append(position[0][i]) oneYList.append(position[1][i]) elif (int(y[0][i]) == 2): twoXList.append(position[0][i]) twoYList.append(position[1][i]) elif (int(y[0][i]) == 3): threeXList.append(position[0][i]) threeYList.append(position[1][i]) plt.scatter(zeroXList, zeroYList, color = 'blue') plt.scatter(oneXList, oneYList, color = 'red') plt.scatter(twoXList, twoYList, color='green') plt.scatter(threeXList, threeYList, color='yellow') return plt def drawline(w, c): a = -(w[1][0] / w[2][0]) b = -(w[0][0] / w[2][0]) foo = lambda x: a * x + b i = 0.0 xList = [] yList = [] while (i < 1.0): xList.append(i) yList.append(foo(i)) i += 0.1 plt.plot(xList, yList, color = c) class Logistic: def __init__(self, label, x, y, w, epoch, alpha): self.label = label self.epoch = epoch self.alpha = alpha self.x = copy.deepcopy(x) self.w = copy.deepcopy(w) self.y = copy.deepcopy(y) for i in range(np.size(self.y)): if (int(self.y[0][i]) == int(label)): self.y[0][i] = 1.0 else: self.y[0][i] = -1.0 # def softmaxCostFunc(self): # cost = np.sum(np.log(1 + np.exp(-(self.y)*np.transpose(np.dot(np.transpose(self.x), self.w))))) # return cost / float(np.size(self.y)) # # def gradientDescentOneStepForSoftmax(self, alpha): # total = np.zeros([3, 1]) # for i in range(np.size(self.y)): # power = np.exp(-self.y[:, i] * np.dot(self.x[:, i], self.w)) # term = power / (1 + power) # total += term * self.y[:, i] * self.x[:, [i]] # # self.w = self.w + alpha * (1 / np.size(self.y)) * total def perceptronCostFunc(self): cost = 0 a = (-(self.y) * np.transpose(np.dot(np.transpose(self.x), self.w)))[0] for i in range(len(a)): cost += a[i] if (a[i] > 0) else 0 return cost / float(np.size(self.y)) def gradientDescentOneStepForPerceptron(self, alpha): total = np.zeros([3, 1]) for i in range(np.size(self.y)): term = -self.y[:, i] * np.dot(self.x[:, [i]].T, self.w) total += 0 if term <= 0 else -self.y[:, i] * self.x[:, [i]] self.w = self.w - alpha * (1 / np.size(self.y)) * total def normalizeW(self): i = 1 sum = 0.0 while (i < np.size(self.w)): sum += self.w[i][0] ** 2 i += 1 return self.w / (sum ** 0.5) # def softmaxTrain(self): # for i in range(self.epoch): # self.gradientDescentOneStepForSoftmax(self.alpha) # return self.normalizeW() def perceptronTrain(self): for i in range(self.epoch): self.gradientDescentOneStepForPerceptron(self.alpha) return self.normalizeW() def statistic(x, y, zeroW, oneW, twoW, threeW): wrong = 0 for i in range(np.shape(x)[1]): zero = np.dot(x[:,[i]].T, zeroW) one = np.dot(x[:,[i]].T, oneW) two = np.dot(x[:,[i]].T, twoW) three = np.dot(x[:,[i]].T, threeW) if (int(y[0][i]) == 0): if not (zero >= one and zero >= two and zero >= three): wrong += 1 elif (int(y[0][i]) == 1): if not (one >= zero and one >= two and one >= three): wrong += 1 elif (int(y[0][i]) == 2): if not (two >= zero and two >= one and two >= three): wrong += 1 elif (int(y[0][i]) == 3): if not (three >= zero and three >= one and three >= two): wrong += 1 return wrong if __name__ == "__main__": csvname = '4class_data.csv' x, y = readData(csvname) w = np.ones([x.shape[0] + 1, 1]) x = np.insert(x, 0, values=np.ones([1, x.shape[1]]), axis=0) plt = drawpoint(x, y) zeroClassifier = Logistic(0.0, x, y, w, 1400, 0.05) zeroW = zeroClassifier.perceptronTrain() drawline(zeroW, 'blue') oneClassifier = Logistic(1.0, x, y, w, 250, 0.05) oneW = oneClassifier.perceptronTrain() drawline(oneW, 'red') w = np.zeros([x.shape[0], 1]) w += 0.05 twoClassifier = Logistic(2.0, x, y, w, 120, 0.01) twoW = twoClassifier.perceptronTrain() drawline(twoW, 'green') threeClassifier = Logistic(3.0, x, y, w, 505, 0.01) threeW = threeClassifier.perceptronTrain() drawline(threeW, 'yellow') plt.show() print("MisClassifications: ", statistic(x, y, zeroW, oneW, twoW, threeW))
PhoeniXuzoo/NU-Projects
EE475/Ch6P15.py
<filename>EE475/Ch6P15.py import numpy as np import matplotlib.pyplot as plt def readData(csvname): data = np.loadtxt(csvname, delimiter=',') x = data[:-1, :] y = data[-1:, :] for i in range(len(x)): x[i] = (x[i] - np.mean(x[i])) / np.std(x[i]) return x, y def perceptronCostFunc(x, y, w, class_index): cost = 0 a = (-y*np.transpose(np.dot(np.transpose(x), w)))[0] for i in range(len(a)): cost += (class_index[i] * a[i]) if (a[i] > 0) else 0 return cost / float(np.size(y)) def gradientDescentOneStepForPerceptron(x, y, w, class_index, alpha=0.00020): total = np.zeros([21,1]) for i in range(np.size(y)): term = -y[:,i] * np.dot(x[:,[i]].T, w) total += 0 if term <= 0 else -y[:,i] * x[:,[i]] * class_index[i] w = w - alpha * (1/np.size(y)) * total return w if __name__ == "__main__": csvname = 'credit_dataset.csv' x, y = readData(csvname) w = np.zeros([x.shape[0] + 1, 1]) w += 0.5 alpha = 0.0025 goodWeight = 1 badWeight = 1 iteration = 5000 x = np.insert(x, 0, values=np.ones([1, x.shape[1]]), axis=0) print("alpha: ", alpha) print("iteration: ", iteration) class_index = [] for i in range(np.size(y)): if (y[:, i] > 0): class_index.append(goodWeight) else: class_index.append(badWeight) xPerceptronList = [0] yPerceptronList = [perceptronCostFunc(x, y, w, class_index)] for i in range(iteration): w = gradientDescentOneStepForPerceptron(x, y, w, class_index, alpha) xPerceptronList.append(i + 1) yPerceptronList.append(perceptronCostFunc(x, y, w, class_index)) plt.plot(xPerceptronList, yPerceptronList) plt.show() plt.close() yPredic = np.transpose(np.dot(np.transpose(x), w)) correct = 0 for i in range(np.size(yPredic)): if ((yPredic[0][i] > 0) == (y[0][i] > 0)): correct += 1 print("Accuracy: ", correct / 1000)
Mitrofanov/enhanced-rds-monitoring
tests/sample_input.py
# Sample RDSOS output from a MySQL instance my_sql_dict = { u"engine": u"MYSQL", u"instanceID": u"ppr9b9oge80i99", u"instanceResourceID": u"db-H4UK4SA7E62QPMAYTMPK5XWETQ", u"timestamp": u"2017-09-19T00:21:07Z", u"version": 1, u"uptime": u"272 days, 1:28:18", u"numVCPUs": 2, u"cpuUtilization": { u"guest": 0, u"irq": 0.72, u"system": 7.25, u"wait": 4.35, u"idle": 78.99, u"user": 3.62, u"total": 21.01, u"steal": 1.45, u"nice": 3.62 }, u"loadAverageMinute": { u"fifteen": 0.4, u"five": 0.3, u"one": 0.29 }, u"memory": { u"writeback": 0, u"hugePagesFree": 0, u"hugePagesRsvd": 0, u"hugePagesSurp": 0, u"cached": 3288068, u"hugePagesSize": 2048, u"free": 2746028, u"hugePagesTotal": 0, u"inactive": 838368, u"pageTables": 5448, u"dirty": 144, u"mapped": 25548, u"active": 3766972, u"total": 7697596, u"slab": 239772, u"buffers": 301616 }, u"tasks": { u"sleeping": 256, u"zombie": 0, u"running": 4, u"stopped": 0, u"total": 262, u"blocked": 2 }, u"swap": { u"cached": 0, u"total": 7703160, u"free": 7703160 }, u"network": [ { u"interface": u"eth0", u"rx": 821640, u"tx": 2856222 } ], u"diskIO": [ { u"writeKbPS": 20, u"readIOsPS": 0, u"await": 1.6, u"readKbPS": 0, u"rrqmPS": 0, u"util": 0.8, u"avgQueueLen": 0.01, u"tps": 5, u"readKb": 0, u"device": u"rdsdev", u"writeKb": 20, u"avgReqSz": 4, u"wrqmPS": 0, u"writeIOsPS": 5 } ], u"fileSys": [ { u"used": 626316, u"name": u"rdsfilesys", u"usedFiles": 523, u"usedFilePercent": 0.01, u"maxFiles": 6553600, u"mountPoint": u"/rdsdbdata", u"total": 103053476, u"usedPercent": 0.61 } ], u"processList": [ { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 0, u"id": 765, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 5.5, u"id": 766, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 0, u"id": 780, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 0, u"id": 781, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 0, u"id": 3305, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 6, u"id": 4111, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 5.5, u"id": 10888, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 6, u"id": 19825, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 5.5, u"id": 31951, u"rss": 703620 }, { u"vss": 6365136, u"name": u"mysqld", u"tgid": 3305, u"parentID": 1, u"memoryUsedPc": 9.14, u"cpuUsedPc": 5.5, u"id": 32015, u"rss": 703620 }, { u"vss": 935380, u"name": u"OS processes", u"tgid": 0, u"parentID": 0, u"memoryUsedPc": 0.35, u"cpuUsedPc": 0.5, u"id": 0, u"rss": 27824 }, { u"vss": 1293204, u"name": u"RDS processes", u"tgid": 0, u"parentID": 0, u"memoryUsedPc": 4.27, u"cpuUsedPc": 4, u"id": 0, u"rss": 329184 } ] } # Sample RDSOS output from a SQL Server instance sql_server_dict = { u"engine": u"SqlServer", u"instanceID": u"trevor", u"instanceResourceID": u"db-YWCA2G6UQEA3NYZ54IS6XEBGUE", u"timestamp": u"2017-09-12T23:58:59Z", u"version": 1, u"uptime": u"0 days, 00:10:43", u"numVCPUs": 1, u"cpuUtilization": { u"idle": 64.13, u"kern": 11.85, u"user": 24.01 }, u"memory": { u"commitTotKb": 1073308, u"commitLimitKb": 1572464, u"commitPeakKb": 1109112, u"physTotKb": 1048176, u"physAvailKb": 125072, u"sysCacheKb": 182448, u"kernTotKb": 192596, u"kernPagedKb": 136120, u"kernNonpagedKb": 56476, u"sqlServerTotKb": 123432, u"pageSize": 4096 }, u"system": { u"handles": 14874, u"threads": 634, u"processes": 44 }, u"disks": [ { u"name": u"rdsdbdata", u"totalKb": 20838336, u"usedKb": 162752, u"usedPc": 0.78, u"availKb": 20675584, u"availPc": 99.22, u"rdCountPS": 0, u"rdBytesPS": 0, u"wrCountPS": 0, u"wrBytesPS": 0 } ], u"network": [ { u"interface": "Ethernet 2", u"rdBytesPS": 0, u"wrBytesPS": 0 } ], u"processList": [ { u"name": u"OS processes", u"cpuUsedPc": 0.3, u"memUsedPc": 8.46, u"workingSetKb": 249908, u"workingSetPrivKb": 88728, u"workingSetShareableKb": 161180, u"virtKb": 57985212532 }, { u"name": u"RDS processes", u"cpuUsedPc": 1.52, u"memUsedPc": 20.61, u"workingSetKb": 367720, u"workingSetPrivKb": 215996, u"workingSetShareableKb": 151724, u"virtKb": 4331792332 }, { u"name": u"sqlwriter.exe", u"pid": 1736, u"cpuUsedPc": 0, u"memUsedPc": 0.1, u"workingSetKb": 5876, u"workingSetPrivKb": 1020, u"workingSetShareableKb": 4856, u"virtKb": 38280 }, { u"name": u"fdlauncher.exe", u"pid": 2436, u"cpuUsedPc": 0, u"memUsedPc": 0.05, u"workingSetKb": 3576, u"workingSetPrivKb": 568, u"workingSetShareableKb": 3008, u"virtKb": 26776 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"cpuUsedPc": 32.22, u"memUsedPc": 11.12, u"workingSetKb": 158676, u"workingSetPrivKb": 116608, u"workingSetShareableKb": 42068, u"virtKb": 2694704 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3628, u"cpuUsedPc": 0 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3636, u"cpuUsedPc": 0 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3644, u"cpuUsedPc": 0 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3648, u"cpuUsedPc": 0 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3660, u"cpuUsedPc": 0 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3664, u"cpuUsedPc": 0 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3668, u"cpuUsedPc": 0 }, { u'name': u"sqlservr.exe", u"pid": 3624, u"tid": 3672, u"cpuUsedPc": 32.22 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3676, u"cpuUsedPc": 0 }, { u"name": u"sqlservr.exe", u"pid": 3624, u"tid": 3680, u"cpuUsedPc": 0 }, ] } # Sample RDSOS output for an Aurora instance aurora_dict = { u"engine": u"Aurora", u"instanceID": u"trevor", u"instanceResourceID": u"db-F3AZGY5JTAS65TT4E7U3CDS6AM", u"timestamp": u"2017-09-12T23:02:53Z", u"version": 1, u"uptime": u"0:44:23", u"numVCPUs": 1, u"cpuUtilization": { u"guest": 0, u"irq": 0.04, u"system": 1.86, u"wait": 2.61, u"idle": 90.08, u"user": 5.01, u"total": 9.92, u"steal": 0.36, u"nice": 0.04 }, u"loadAverageMinute": { u"fifteen": 0.13, u"five": 0.1, u"one": 0.3 }, u"memory": { u"writeback": 0, u"hugePagesFree": 1024, u"hugePagesRsvd": 0, u"hugePagesSurp": 0, u"cached": 489508, u"hugePagesSize": 2048, u"free": 99192, u"hugePagesTotal": 367616, u"inactive": 260216, u"pageTables": 6244, u"dirty": 1400, u"mapped": 84816, u"active": 890436, u"total": 2052380, u"slab": 38916, u"buffers": 67024 }, u"tasks": { u"sleeping": 233, u"zombie": 0, u"running": 2, u"stopped": 0, u"total": 235, u"blocked": 0 }, u"swap": { u"cached": 0, u"total": 0, u"free": 0 }, u"network": [ { u"interface": u"eth0", u"rx": 58724202.3, u"tx": 2613197.9 } ], u"diskIO": [ { u"readLatency": 2.74, u"writeLatency": 1.5, u"writeThroughput": 476409.8, u"readThroughput": 665190.4, u"readIOsPS": 40.6, u"diskQueueDepth": 0, u"writeIOsPS": 1426.8 } ], u"fileSys": [ { u"used": 68956, u"name": u"rdsfilesys", u"usedFiles": 211, u"usedFilePercent": 0.01, u"maxFiles": 2097152, u"mountPoint": u"/rdsdbdata", u"total": 32890736, u"usedPercent": 0.21 } ], u"processList": [ { u"vss": 1232180, u"name": u"aurora", u"tgid": 5439, u"vmlimit": 2052380, u"parentID": 1, u"memoryUsedPc": 12.47, u"cpuUsedPc": 0, u"id": 5439, u"rss": 255908 }, { u"vss": 691648, u"name": u"OS processes", u"tgid": 0, u"vmlimit": u"", u"parentID": 0, u"memoryUsedPc": 1.15, u"cpuUsedPc": 0, u"id": 0, u"rss": 23888 }, { u"vss": 3247344, u"name": u"RDS processes", u"tgid": 0, u"vmlimit": u"", u"parentID": 0, u"memoryUsedPc": 20.91, u"cpuUsedPc": 0, u"id": 0, u"rss": 429908 } ] }
Mitrofanov/enhanced-rds-monitoring
enhanced_rds/metric_maps.py
<filename>enhanced_rds/metric_maps.py<gh_stars>0 """ The default structure(s) of the metric payloads delivered to the Lambda. Note that the Aurora section only contains the structures in which there is meaningful difference from the standard version. """ # Standard set of metric info METRICS = [ u'cpuUtilization', u'diskIO', u'fileSys', u'loadAverageMinute', u'memory', u'network', u'swap', u'tasks', u'OSprocesses', u'RDSprocesses' ] PROCESS_METRICS = [ u'vss', u'rss', u'memoryUsedPc', u'cpuUsedPc' ] METRICS_DIMS = { u'diskIO': [u'device'], u'fileSys': [u'name', u'mountPoint'], u'network': [u'interface'] } # Metric info for Aurora instances. METRICS_AURORA_DIMS = { u'diskIO': [], # Workaround to account for Aurora diskIO metrics u'fileSys': [u'name', u'mountPoint'], u'network': [u'interface'] } # Metric info for Microsoft SQL instances. METRICS_MICROSOFT = [ u'cpuUtilization', u'disks', u'memory', u'network', u'OSprocesses', u'RDSprocesses', u'system' ] PROCESS_METRICS_MICROSOFT = [ u'cpuUsedPc', u'memUsedPc', u'workingSetKb', u'workingSetPrivKb', u'workingSetShareableKb', u'virtKb' ] METRICS_MICROSOFT_DIMS = { u'disks': [u'name'], u'network': [u'interface'] }
Mitrofanov/enhanced-rds-monitoring
setup.py
#!/usr/bin/env python from setuptools import setup, find_packages with open('requirements.txt') as f: requirements = [line.strip() for line in f.readlines()] setup( name='enhanced_rds', version='0.1.0', description='AWS Lambda function wrapper to capture metrics from enhanced RDS monitoring logs', zip_safe=True, packages=find_packages(), install_requires=requirements, author='<NAME>', author_email='<EMAIL>', license='Apache Software License v2', url='https://github.com/signalfx/enhanced-rds-monitoring' )
krava2020/pong
Ping_Pong.py
import pygame pygame.init() mw_w = 500 mw_h = 500 mw = pygame.display.set_mode((mw_w, mw_h)) mw.fill((0,255,0)) clock = pygame.time.Clock() FPS = 60 class Sprite: def __init__(self, image_name, x, y, width, hight, speed): self.image = pygame.transform.scale( pygame.image.load(image_name), (width, hight)) self.rect = self.image.get_rect() self.rect.x = x self.rect.y = y self.speed = speed def reset(self): mw.blit(self.image, (self.rect.x, self.rect.y)) #qwqwqwqwqwqwqwqwq class Rockets(Sprite): #def __init__(self, image_name, x, y, width, #hight, speed, ball_speed1, ball_speed2): #super().__init__(self, image_name, x, y, width, hight, speed) #self.b_speed1 = ball_speed1 #self.b_speed2 = ball_speed2 def instant_(self): self.spped_x = 3 self.speed_y = 3 def rocket_l(self): key_ = pygame.key.get_pressed() if key_[pygame.K_w] and self.rect.y > 0: self.rect.y -= self.speed if key_[pygame.K_s] and self.rect.y < mw_h - 106: self.rect.y += self.speed def rocket_r(self): key_ = pygame.key.get_pressed() if key_[pygame.K_UP] and self.rect.y > 0: self.rect.y -= self.speed if key_[pygame.K_DOWN] and self.rect.y < mw_h - 106: self.rect.y += self.speed def move(self): speed1 = 3 speed2 = 3 self.rect.x += speed1 self.rect.y += speed2 if self.rect.x > 500: speed1 *= -1 if self.rect.y > mw_h - 26 or self.rect.y < 0: speed2 *= -1 rct_l = Rockets('rc.png', 5, 0, 16, 106, 3) rct_r = Rockets('rc.png', mw_w - 20, 0, 16, 106, 3) ball = Rockets('ball.png', 220, 4, 26, 26, 5) ball.instand_() game = True speed1 = 3 speed2 = 3 while game: for e in pygame.event.get(): if e.type == pygame.QUIT: game = False if game: mw.fill((0,255,0)) rct_l.rocket_l() rct_r.rocket_r() #ball.move() ball.rect.x += speed1 ball.rect.y += speed2 if ball.rect.x > mw_w - 26 or ball.rect.x < 0: speed1 *= -1 if ball.rect.y > mw_h - 26 or ball.rect.y < 0: speed2 *= -1 if pygame.sprite.collide_rect rct_l.reset() rct_r.reset() ball.reset() pygame.display.update() clock.tick(FPS)
mbeyers/NCI-DOE-Collab-Pilot1-Tumor_Classifier
Pilot1/TC1/tc1_infer.py
from __future__ import print_function import pandas as pd import numpy as np import os import sys import gzip import argparse try: import configparser except ImportError: import ConfigParser as configparser from keras import backend as K from keras.layers import Input, Dense, Dropout, Activation, Conv1D, MaxPooling1D, Flatten from keras.optimizers import SGD, Adam, RMSprop from keras.models import Sequential, Model, model_from_json, model_from_yaml from keras.utils import np_utils from keras.callbacks import ModelCheckpoint, CSVLogger, ReduceLROnPlateau from sklearn.metrics import accuracy_score from sklearn.preprocessing import StandardScaler, MinMaxScaler, MaxAbsScaler file_path = os.path.dirname(os.path.realpath(__file__)) lib_path2 = os.path.abspath(os.path.join(file_path, '..', '..', 'common')) sys.path.append(lib_path2) import tc1 as bmk import candle def initialize_parameters(default_model = 'tc1_default_model.txt'): # Build benchmark object tc1Bmk = bmk.BenchmarkTC1(file_path, default_model, 'keras', prog='tc1_baseline', desc='Multi-task (DNN) for data extraction from clinical reports - Pilot 3 Benchmark 1') # Initialize parameters gParameters = candle.finalize_parameters(tc1Bmk) #benchmark.logger.info('Params: {}'.format(gParameters)) return gParameters def run(gParameters): X_train, Y_train, X_test, Y_test = bmk.load_data(gParameters) print('X_test shape:', X_test.shape) print('Y_test shape:', Y_test.shape) # this reshaping is critical for the Conv1D to work X_test = np.expand_dims(X_test, axis=2) print('X_test shape:', X_test.shape) output_dir = gParameters['output_dir'] if not os.path.exists(output_dir): os.makedirs(output_dir) model_name = gParameters['model_name'] # load json and create model json_file = open('{}/{}.model.json'.format(output_dir, model_name), 'r') loaded_model_json = json_file.read() json_file.close() loaded_model_json = model_from_json(loaded_model_json) # load yaml and create model yaml_file = open('{}/{}.model.yaml'.format(output_dir, model_name), 'r') loaded_model_yaml = yaml_file.read() yaml_file.close() loaded_model_yaml = model_from_yaml(loaded_model_yaml) # load weights into new model loaded_model_json.load_weights('{}/{}.model.h5'.format(output_dir, model_name)) print("Loaded json model from disk") # evaluate json loaded model on test data loaded_model_json.compile(loss=gParameters['loss'], optimizer=gParameters['optimizer'], metrics=[gParameters['metrics']]) score_json = loaded_model_json.evaluate(X_test, Y_test, verbose=0) print('json Test score:', score_json[0]) print('json Test accuracy:', score_json[1]) print("json %s: %.2f%%" % (loaded_model_json.metrics_names[1], score_json[1]*100)) # load weights into new model loaded_model_yaml.load_weights('{}/{}.model.h5'.format(output_dir, model_name)) print("Loaded yaml model from disk") # evaluate loaded model on test data loaded_model_yaml.compile(loss=gParameters['loss'], optimizer=gParameters['optimizer'], metrics=[gParameters['metrics']]) score_yaml = loaded_model_yaml.evaluate(X_test, Y_test, verbose=0) print('yaml Test score:', score_yaml[0]) print('yaml Test accuracy:', score_yaml[1]) print("yaml %s: %.2f%%" % (loaded_model_yaml.metrics_names[1], score_yaml[1]*100)) def main(): gParameters = initialize_parameters() run(gParameters) if __name__ == '__main__': main() try: K.clear_session() except AttributeError: # theano does not have this function pass
quantori/nsaph-platform-deployment
project/examples/test-workflow.py
#!/usr/bin/env python3 from cwl_airflow.extensions.cwldag import CWLDAG args = { "cwl": { "debug": True, "parallel": True } } dag = CWLDAG( workflow="/opt/airflow/project/examples/test-workflow.cwl", dag_id="Test-Workflow", default_args=args )
quantori/nsaph-platform-deployment
project/python_sample_project/setup.py
from setuptools import setup, find_packages with open("README.md", "r") as readme: long_description = readme.read() setup( name='pisample', version="0.1", url='', license='Apache 2.0', author='<NAME>', author_email='<EMAIL>', description='Sample Python project for CWL-Airflow', long_description = long_description, long_description_content_type = "text/markdown", py_modules = ['pi'], packages=find_packages(), classifiers=[ "Programming Language :: Python :: 3", "License :: Harvard University :: Development", "Operating System :: OS Independent"] )
quantori/nsaph-platform-deployment
project/examples/medicaid.py
<gh_stars>0 #!/usr/bin/env python3 from cwl_airflow.extensions.cwldag import CWLDAG args = { "cwl": { "debug": True, "parallel": True } } dag = CWLDAG( workflow="/opt/airflow/project/cms/src/cwl/medicaid.cwl", dag_id="medicaid", default_args=args )
quantori/nsaph-platform-deployment
project/examples/1st-tool.py
#!/usr/bin/env python3 from cwl_airflow.extensions.cwldag import CWLDAG args = { "cwl": { "debug": True, "parallel": True } } dag = CWLDAG( workflow="/opt/airflow/project/examples/1st-tool.cwl", dag_id="1st-tool", default_args=args )
quantori/nsaph-platform-deployment
project/python_sample_project/pi.py
<reponame>quantori/nsaph-platform-deployment import sys def calculate(n:int) -> float: k = 1 s = 0 for i in range(n): if i % 2 == 0: s += 4/k else: s -= 4/k k += 2 return s if __name__ == '__main__': print(calculate(int(sys.argv[1])))
quantori/nsaph-platform-deployment
project/examples/airnow.py
<gh_stars>0 #!/usr/bin/env python3 from cwl_airflow.extensions.cwldag import CWLDAG args = { "cwl": { "debug": True, "parallel": True } } dag = CWLDAG( workflow="/opt/airflow/project/epa/src/cwl/airnow.cwl", dag_id="AirNow", default_args=args )
quantori/nsaph-platform-deployment
project/examples/rpi.py
#!/usr/bin/env python3 from cwl_airflow.extensions.cwldag import CWLDAG args = { "cwl": { "debug": True, "parallel": True } } dag = CWLDAG( workflow="/opt/airflow/project/examples/rpi.cwl", dag_id="rpi", default_args=args )
andrewschoen/cephci
tests/ceph_installer/test_ceph_deploy.py
<reponame>andrewschoen/cephci import logging import time from ceph.utils import keep_alive, setup_deb_repos from ceph.utils import setup_repos, create_ceph_conf logger = logging.getLogger(__name__) log = logger def run(**kw): log.info("Running test") ceph_nodes = kw.get('ceph_nodes') config = kw.get('config') if config.get('ubuntu_repo'): ubuntu_repo = config.get('ubuntu_repo') if config.get('base_url'): base_url = config.get('base_url') installer_url = None if config.get('installer_url'): installer_url = config.get('installer_url') if config.get('skip_setup') is True: log.info("Skipping setup of ceph cluster") return 0 ceph1 = ceph_nodes[0] out, _ = ceph1.exec_command(cmd='uuidgen') uuid = out.read().strip().decode() ceph_mon_nodes = [] mon_names = '' all_nodes = '' for ceph in ceph_nodes: if ceph.role == 'mon': ceph_mon_nodes.append(ceph) mon_names = mon_names + ceph.shortname + ' ' all_nodes = all_nodes + ceph.shortname + ' ' ceph_conf = create_ceph_conf(fsid=uuid, mon_hosts=ceph_mon_nodes) keys = '' hosts = '' hostkeycheck = 'Host *\n\tStrictHostKeyChecking no\n\tServerAliveInterval 2400\n' for ceph in ceph_nodes: ceph.generate_id_rsa() keys = keys + ceph.id_rsa_pub hosts = hosts + ceph.ip_address + "\t" + ceph.hostname \ + "\t" + ceph.shortname + "\n" for ceph in ceph_nodes: keys_file = ceph.write_file( file_name='.ssh/authorized_keys', file_mode='a') hosts_file = ceph.write_file( sudo=True, file_name='/etc/hosts', file_mode='a') ceph.exec_command( cmd='[ -f ~/.ssh/config ] && chmod 700 ~/.ssh/config', check_ec=False) ssh_config = ceph.write_file(file_name='.ssh/config', file_mode='w') keys_file.write(keys) hosts_file.write(hosts) ssh_config.write(hostkeycheck) keys_file.flush() hosts_file.flush() ssh_config.flush() ceph.exec_command(cmd='chmod 600 ~/.ssh/authorized_keys') ceph.exec_command(cmd='chmod 400 ~/.ssh/config') for ceph in ceph_nodes: if config.get('use_cdn') is False: if ceph.pkg_type == 'deb': setup_deb_repos(ceph, ubuntu_repo) # install python2 on xenial ceph.exec_command(cmd='sudo apt-get install -y python') else: setup_repos(ceph, base_url, installer_url) else: log.info("Using the cdn repo for the test") log.info("Updating metadata") if ceph.pkg_type == 'rpm': ceph.exec_command(cmd='sudo yum update metadata') ceph1.exec_command(cmd='mkdir cd') ceph1.exec_command(sudo=True, cmd='cd cd; yum install -y ceph-deploy') ceph1.exec_command( cmd='cd cd; ceph-deploy new {mons}'.format(mons=mon_names)) cc = ceph1.write_file(file_name='cd/ceph.conf', file_mode='w') cc.write(ceph_conf) cc.flush() out, err = ceph1.exec_command( cmd='cd cd; ceph-deploy install {all_n}'.format( all_n=all_nodes), timeout=600, check_ec=False) running = True while running: keep_alive(ceph_nodes) log.info("Wait for 120 seconds before next check") time.sleep(120) if out.channel.exit_status_ready(): log.info( "Command completed on remote node %d", out.channel.recv_exit_status()) running = False log.info(out.read().decode()) log.info(err.read().decode()) else: log.info("Command still running") out, err = ceph1.exec_command( cmd='cd cd; ceph-deploy mon create-initial', timeout=300) if ceph1.exit_status != 0: log.error("Failed during mon create-initial") return ceph1.exit_status for cnode in ceph_nodes: if cnode.role == 'osd': hostname = cnode.shortname devices = cnode.no_of_volumes dev = 98 # start with b for vol in range(0, devices): device = hostname + ':' + '/dev/vd' + chr(dev) # --dmcrypt {device} ceph1.exec_command( cmd='cd cd; ceph-deploy osd prepare {device}'.format(device=device), timeout=300) device = hostname + ':' + '/dev/vd' + chr(dev) + '1' ceph1.exec_command( cmd='cd cd; ceph-deploy osd activate {device}'.format(device=device), timeout=60) time.sleep(2) dev = dev + 1 if ceph1.exit_status != 0: log.error("Failed during osd activate") return ceph1.exit_status elif cnode.role == 'client': ceph1.exec_command( cmd='cd cd; ceph-deploy admin ' + cnode.shortname ) return 0
andrewschoen/cephci
tests/ceph_ansible/shrink_mon.py
<reponame>andrewschoen/cephci<filename>tests/ceph_ansible/shrink_mon.py import logging import re log = logging.getLogger(__name__) def run(ceph_cluster, **kw): """ Remove monitor from cluster using shrink-mon.yml Args: ceph_cluster (ceph.ceph.Ceph): ceph cluster Returns: int: non-zero on failure, zero on pass """ log.info("Shrinking monitor") config = kw.get('config') build = config.get('build', config.get('rhbuild')) mon_to_kill_list = config.get('mon-to-kill') mon_to_kill = None mon_short_name_list = [ceph_node.shortname for ceph_node in ceph_cluster.get_nodes('mon')] for _mon_to_kill in mon_to_kill_list: matcher = re.compile(_mon_to_kill) matched_short_names = list(filter(matcher.match, mon_short_name_list)) if len(matched_short_names) > 0: shrinked_nodes = [ceph_node for ceph_node in ceph_cluster if ceph_node.shortname in matched_short_names] for ceph_node in shrinked_nodes: ceph_node.remove_ceph_object(ceph_node.get_ceph_objects('mon')[0]) else: raise RuntimeError('No match for {node_name}'.format(node_name=_mon_to_kill)) mon_to_kill = ','.join([mon_to_kill, ','.join(matched_short_names)]) if mon_to_kill else ','.join( matched_short_names) ceph_installer = ceph_cluster.get_ceph_object('installer') ceph_installer.node.obtain_root_permissions('/var/log') ansible_dir = ceph_installer.ansible_dir ceph_installer.exec_command(sudo=True, cmd='cp -R {ansible_dir}/infrastructure-playbooks/shrink-mon.yml ' '{ansible_dir}/shrink-mon.yml'.format(ansible_dir=ansible_dir)) out, err = ceph_installer.exec_command(cmd='export ANSIBLE_DEPRECATION_WARNINGS=False ; cd {ansible_dir} ; ' 'ansible-playbook -vvvv -e ireallymeanit=yes shrink-mon.yml ' '-e mon_to_kill={mon_to_kill} -i hosts'.format(ansible_dir=ansible_dir, mon_to_kill=mon_to_kill), long_running=True) if err != 0: log.error("Failed during ansible playbook run") return err return ceph_cluster.check_health(build)
slimpotatoes/STEM_Moire_GPA
src/gpa.py
<reponame>slimpotatoes/STEM_Moire_GPA # GPA Module import mask as mask import data as data import numpy as np from skimage.restoration import unwrap_phase def gpa(mask_id, datastruct): """Pefrom GPA on the GUI id, mask_id using the data in datastructure: 1. Load elements of GUI (center, radius) to create the mask 2. Load the Fourier transform of the ISMHexp (element to mask) 3. Create the mask in the image space 4. Store the approximation of the unstrain reference as the center of the circle g_uns 5. Mask the Fourier transform of ISMHexp 6. Calculate the phase of the masked section of ISMHexp 7. Calculate the g_M vector by taking the gradient of the phase 8. Calculate the variation of the g_M vector as the difference of g_M and g_uns 9. Calculate the phase corrected by removing the contribution of g_uns to be only related to delta_g 10. Store delta_g and the phase related to delta_g""" # Load the elements center = data.SMGData.load_g(datastruct, mask_id, 'Mask')[0] r = data.SMGData.load_g(datastruct, mask_id, 'Mask')[1] ft_ismh_exp = data.SMGData.load(datastruct, 'FTISMHexp') # Generate the mask in the image space m, g_uns = mask.mask_gaussian(center, r, ft_ismh_exp.shape) # Store the unstrain reference in the datastructure data.SMGData.store_g(datastruct, mask_id, 'gMuns', g_uns) # Mask and calculate the phase component masked_ft_ismh_exp = np.multiply(m, np.fft.fftshift(ft_ismh_exp)) phase_g_m = np.angle(np.fft.ifft2(np.fft.ifftshift(masked_ft_ismh_exp))) data.SMGData.store_g(datastruct, mask_id, 'phaseraw', phase_g_m) # Calculate g_m and the variation of g_M g_m = np.array([1 / (2 * np.pi) * np.gradient(unwrap_phase(phase_g_m))[0], 1 / (2 * np.pi) * np.gradient(unwrap_phase(phase_g_m))[1]]) delta_g_m = np.subtract(g_m, g_uns) # Calculate phase corrected by removing the contribution of g_uns to be only related to delta_g mesh_x, mesh_y = np.meshgrid(np.arange(phase_g_m.shape[0]), np.arange(phase_g_m.shape[1])) unwrapped_phase_delta_g_m = np.array(unwrap_phase(phase_g_m)) - 2 * np.pi * ( np.multiply(g_uns[1], mesh_x) + np.multiply(g_uns[0], mesh_y)) phase_delta_g_m = unwrapped_phase_delta_g_m - np.round(unwrapped_phase_delta_g_m / (2 * np.pi)) * 2 * np.pi # Store the final data data.SMGData.store_g(datastruct, mask_id, 'deltagM', delta_g_m) data.SMGData.store_g(datastruct, mask_id, 'phasegM', phase_delta_g_m)
slimpotatoes/STEM_Moire_GPA
src/userinput.py
# Input Module import dm3_lib as dm3_lib import numpy as np import data as data def load_files(file_path_smh, file_path_ic, datastruct): """Load the files by asking the controller the filepaths smh and ic provided by the user and store the appropriate data in the data structure object (dictionary). Before storing, the data are verified""" dm3_meta_smh = dm3_lib.DM3(file_path_smh) dm3_meta_ic = dm3_lib.DM3(file_path_ic) verify_i(dm3_meta_smh.imagedata, dm3_meta_ic.imagedata) pixel_smh = dm3_meta_smh.pxsize pixel_ic = dm3_meta_ic.pxsize verify_p(pixel_smh[0], pixel_ic[0]) verify_p_unit(pixel_smh[1].decode("ascii"), pixel_ic[1].decode("ascii")) data.SMGData.store(datastruct, 'ISMHexp', dm3_meta_smh.imagedata) data.SMGData.store(datastruct, 'p', pixel_smh[0]) data.SMGData.store(datastruct, 'ICref', dm3_meta_ic.imagedata) data.SMGData.store(datastruct, 'pref', pixel_ic[0]) print('Files loaded') print('Pixel size SMH: ', pixel_smh[0], 'nm') print('Pixel size Reference: ', pixel_ic[0], 'nm') def verify_i(ismh, ic): """Verify if the input images ismh and ic are 2D arrays of real numbers""" for value in np.nditer(np.isreal(ismh)): if value is False: raise Exception('The STEM Moire hologram is not composed of real numbers.') for value in np.nditer(np.isreal(ic)): if value is False: raise Exception('The Reference Image is not composed of real numbers.') def verify_p(p, pref): """Verify if the pixel size input p and pref are real numbers strictly positive and if there are different.""" if p < 0 or isinstance(p, float) is False: raise Exception('The pixel size of the STEM Moire hologram is not a real number strictly positive.') if pref < 0 or isinstance(pref, float) is False: raise Exception('The pixel size of the Reference image is not a real number strictly positive.') if p == pref: print('The pixel size is the same for the STEM Moire hologram and the Reference image. The software ' 'is now running in classic HRSTEM GPA mode.') # Improve it with a raise warning but it doesn't continue def verify_p_unit(unit_p, unit_pref): """Verify if the unit of the pixel size inputs unit_p and unit_pref are in nm""" if unit_p != 'nm': raise Exception('The pixel size of the STEM Moire hologram is not in nanometer.') if unit_pref != 'nm': raise Exception('The pixel size of the Reference image is not in nanometer.') if unit_p != unit_pref: raise Exception('The units used for the pixel size between the two images are different.')
slimpotatoes/STEM_Moire_GPA
src/test_mask.py
import pytest import mask as mask import numpy as np # Test cases test_case_0 = dict() test_case_0['center'] = (2, 3) test_case_0['radius'] = 1 test_case_0['image_test'] = np.ones(shape=(6, 6)) test_case_0['image_to_assert'] = np.array([[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 1, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0]]) test_case_0['g0'] = np.array([0, -1/6]) test_case_1 = dict() test_case_1['center'] = (2, 3) test_case_1['radius'] = 1 test_case_1['image_test'] = np.ones(shape=(5, 5)) test_case_1['image_to_assert'] = np.array([[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 1, 0, 0], [0, 0, 0, 0, 0]]) test_case_1['g0'] = np.array([1/10, -1/10]) test_case_2 = dict() test_case_2['center'] = (3, 2) test_case_2['radius'] = 1 test_case_2['image_test'] = np.ones(shape=(6, 6)) test_case_2['image_to_assert'] = np.array([[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0]]) test_case_2['g0'] = np.array([-1/6, 0]) test_case_3 = dict() test_case_3['center'] = (4, 3) test_case_3['radius'] = 1 test_case_3['image_test'] = np.ones(shape=(6, 6)) test_case_3['image_to_assert'] = np.array([[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 1, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0]]) test_case_3['g0'] = np.array([0, 1/6]) test_case_4 = dict() test_case_4['center'] = (3, 4) test_case_4['radius'] = 1 test_case_4['image_test'] = np.ones(shape=(6, 6)) test_case_4['image_to_assert'] = np.array([[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 1, 0, 0], [0, 0, 0, 0, 0, 0]]) test_case_4['g0'] = np.array([1/6, 0]) test_cases = [test_case_0, test_case_1, test_case_2, test_case_3, test_case_4] # ------------------------------------------------- # Test of classic mask on test cases # ------------------------------------------------- # ------------------------------------------------- # Test of gaussian mask checking position of center and g_0 on test cases # ------------------------------------------------- '''@pytest.mark.parametrize('image, image_g', image_test_after_classic_mask) def test_gaussian_mask(image, image_g): image_masked = mask.mask_gaussian(center, radius, image.shape) assert(image_masked[0][3, 2] == 1) assert(np.all(image_masked[1] == image_g))''' @pytest.mark.parametrize("test_case", test_cases) def test_classic_masking(test_case): image_masked = np.array(mask.mask_classic(test_case['center'], test_case['radius'], np.shape(test_case['image_test']))[0]) #print(image_masked) #print(test_case['image_to_assert']) assert np.all(image_masked==test_case['image_to_assert']) @pytest.mark.parametrize("test_case", test_cases) def test_classic_g0(test_case): image_masked = np.array(mask.mask_classic(test_case['center'], test_case['radius'], np.shape(test_case['image_test']))[1][:,0,0]) #print(image_masked) #print(test_case['g0']) assert np.all(image_masked==test_case['g0']) @pytest.mark.parametrize("test_case", test_cases) def test_gaussian_g0(test_case): image_masked = np.array(mask.mask_gaussian(test_case['center'], test_case['radius'], np.shape(test_case['image_test']))[1][:,0,0]) #print(image_masked) #print(test_case['g0']) assert np.all(image_masked==test_case['g0'])
slimpotatoes/STEM_Moire_GPA
src/rotatecalc.py
import numpy as np import math import data def angle_rotation(gui): first_position = [gui.line_rot.LineCoords[0][1], gui.line_rot.LineCoords[0][0]] second_position = [gui.line_rot.LineCoords[1][1], gui.line_rot.LineCoords[1][0]] print('first position = ', first_position) print('second position = ', second_position) first_point = [first_position[1], -first_position[0]] second_point = [second_position[1], -second_position[0]] print('first point = ', first_point) print('second point = ', second_point) # x1 = math.sqrt((second_point[0] - first_point[0]) ** 2) # x2 = math.sqrt((second_point[1] - first_point[1]) ** 2) norm = math.sqrt((second_point[0] - first_point[0]) ** 2 + (second_point[1] - first_point[1]) ** 2) horizontal_proj = (second_point[0] - first_point[0]) / norm vertical_proj = (second_point[1] - first_point[1]) / norm if vertical_proj >= 0: theta = np.arccos(horizontal_proj) else: theta = - np.arccos(horizontal_proj) print('theta = ', theta) return theta def rotate_tensor(datastruct, gui): theta = angle_rotation(gui) exx = data.SMGData.load(datastruct, 'Exx') eyy = data.SMGData.load(datastruct, 'Eyy') exy = data.SMGData.load(datastruct, 'Exy') rxy = data.SMGData.load(datastruct, 'Rxy') epsilon = np.array([[exx, exy], [exy, eyy]]) omega = np.array([[np.zeros(rxy.shape), rxy], [-rxy, np.zeros(rxy.shape)]]) r = np.array([[np.cos(theta), -np.sin(theta)], [np.sin(theta), np.cos(theta)]]) epsilon = np.transpose(epsilon, axes=[2, 3, 0, 1]) omega = np.transpose(omega, axes=[2, 3, 0, 1]) for i in range(0, epsilon[:, :, 0, 0].shape[0]): for j in range(0, epsilon[:, :, 0, 0].shape[1]): epsilon[i, j] = np.dot(np.transpose(r), np.dot(epsilon[i, j], r)) omega[i, j] = np.dot(np.transpose(r), np.dot(omega[i, j], r)) epsilon = np.transpose(epsilon, axes=[2, 3, 0, 1]) omega = np.transpose(omega, axes=[2, 3, 0, 1]) data.SMGData.store(datastruct, 'Exx', epsilon[0, 0]) data.SMGData.store(datastruct, 'Eyy', epsilon[1, 1]) data.SMGData.store(datastruct, 'Exy', epsilon[1, 0]) data.SMGData.store(datastruct, 'Rxy', omega[1, 0])
slimpotatoes/STEM_Moire_GPA
src/test_conversion.py
<reponame>slimpotatoes/STEM_Moire_GPA import pytest import data import numpy as np import conversion mask = 'Mask1' # Create elements used by conversion in the data structure # ------------------------------- # Test cases # ------------------------------- # Test 24 in TestPlan.pdf test_case_0 = dict() datastruct_0 = data.SMGData() datastruct_0.create_branch(mask) datastruct_0.store('p', 1/3) datastruct_0.store_g(mask, 'gMuns', np.transpose(np.array([[[1, -1]]]), axes=(2, 0, 1))) test_case_0['datastruct'] = datastruct_0 test_case_0['n_horizontal'] = 1 test_case_0['m_vertical'] = 2 test_case_0['to_assert'] = np.transpose(np.array([[[-1, 0]]]), axes=(2, 0, 1)) test_case_1 = dict() datastruct_1 = data.SMGData() datastruct_1.create_branch(mask) datastruct_1.store('p', 1/3) datastruct_1.store_g(mask, 'gMuns', np.transpose(np.array([[[1, -1]]]), axes=(2, 0, 1))) test_case_1['datastruct'] = datastruct_1 test_case_1['n_horizontal'] = -1 test_case_1['m_vertical'] = -2 test_case_1['to_assert'] = np.transpose(np.array([[[3, -2]]]), axes=(2, 0, 1)) test_case_2 = dict() datastruct_2 = data.SMGData() datastruct_2.create_branch(mask) datastruct_2.store('p', 1/3) datastruct_2.store_g(mask, 'gMuns', np.transpose(np.array([[[1, 0]]]), axes=(2, 0, 1))) test_case_2['datastruct'] = datastruct_2 test_case_2['n_horizontal'] = 0 test_case_2['m_vertical'] = -2 test_case_2['to_assert'] = np.transpose(np.array([[[3, 0]]]), axes=(2, 0, 1)) test_case_3 = dict() datastruct_3 = data.SMGData() datastruct_3.create_branch(mask) datastruct_3.store('p', 1/3) datastruct_3.store_g(mask, 'gMuns', np.transpose(np.array([[[-1, 0]]]), axes=(2, 0, 1))) test_case_3['datastruct'] = datastruct_3 test_case_3['n_horizontal'] = 0 test_case_3['m_vertical'] = 2 test_case_3['to_assert'] = np.transpose(np.array([[[-3, 0]]]), axes=(2, 0, 1)) test_cases = [test_case_0, test_case_1, test_case_2, test_case_3] @pytest.mark.parametrize("test_case", test_cases) def test_conversion(test_case): conversion.conversion(mask, test_case['n_horizontal'], test_case['m_vertical'], test_case['datastruct']) assert np.all(np.array(test_case['datastruct'].load_g(mask, 'gCuns')) == test_case['to_assert'])
slimpotatoes/STEM_Moire_GPA
src/test_data.py
""" Tests of the SMGData class from data.py """ import pytest import data # Test setup # SMGData instance smgData = data.SMGData() smgData.create_branch('Mask1') smgData.create_branch('Mask2') # This definition should have been provided by the SMGData interface possibleKeys = [ "ICref", "pref", "FTISMHexp", "FTISMHsim", "FTISMHsimDisplay", "Uref", "Exx", "Eyy", "Exy", "Rxy", ] # This definition should have been provided by the SMGData interface possibleKeysOfGuiIds = [ "Mask", "gMuns", "deltagM", "phasegM", "shiftg", "gCuns", ] possibleGuiIds = [ "Mask1", "Mask2", ] # ------------------------------------------------- # Test of positive (OK) cases of data stored # ------------------------------------------------- @pytest.mark.parametrize("key", possibleKeys) def test_data_store_load_ok(key): smgData.store(key, key) assert(smgData.SMGData[key] == key) # ------------------------------------------------- # Test of positive (OK) cases of data stored in an existing branch # ------------------------------------------------- @pytest.mark.parametrize("key", possibleKeysOfGuiIds) @pytest.mark.parametrize("gui_id", possibleGuiIds) def test_data_store_load_g_ok(gui_id, key): smgData.store_g(gui_id, key, key) assert(smgData.SMGData[gui_id][key] == key) assert(smgData.load_g(gui_id, key) == key) # ------------------------------------------------- # Test of positive cases of data to load # ------------------------------------------------- @pytest.mark.parametrize("key", possibleKeys) def test_data_store_load_ok(key): smgData.store(key, key) assert (smgData.load(key) == key) # ------------------------------------------------- # Test of positive cases of data to load from an existing branch # ------------------------------------------------- @pytest.mark.parametrize("key", possibleKeysOfGuiIds) @pytest.mark.parametrize("gui_id", possibleGuiIds) def test_data_store_load_g_ok(gui_id, key): smgData.store_g(gui_id, key, key) assert (smgData.load_g(gui_id, key) == key) # ------------------------------------------------- # Test of positive cases of non-string data to store # ------------------------------------------------- def test_data_save_load_non_string_ok(): smgData.store(possibleKeys[0], 1) # an integer assert(smgData.load(possibleKeys[0]) == 1) smgData.store(possibleKeys[1], 1.25) # a float assert(smgData.load(possibleKeys[1]) == 1.25) arr_len = 100000 # a big array my_array = [None] * arr_len for i in range(0, arr_len): my_array[i] = i smgData.store(possibleKeys[2], my_array) my_array2 = smgData.load(possibleKeys[2]) assert(my_array == my_array2) # ------------------------------------------------- # Test of negative (NOK) cases of data save and load # ------------------------------------------------- def test_data_store_nok_illegal_key(): with pytest.raises(Exception) as e_info: smgData.store('an illegal key', 0) assert(str(e_info.value) == 'Key does not exist in the data structure') def test_data_load_nok_illegal_key(): with pytest.raises(Exception) as e_info: smgData.load('an illegal key') assert(str(e_info.value) == 'Key does not exist in the data structure')
slimpotatoes/STEM_Moire_GPA
src/conversion.py
# Conversion Module import numpy as np import data as data import math as math def conversion(mask_id, n_horizontal, m_vertical, datastruct): """Convert the Moire 3D array into the Crystal 2D array and store it in datastruct. The Moire 3D array is loaded from the data structure (datastruct) using the keyword (mask_id = string). The horizontal and vertical component of the Moire 3D array are separated and each component are converted using the integer n_horizontal and m_vertical and the pixel size (strictly positive real number loaded from datastuct).""" # Load the pixel size (p) from the data structure and check if p is strictly positive p = data.SMGData.load(datastruct, 'p') if p <= 0: raise Exception('Pixel size negative or zero, conversion cannot be performed') # Normalize the unstrained reference Moire 3D array (gMuns = 3D array -- 2D vector on each pixel of a # 2D image and separate components) g_m_uns = data.SMGData.load_g(datastruct, mask_id, 'gMuns') # Generate the correction 3D array to apply on the unstrained reference Moire 3D array on each component correction = np.ones(g_m_uns.shape) # Warning g[0] component along x (vertical axis pointing down) correction[0, :, :] = - m_vertical * correction[0, :, :] # Warning g[1] component along y (horizontal axis pointing right) correction[1, :, :] = n_horizontal * correction[1, :, :] # Apply correction to get the unstrained reference crystalline 3D array and store it in the data structure g_c_uns = g_m_uns + correction data.SMGData.store_g(datastruct, mask_id, 'gCuns', g_c_uns) # Inform user of the completion of the conversion and provide the norm of the crystalline wave vector norm = 1 / p * math.sqrt(g_c_uns[0, 0, 0] ** 2 + g_c_uns[1, 0, 0] ** 2) print('Conversion done !!') print('g norm = ', norm, ' nm-1')
slimpotatoes/STEM_Moire_GPA
src/main.py
<gh_stars>1-10 # ############################# # # # # STEM Moire GPA Software # # # # ############################# # # # ##################################################################################### # # Python script calculating the 2D relative strain maps from a STEM Moire hologram. # <NAME> <<EMAIL>> # https://github.com/slimpotatoes/STEM_Moire_GPA # 18/06/2019 # # ########################################### # STEM Moire GPA control Module import matplotlib.pyplot as plt import gui as gui import data as data import userinput as userinput import smhsimulation as smhsimu import gpa as gpa import unstrainref as uref import conversion as conversion import straincalc as strain import rotatecalc as rotate def main(): """Connection of the different events (button clicked by user) with the process steps of STEM Moire GPA processing""" def flow_input(event): """Input Process 1. Call the GUI to create an open file dialog for the user to input files. 2. Verify and import files in smgdata. 3. Display the SMH and ICref images to the user.""" if not event.inaxes == smggui.event_input.ax: raise Exception('Improper input axis') file_path_smh, file_path_ic = smggui.open_files() userinput.load_files(file_path_smh, file_path_ic, smgdata) smggui.guismhexp(smgdata) def flow_smhsim(event): """Simulation of the STEM Moire hologram Process 1. Call smh_sim function in smhsimulation module to simulate the STEM Moire hologram from ICref and store the results in smgdata. 2. Display the results of the simulation to the user using guismhsim window.""" if not event.inaxes == smggui.event_smhsim.ax: raise Exception('Improper shmsim axis') smhsimu.smh_sim(smgdata) smggui.guismhsim(smgdata) def flow_gpa(event): """Geometrical Phase Analysis Process 1. Collect the mask selected by the user on the guismhsim window. 2. Perform the GPA calculation on the selected mask and store the results in smgdata. 3. Display the GPA result (phase image) to the user using guiphase window.""" if not event.inaxes == smggui.event_gpa.ax: raise Exception('Improper gpa axis') mask_selected = smggui.mask_selection() gpa.gpa(mask_selected, smgdata) smggui.guiphase(mask_selected, smgdata) def flow_ref(event): """Unstrained reference definition Process. On the two phase images calculated by GPA, 1. Call the update_zerostrain function in unstrainref module to update the unstrain reference on the phase image and store the results in smgdata. 2. Display the updated phase image with the new unstrained reference on the guiphase window.""" if not event.inaxes == smggui.event_ref.ax: raise Exception('Improper ref axis') for mask_id in ['Mask1', 'Mask2']: uref.update_zerostrain(mask_id, smgdata) smggui.update_phase(mask_id, smgdata) def flow_convert(event): """Moire to crystal data conversion Process. Call the conversion function in the conversion module for both masks.""" if not event.inaxes == smggui.event_convert.ax: raise Exception('Improper convert axis') print(smggui.h_1, smggui.h_2, smggui.v_1, smggui.v_2) conversion.conversion('Mask1', smggui.h_1, smggui.v_1, smgdata) conversion.conversion('Mask2', smggui.h_2, smggui.v_2, smgdata) def flow_strain(event): """Strain tensor calculation from two non collinear crystalline wave vector Process""" if not event.inaxes == smggui.event_strain.ax: raise Exception('Improper strain axis') strain.strain_calculation('Mask1', 'Mask2', smgdata) rotate.rotate_tensor(smgdata, smggui) smggui.guistrain(smgdata) """Creation of the GUI and the Data object""" smgdata = data.SMGData() smggui = gui.SMGGUI(smgdata) """Call of the GUI module functions to pop up the initial windows for the user""" smggui.guiconv() smggui.guiflow() """Connection of the event "button clicked by the user" to a function""" smggui.event_input.on_clicked(flow_input) smggui.event_smhsim.on_clicked(flow_smhsim) smggui.event_gpa.on_clicked(flow_gpa) smggui.event_ref.on_clicked(flow_ref) smggui.event_convert.on_clicked(flow_convert) smggui.event_strain.on_clicked(flow_strain) plt.show() if __name__ == "__main__": main()
slimpotatoes/STEM_Moire_GPA
src/data.py
<filename>src/data.py<gh_stars>1-10 # Data Structure Module class SMGData(object): """Module storing and loading intermediate data used during the processing""" def __init__(self): self.SMGData = dict() self.SMGData['ISMHexp'] = None self.SMGData['p'] = None self.SMGData['ICref'] = None self.SMGData['pref'] = None self.SMGData['FTISMHexp'] = None self.SMGData['FTISMHsim'] = None self.SMGData['FTISMHsimDisplay'] = None self.SMGData['Uref'] = None self.SMGData['Exx'] = None self.SMGData['Eyy'] = None self.SMGData['Exy'] = None self.SMGData['Rxy'] = None def create_branch(self, gui_id): """Create a (sub)dictionary in dictionary SMGData associated with the string key gui_id representing the id of a the mask GUI object.""" if gui_id not in self.SMGData.keys(): self.SMGData[gui_id] = dict() self.SMGData[gui_id]['Mask'] = None self.SMGData[gui_id]['gMuns'] = None self.SMGData[gui_id]['phaseraw'] = None self.SMGData[gui_id]['deltagM'] = None self.SMGData[gui_id]['phasegM'] = None self.SMGData[gui_id]['shiftg'] = None self.SMGData[gui_id]['gCuns'] = None else: raise Exception('Key gui_id already exists, branch creation aborted') def remove_branch(self, gui_id): """Remove the (sub)dictionary in dictionary SMGData associated with the string key gui_id representing the id of a the mask GUI object.""" if gui_id in self.SMGData.keys(): del self.SMGData[gui_id] else: return # raise Warning('Key gui_id does not exist, the deletion process did not occur') def store(self, key, a): """Store in dictionary SMGData an object a associated with the string key.""" if key in self.SMGData.keys(): self.SMGData[key] = a else: raise Exception('Key does not exist in the data structure') def load(self, key): """Return from dictionary SMGData the object associated with the string key.""" if key in self.SMGData.keys(): return self.SMGData[key] else: raise Exception('Key does not exist in the data structure') def store_g(self, gui_id, key, a): """Store in (sub)dictionary SMGData[gui_id] an object a associated with the string key.""" if gui_id in self.SMGData.keys(): if key in self.SMGData[gui_id].keys(): self.SMGData[gui_id][key] = a else: raise Exception('Key is gui_id does not exist in the data structure') else: raise Exception('Gui id does not exist in the data structure') def load_g(self, gui_id, key): """Return from dictionary SMGData[gui_id] the object associated with the string key.""" if gui_id in self.SMGData.keys(): if key in self.SMGData[gui_id].keys(): return self.SMGData[gui_id][key] else: raise Exception('Key is gui_id does not exist in the data structure') else: raise Exception('Gui id does not exist in the data structure')
slimpotatoes/STEM_Moire_GPA
src/manual_test_gpa.py
<reponame>slimpotatoes/STEM_Moire_GPA # Test of GPA Module import numpy as np import math as math import data as data import gpa as gpa import matplotlib.pyplot as plt import statistics # ####################################### # Test #2 in TestPlan document # ####################################### data_test_2 = data.SMGData() # Data structure for test 2 # Generated STEM Moire hologram (256x256 pixels) using numpy arrays with a unique periodicity of 16 pixels along # horizontal axis. (Very easy test case) x2 = np.linspace(0, 255, 256) y2 = np.linspace(0, 255, 256) mx2, my2 = np.meshgrid(x2, y2) ismh2 = np.sin(mx2 * 2 * np.pi / 16) ft_ismh2 = np.fft.fft2(ismh2) # Store data in datastructure data_test_2.store('ISMHexp', ismh2) data_test_2.store('FTISMHexp', ft_ismh2) # Create branch for mask data_test_2.create_branch('Mask1') # Circle of radius 1 centered around coordinate (192, 128) r2 = 1 center2 = (144, 128) circle2 = center2, r2 # Store mask properties into datastructure data_test_2.store_g('Mask1', 'Mask', circle2) # Entering gpa gpa.gpa('Mask1', data_test_2) # Display data # Input/Output data - Phase is supposed to be constant and equal to 0, anything different from 0 represents the error. fig_test2 = plt.figure(figsize=(13, 9)) fig_test2_ax1 = fig_test2.add_subplot(2, 3, 1) fig_test2_ax2 = fig_test2.add_subplot(2, 3, 4) fig_test2_ax3 = fig_test2.add_subplot(2, 3, 2) fig_test2_ax4 = fig_test2.add_subplot(2, 3, 5) fig_test2_ax5 = fig_test2.add_subplot(2, 3, 3) fig_test2_ax6 = fig_test2.add_subplot(2, 3, 6) fig_test2_ax1.imshow(ismh2, cmap='gray') fig_test2_ax1.set_title('I_SMH') fig_test2_ax2.imshow(np.log1p(np.fft.fftshift(np.abs(ft_ismh2 ** 2))), cmap='gray') fig_test2_ax2.set_title('Fourier Transform of I_SMH') fig_test2_ax3.imshow(data_test_2.load_g('Mask1', 'phaseraw'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_test2_ax3.set_title('Raw Phase') fig_test2_ax4.imshow(data_test_2.load_g('Mask1', 'phasegM'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_test2_ax4.set_title('Phase corrected') fig_test2_ax5.imshow(data_test_2.load_g('Mask1', 'deltagM')[0], cmap='gray', vmin=-1, vmax=1) fig_test2_ax5.set_title('Vertical component of Δg') fig_test2_ax6.imshow(data_test_2.load_g('Mask1', 'deltagM')[1], cmap='gray', vmin=-1, vmax=1) fig_test2_ax6.set_title('Horizontal component of Δg') fig_test2.savefig('/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/' 'Test_2_explanation.png', dpi=300, bbox_inches='tight') # Input/Output data in 1D fig_test2_1d = plt.figure(figsize=(13, 9)) fig_test2_1d_ax1 = fig_test2_1d.add_subplot(2, 3, 1) fig_test2_1d_ax2 = fig_test2_1d.add_subplot(2, 3, 4) fig_test2_1d_ax3 = fig_test2_1d.add_subplot(2, 3, 2) fig_test2_1d_ax4 = fig_test2_1d.add_subplot(2, 3, 5) fig_test2_1d_ax5 = fig_test2_1d.add_subplot(2, 3, 3) fig_test2_1d_ax6 = fig_test2_1d.add_subplot(2, 3, 6) fig_test2_1d_ax1.plot(ismh2[128, :]) fig_test2_1d_ax1.set_title('I_SMH') fig_test2_1d_ax2.plot(x2, np.log1p(np.fft.fftshift(np.abs(ft_ismh2 ** 2)))[128, :]) fig_test2_1d_ax2.set_title('Fourier Transform of I_SMH') fig_test2_1d_ax3.plot(x2, data_test_2.load_g('Mask1', 'phaseraw')[128, :]) fig_test2_1d_ax3.set_ylim(-np.pi, np.pi) fig_test2_1d_ax3.set_title('Raw Phase') fig_test2_1d_ax4.plot(x2, data_test_2.load_g('Mask1', 'phasegM')[128, :]) fig_test2_1d_ax4.set_ylim(-np.pi, np.pi) fig_test2_1d_ax4.set_title('Phase corrected') fig_test2_1d_ax5.plot(x2, data_test_2.load_g('Mask1', 'deltagM')[0, 128, :]) fig_test2_1d_ax5.set_ylim(-1, 1) fig_test2_1d_ax5.set_title('Vertical component of Δg') fig_test2_1d_ax6.plot(x2, data_test_2.load_g('Mask1', 'deltagM')[1, 128, :]) fig_test2_1d_ax6.set_ylim(-1, 1) fig_test2_1d_ax6.set_title('Horizontal component of Δg') fig_test2_1d.savefig('/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/' 'Test_2_explanation_1D.png', dpi=300, bbox_inches='tight') plt.show() # -------------------- Test #2 Improvement --------------- # periodicity in pixels: q q = [3, 4, 4.1, 4.2, 4.5, 16, 100] data2 = [] for periodicity in q: ismh = np.sin(mx2 * 2 * np.pi / periodicity) ft_ismh = np.fft.fft2(ismh) circle = (128 + round(256 / periodicity), 128), r2 data_i = data.SMGData() data_i.store('ISMHexp', ismh) data_i.store('FTISMHexp', ft_ismh) data_i.create_branch('Mask1') data_i.store_g('Mask1', 'Mask', circle) gpa.gpa('Mask1', data_i) data2.append(data_i) fig_test2_multiple_q = plt.figure(figsize=(13, 6)) fig_test2_multiple_q_ax1 = fig_test2_multiple_q.add_subplot(1, 2, 1) fig_test2_multiple_q_ax2 = fig_test2_multiple_q.add_subplot(1, 2, 2) fig_test2_multiple_q_ismh = plt.figure(figsize=(13, 6)) fig_test2_multiple_q_ismh_ax1 = fig_test2_multiple_q_ismh.add_subplot(1, 2, 1) fig_test2_multiple_q_ismh_ax2 = fig_test2_multiple_q_ismh.add_subplot(1, 2, 2) fig_test2_error_delta_g = plt.figure(figsize=(13, 9)) fig_test2_error_delta_g_ax = fig_test2_error_delta_g.add_subplot(1, 1, 1) count = 0 mean_test_2 = [] stand_dev_test_2 = [] for elements in data2: fig_test2_multiple_q_ax1.plot(x2, elements.load_g('Mask1', 'phasegM')[128, :], linewidth=3, label=str(q[count])) fig_test2_multiple_q_ax2.plot(x2, elements.load_g('Mask1', 'deltagM')[1, 128, :], linewidth=3, label=str(q[count])) fig_test2_multiple_q_ismh_ax1.plot(x2[160:180], elements.load('ISMHexp')[128, 160:180], linewidth=3, label=str(q[count])) fig_test2_multiple_q_ismh_ax2.plot(x2, np.log1p(np.fft.fftshift(np.abs(elements.load('FTISMHexp') ** 2)))[128, :], linewidth=3, label=str(q[count])) error_delta_g = np.abs(elements.load_g('Mask1', 'deltagM')[1, 128, :]) fig_test2_error_delta_g_ax.plot(x2, error_delta_g, linewidth=3, label=str(q[count])) mean_test_2.append(statistics.mean(elements.load_g('Mask1', 'deltagM')[1, 128, :])) stand_dev_test_2.append(statistics.stdev(elements.load_g('Mask1', 'deltagM')[1, 128, :])) count += 1 fig_test2_multiple_q_ax1.set_ylim(-np.pi, np.pi) fig_test2_multiple_q_ax1.legend() fig_test2_multiple_q_ax1.set_title('Phase corrected') fig_test2_multiple_q_ax2.set_title('Horizontal component of Δg') fig_test2_multiple_q_ax2.set_ylim(-0.01, 0.01) fig_test2_multiple_q_ismh_ax1.set_title('I_SMH') fig_test2_multiple_q_ismh_ax2.set_title('Fourier Transform of I_SMH') fig_test2_multiple_q_ismh_ax2.legend() fig_test2_error_delta_g_ax.set_ylim(0, 0.0025) fig_test2_error_delta_g_ax.legend() fig_test2_error_delta_g_ax.set_title('Error of the horizontal component of Δg ') fig_test2_multiple_q.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_2_test_results.png', dpi=300, bbox_inches='tight') fig_test2_multiple_q_ismh.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_2_test_cases.png', dpi=300, bbox_inches='tight') fig_test2_error_delta_g.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_2_test_error.png', dpi=300, bbox_inches='tight') plt.show() print('Mean test 2 = ', mean_test_2) print('StDev test 2 = ', stand_dev_test_2) # -------------------- Testing the vertical direction ------------------- data_test_2_v = data.SMGData() # Data structure for test 2 # Generated STEM Moire hologram (256x256 pixels) using numpy arrays with a unique periodicity of 16 pixels along # horizontal axis. (Very easy test case) x2_v = np.linspace(0, 255, 256) y2_v = np.linspace(0, 255, 256) mx2_v, my2_v = np.meshgrid(x2_v, y2_v) ismh2_v = np.sin(my2_v * 2 * np.pi / 16) ft_ismh2_v = np.fft.fft2(ismh2_v) # Store data in datastructure data_test_2_v.store('ISMHexp', ismh2_v) data_test_2_v.store('FTISMHexp', ft_ismh2_v) # Create branch for mask data_test_2_v.create_branch('Mask1') # Circle of radius 1 centered around coordinate (192, 128) r2_v = 1 center2_v = (128, 144) circle2_v = center2_v, r2_v # Store mask properties into datastructure data_test_2_v.store_g('Mask1', 'Mask', circle2_v) # Entering gpa gpa.gpa('Mask1', data_test_2_v) # Display data # Input/Output data - Phase is supposed to be constant and equal to 0, anything different from 0 represents the error. fig_test2_v = plt.figure(figsize=(13, 9)) fig_test2_ax1_v = fig_test2_v.add_subplot(2, 3, 1) fig_test2_ax2_v = fig_test2_v.add_subplot(2, 3, 4) fig_test2_ax3_v = fig_test2_v.add_subplot(2, 3, 2) fig_test2_ax4_v = fig_test2_v.add_subplot(2, 3, 5) fig_test2_ax5_v = fig_test2_v.add_subplot(2, 3, 3) fig_test2_ax6_v = fig_test2_v.add_subplot(2, 3, 6) fig_test2_ax1_v.imshow(ismh2_v, cmap='gray') fig_test2_ax1_v.set_title('I_SMH') fig_test2_ax2_v.imshow(np.log1p(np.fft.fftshift(np.abs(ft_ismh2_v ** 2))), cmap='gray') fig_test2_ax2_v.set_title('Fourier Transform of I_SMH') fig_test2_ax3_v.imshow(data_test_2_v.load_g('Mask1', 'phaseraw'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_test2_ax3_v.set_title('Raw Phase') fig_test2_ax4_v.imshow(data_test_2_v.load_g('Mask1', 'phasegM'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_test2_ax4_v.set_title('Phase corrected') fig_test2_ax5_v.imshow(data_test_2_v.load_g('Mask1', 'deltagM')[0], cmap='gray', vmin=-1, vmax=1) fig_test2_ax5_v.set_title('Vertical component of Δg') fig_test2_ax6_v.imshow(data_test_2_v.load_g('Mask1', 'deltagM')[1], cmap='gray', vmin=-1, vmax=1) fig_test2_ax6_v.set_title('Horizontal component of Δg') fig_test2_v.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_2_v_explanation.png', dpi=300, bbox_inches='tight') # Input/Output data in 1D fig_test2_1d_v = plt.figure(figsize=(13, 9)) fig_test2_1d_ax1_v = fig_test2_1d_v.add_subplot(2, 3, 1) fig_test2_1d_ax2_v = fig_test2_1d_v.add_subplot(2, 3, 4) fig_test2_1d_ax3_v = fig_test2_1d_v.add_subplot(2, 3, 2) fig_test2_1d_ax4_v = fig_test2_1d_v.add_subplot(2, 3, 5) fig_test2_1d_ax5_v = fig_test2_1d_v.add_subplot(2, 3, 3) fig_test2_1d_ax6_v = fig_test2_1d_v.add_subplot(2, 3, 6) fig_test2_1d_ax1_v.plot(ismh2_v[:, 128]) fig_test2_1d_ax1_v.set_title('I_SMH') fig_test2_1d_ax2_v.plot(y2_v, np.log1p(np.fft.fftshift(np.abs(ft_ismh2_v ** 2)))[:, 128]) fig_test2_1d_ax2_v.set_title('Fourier Transform of I_SMH') fig_test2_1d_ax3_v.plot(y2_v, data_test_2_v.load_g('Mask1', 'phaseraw')[:, 128]) fig_test2_1d_ax3_v.set_ylim(-np.pi, np.pi) fig_test2_1d_ax3_v.set_title('Raw Phase') fig_test2_1d_ax4_v.plot(y2_v, data_test_2_v.load_g('Mask1', 'phasegM')[:, 128]) fig_test2_1d_ax4_v.set_ylim(-np.pi, np.pi) fig_test2_1d_ax4_v.set_title('Phase corrected') fig_test2_1d_ax5_v.plot(y2_v, data_test_2_v.load_g('Mask1', 'deltagM')[0, :, 128]) fig_test2_1d_ax5_v.set_ylim(-1, 1) fig_test2_1d_ax5_v.set_title('Vertical component of Δg') fig_test2_1d_ax6_v.plot(y2_v, data_test_2_v.load_g('Mask1', 'deltagM')[1, :, 128]) fig_test2_1d_ax6_v.set_ylim(-1, 1) fig_test2_1d_ax6_v.set_title('Horizontal component of Δg') fig_test2_1d_v.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_2_v_explanation_1D.png', dpi=300, bbox_inches='tight') plt.show() # -------------------- Test #2 Improvement --------------- # periodicity in pixels: q q_v = [3, 4, 4.1, 4.2, 4.5, 16, 100] data2_v = [] for periodicity in q_v: ismh_v = np.sin(my2_v * 2 * np.pi / periodicity) ft_ismh_v = np.fft.fft2(ismh_v) circle_v = (128, 128 + round(256 / periodicity)), r2_v data_i_v = data.SMGData() data_i_v.store('ISMHexp', ismh_v) data_i_v.store('FTISMHexp', ft_ismh_v) data_i_v.create_branch('Mask1') data_i_v.store_g('Mask1', 'Mask', circle_v) gpa.gpa('Mask1', data_i_v) data2_v.append(data_i_v) fig_test2_multiple_q_v = plt.figure(figsize=(13, 6)) fig_test2_multiple_q_ax1_v = fig_test2_multiple_q_v.add_subplot(1, 2, 1) fig_test2_multiple_q_ax2_v = fig_test2_multiple_q_v.add_subplot(1, 2, 2) fig_test2_multiple_q_ismh_v = plt.figure(figsize=(13, 6)) fig_test2_multiple_q_ismh_ax1_v = fig_test2_multiple_q_ismh_v.add_subplot(1, 2, 1) fig_test2_multiple_q_ismh_ax2_v = fig_test2_multiple_q_ismh_v.add_subplot(1, 2, 2) fig_test2_error_delta_g_v = plt.figure(figsize=(13, 9)) fig_test2_error_delta_g_ax_v = fig_test2_error_delta_g_v.add_subplot(1, 1, 1) count = 0 mean_test_2_v = [] stand_dev_test_2_v = [] for elements in data2_v: fig_test2_multiple_q_ax1_v.plot(y2_v, elements.load_g('Mask1', 'phasegM')[:, 128], linewidth=3, label=str(q[count])) fig_test2_multiple_q_ax2_v.plot(y2_v, elements.load_g('Mask1', 'deltagM')[0, :, 128], linewidth=3, label=str(q[count])) fig_test2_multiple_q_ismh_ax1_v.plot(y2_v[160:180], elements.load('ISMHexp')[160:180, 128], linewidth=3, label=str(q_v[count])) fig_test2_multiple_q_ismh_ax2_v.plot(y2_v, np.log1p(np.fft.fftshift(np.abs(elements.load('FTISMHexp') ** 2)))[:, 128], linewidth=3, label=str(q_v[count])) error_delta_g_v = np.abs(elements.load_g('Mask1', 'deltagM')[0, :, 128]) fig_test2_error_delta_g_ax_v.plot(y2_v, error_delta_g_v, linewidth=3, label=str(q_v[count])) mean_test_2_v.append(statistics.mean(elements.load_g('Mask1', 'deltagM')[0, :, 128])) stand_dev_test_2_v.append(statistics.stdev(elements.load_g('Mask1', 'deltagM')[0, :, 128])) count += 1 fig_test2_multiple_q_ax1_v.set_ylim(-np.pi, np.pi) fig_test2_multiple_q_ax1_v.legend() fig_test2_multiple_q_ax1_v.set_title('Phase corrected') fig_test2_multiple_q_ax2_v.set_title('Vertical component of Δg') fig_test2_multiple_q_ax2_v.set_ylim(-0.01, 0.01) fig_test2_multiple_q_ismh_ax1_v.set_title('I_SMH') fig_test2_multiple_q_ismh_ax2_v.set_title('Fourier Transform of I_SMH') fig_test2_multiple_q_ismh_ax2_v.legend() fig_test2_error_delta_g_ax_v.set_ylim(0, 0.0025) fig_test2_error_delta_g_ax_v.legend() fig_test2_error_delta_g_ax_v.set_title('Error of the vertical component of Δg ') fig_test2_multiple_q_v.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_2_v_test_results.png', dpi=300, bbox_inches='tight') fig_test2_multiple_q_ismh_v.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_2_v_test_cases.png', dpi=300, bbox_inches='tight') fig_test2_error_delta_g_v.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_2_v_test_error.png', dpi=300, bbox_inches='tight') plt.show() print('Mean test 2 = ', mean_test_2_v) print('StDev test 2 = ', stand_dev_test_2_v) # ####################################### # Test #3 in TestPlan document # ####################################### data_test_3 = data.SMGData() # STEM Moire hologram with a unique periodicity of 4 pixels along horizontal axis on half image and a periodicity of # 4.5 pixel along the same axis on the second half of the image (easy case). This leads to o a 'delta g' of # 1/4-1/4.5 = 0.0277778. x3a = np.linspace(0, 127, 128) x3b = np.linspace(128, 255, 128) y3 = np.linspace(0, 255, 256) mx3a, my3a = np.meshgrid(x3a, y3) mx3b, my3b = np.meshgrid(x3b, y3) ismh3a = np.sin(mx3a * 2 * np.pi / 4) ismh3b = np.sin(mx3b * 2 * np.pi / 4.5) ismh3 = np.concatenate((ismh3a, ismh3b), axis=1) ft_ismh3 = np.fft.fft2(ismh3) # Store data in datastructure data_test_3.store('ISMHexp', ismh3) data_test_3.store('FTISMHexp', ft_ismh3) # Create branch for mask data_test_3.create_branch('Mask1') # Circle of radius R centered around coordinate (192, 128) r3 = 20 center3 = (192, 128) circle3 = center3, r3 # Store mask properties into datastructure data_test_3.store_g('Mask1', 'Mask', circle3) # Entering gpa gpa.gpa('Mask1', data_test_3) # Display data # Input/Output data - Phase is supposed to be constant and equal to 0, anything different from 0 represents the error. fig_test3 = plt.figure(figsize=(13, 9)) fig_test3_ax1 = fig_test3.add_subplot(2, 3, 1) fig_test3_ax2 = fig_test3.add_subplot(2, 3, 4) fig_test3_ax3 = fig_test3.add_subplot(2, 3, 2) fig_test3_ax4 = fig_test3.add_subplot(2, 3, 5) fig_test3_ax5 = fig_test3.add_subplot(2, 3, 3) fig_test3_ax6 = fig_test3.add_subplot(2, 3, 6) fig_test3_ax1.imshow(ismh3, cmap='gray') fig_test3_ax2.imshow(np.log1p(np.fft.fftshift(np.abs(ft_ismh3 ** 2))), cmap='gray') fig_test3_ax3.imshow(data_test_3.load_g('Mask1', 'phaseraw'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_test3_ax4.imshow(data_test_3.load_g('Mask1', 'phasegM'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_test3_ax5.imshow(data_test_3.load_g('Mask1', 'deltagM')[0], cmap='gray', vmin=-1, vmax=1) fig_test3_ax6.imshow(data_test_3.load_g('Mask1', 'deltagM')[1], cmap='gray', vmin=-1, vmax=1) fig_test3_ax1.set_title('I_SMH') fig_test3_ax2.set_title('Fourier Transform of I_SMH') fig_test3_ax3.set_title('Raw Phase') fig_test3_ax4.set_title('Phase corrected') fig_test3_ax5.set_title('Vertical component of Δg') fig_test3_ax6.set_title('Horizontal component of Δg') fig_test3.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_3_explanation.png', dpi=300, bbox_inches='tight') # Input/Output data in 1D fig_test3_1d = plt.figure(figsize=(13, 9)) fig_test3_1d_ax1 = fig_test3_1d.add_subplot(2, 3, 1) fig_test3_1d_ax2 = fig_test3_1d.add_subplot(2, 3, 4) fig_test3_1d_ax3 = fig_test3_1d.add_subplot(2, 3, 2) fig_test3_1d_ax4 = fig_test3_1d.add_subplot(2, 3, 5) fig_test3_1d_ax5 = fig_test3_1d.add_subplot(2, 3, 3) fig_test3_1d_ax6 = fig_test3_1d.add_subplot(2, 3, 6) fig_test3_1d_ax1.plot(x2[50:206], ismh3[128, 50:206]) fig_test3_1d_ax2.plot(x2, np.log1p(np.fft.fftshift(np.abs(ft_ismh3 ** 2)))[128, :]) fig_test3_1d_ax3.plot(x2, data_test_3.load_g('Mask1', 'phaseraw')[128, :]) fig_test3_1d_ax3.set_ylim(-np.pi, np.pi) fig_test3_1d_ax4.plot(x2, data_test_3.load_g('Mask1', 'phasegM')[128, :]) fig_test3_1d_ax4.set_ylim(-np.pi, np.pi) fig_test3_1d_ax5.plot(x2, data_test_3.load_g('Mask1', 'deltagM')[0, 128, :]) fig_test3_1d_ax5.set_ylim(-0.1, 0.1) fig_test3_1d_ax6.plot(x2, data_test_3.load_g('Mask1', 'deltagM')[1, 128, :]) fig_test3_1d_ax6.set_ylim(-0.1, 0.1) fig_test3_1d_ax1.set_title('I_SMH') fig_test3_1d_ax2.set_title('Fourier Transform of I_SMH') fig_test3_1d_ax3.set_title('Raw Phase') fig_test3_1d_ax4.set_title('Phase corrected') fig_test3_1d_ax5.set_title('Vertical component of Δg') fig_test3_1d_ax6.set_title('Horizontal component of Δg') fig_test3_1d.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_3_explanation_1D.png', dpi=300, bbox_inches='tight') plt.show() # -------------------- Test #3 Improvement --------------- # Different strain level: dg dg = [1/4 - 1/3, 1/4 - 1/3.8, 1/4 - 1/3.9, 1/4 - 1/4, 1/4 - 1/4.1, 1/4 - 1/4.2, 1/4 - 1/5] # Different strain periodicity dq = [3, 3.8, 3.9, 4, 4.1, 4.2, 5] data3 = [] for strain in dq: ismhb = np.sin(mx3b * 2 * np.pi / strain) ismh = np.concatenate((ismh3a, ismhb), axis=1) ft_ismh = np.fft.fft2(ismh) circle = (128 + round(256 / 4), 128), r3 print(circle) data_i = data.SMGData() data_i.store('ISMHexp', ismh) data_i.store('FTISMHexp', ft_ismh) data_i.create_branch('Mask1') data_i.store_g('Mask1', 'Mask', circle) gpa.gpa('Mask1', data_i) data3.append(data_i) fig_test3_multiple_dq = plt.figure(figsize=(13, 6)) fig_test3_multiple_dq_ax1 = fig_test3_multiple_dq.add_subplot(1, 2, 1) fig_test3_multiple_dq_ax2 = fig_test3_multiple_dq.add_subplot(1, 2, 2) count = 0 strain_values = [] for elements in data3: fig_test3_multiple_dq_ax1.plot(x2[100:200], elements.load_g('Mask1', 'phasegM')[128, 100:200], linewidth=3, label=str(dq[count])) fig_test3_multiple_dq_ax2.plot(x2, elements.load_g('Mask1', 'deltagM')[1, 128, :], linewidth=3, label=str(dq[count])) strain_value = elements.load_g('Mask1', 'deltagM')[1, 128, 200] strain_values.append(strain_value) count += 1 fig_test3_multiple_dq_ax1.set_ylim(-np.pi, np.pi) fig_test3_multiple_dq_ax1.legend() fig_test3_multiple_dq_ax1.set_title('Raw phase') fig_test3_multiple_dq_ax2.set_title('Horizontal component of Δg') fig_test3_multiple_dq_ax2.set_ylim(-0.1, 0.1) fig_test3_multiple_dq.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_3_test_results.png', dpi=300, bbox_inches='tight') print('dg = ', dg) print('strain = ', strain_values) print('error = ', np.abs(np.array(dg)-np.array(strain_values))) plt.show() # ----------------------- Test vertical component ---------------------- data_test_3_v = data.SMGData() # STEM Moire hologram with a unique periodicity of 4 pixels along horizontal axis on half image and a periodicity of # 4.5 pixel along the same axis on the second half of the image (easy case). This leads to o a 'delta g' of # 1/4-1/4.5 = 0.0277778. y3a_v = np.linspace(0, 127, 128) y3b_v = np.linspace(128, 255, 128) x3_v = np.linspace(0, 255, 256) mx3a_v, my3a_v = np.meshgrid(x3_v, y3a_v) mx3b_v, my3b_v = np.meshgrid(x3_v, y3b_v) ismh3a_v = np.sin(my3a_v * 2 * np.pi / 4) ismh3b_v = np.sin(my3b_v * 2 * np.pi / 4.5) ismh3_v = np.concatenate((ismh3a_v, ismh3b_v), axis=0) ft_ismh3_v = np.fft.fft2(ismh3_v) # Store data in datastructure data_test_3_v.store('ISMHexp', ismh3_v) data_test_3_v.store('FTISMHexp', ft_ismh3_v) # Create branch for mask data_test_3_v.create_branch('Mask1') # Circle of radius R centered around coordinate (192, 128) r3_v = 20 center3_v = (128, 192) circle3_v = center3_v, r3_v # Store mask properties into datastructure data_test_3_v.store_g('Mask1', 'Mask', circle3_v) # Entering gpa gpa.gpa('Mask1', data_test_3_v) # Display data # Input/Output data - Phase is supposed to be constant and equal to 0, anything different from 0 represents the error. fig_test3_v = plt.figure(figsize=(13, 9)) fig_test3_ax1_v = fig_test3_v.add_subplot(2, 3, 1) fig_test3_ax2_v = fig_test3_v.add_subplot(2, 3, 4) fig_test3_ax3_v = fig_test3_v.add_subplot(2, 3, 2) fig_test3_ax4_v = fig_test3_v.add_subplot(2, 3, 5) fig_test3_ax5_v = fig_test3_v.add_subplot(2, 3, 3) fig_test3_ax6_v = fig_test3_v.add_subplot(2, 3, 6) fig_test3_ax1_v.imshow(ismh3_v, cmap='gray') fig_test3_ax2_v.imshow(np.log1p(np.fft.fftshift(np.abs(ft_ismh3_v ** 2))), cmap='gray') fig_test3_ax3_v.imshow(data_test_3_v.load_g('Mask1', 'phaseraw'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_test3_ax4_v.imshow(data_test_3_v.load_g('Mask1', 'phasegM'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_test3_ax5_v.imshow(data_test_3_v.load_g('Mask1', 'deltagM')[0], cmap='gray', vmin=-1, vmax=1) fig_test3_ax6_v.imshow(data_test_3_v.load_g('Mask1', 'deltagM')[1], cmap='gray', vmin=-1, vmax=1) fig_test3_ax1_v.set_title('I_SMH') fig_test3_ax2_v.set_title('Fourier Transform of I_SMH') fig_test3_ax3_v.set_title('Raw Phase') fig_test3_ax4_v.set_title('Phase corrected') fig_test3_ax5_v.set_title('Vertical component of Δg') fig_test3_ax6_v.set_title('Horizontal component of Δg') fig_test3_v.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_3_v_explanation.png', dpi=300, bbox_inches='tight') # Input/Output data in 1D fig_test3_1d_v = plt.figure(figsize=(13, 9)) fig_test3_1d_ax1_v = fig_test3_1d_v.add_subplot(2, 3, 1) fig_test3_1d_ax2_v = fig_test3_1d_v.add_subplot(2, 3, 4) fig_test3_1d_ax3_v = fig_test3_1d_v.add_subplot(2, 3, 2) fig_test3_1d_ax4_v = fig_test3_1d_v.add_subplot(2, 3, 5) fig_test3_1d_ax5_v = fig_test3_1d_v.add_subplot(2, 3, 3) fig_test3_1d_ax6_v = fig_test3_1d_v.add_subplot(2, 3, 6) fig_test3_1d_ax1_v.plot(x2[50:206], ismh3_v[50:206, 128]) fig_test3_1d_ax2_v.plot(x2, np.log1p(np.fft.fftshift(np.abs(ft_ismh3_v ** 2)))[:, 128]) fig_test3_1d_ax3_v.plot(x2, data_test_3_v.load_g('Mask1', 'phaseraw')[:, 128]) fig_test3_1d_ax3_v.set_ylim(-np.pi, np.pi) fig_test3_1d_ax4_v.plot(x2, data_test_3_v.load_g('Mask1', 'phasegM')[:, 128]) fig_test3_1d_ax4_v.set_ylim(-np.pi, np.pi) fig_test3_1d_ax5_v.plot(x2, data_test_3_v.load_g('Mask1', 'deltagM')[0, :, 128]) fig_test3_1d_ax5_v.set_ylim(-0.1, 0.1) fig_test3_1d_ax6_v.plot(x2, data_test_3_v.load_g('Mask1', 'deltagM')[1, :, 128]) fig_test3_1d_ax6_v.set_ylim(-0.1, 0.1) fig_test3_1d_ax1_v.set_title('I_SMH') fig_test3_1d_ax2_v.set_title('Fourier Transform of I_SMH') fig_test3_1d_ax3_v.set_title('Raw Phase') fig_test3_1d_ax4_v.set_title('Phase corrected') fig_test3_1d_ax5_v.set_title('Vertical component of Δg') fig_test3_1d_ax6_v.set_title('Horizontal component of Δg') fig_test3_1d_v.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_3_v_explanation_1D.png', dpi=300, bbox_inches='tight') # -------------------- Test #3 Improvement --------------- # Different strain level: dg dg = [1/4 - 1/3, 1/4 - 1/3.8, 1/4 - 1/3.9, 1/4 - 1/4, 1/4 - 1/4.1, 1/4 - 1/4.2, 1/4 - 1/5] # Different strain periodicity dq = [3, 3.8, 3.9, 4, 4.1, 4.2, 5] data3 = [] for strain in dq: ismhb = np.sin(mx3b * 2 * np.pi / strain) ismh = np.concatenate((ismh3a, ismhb), axis=1) ft_ismh = np.fft.fft2(ismh) circle = (128 + round(256 / 4), 128), r3 print(circle) data_i = data.SMGData() data_i.store('ISMHexp', ismh) data_i.store('FTISMHexp', ft_ismh) data_i.create_branch('Mask1') data_i.store_g('Mask1', 'Mask', circle) gpa.gpa('Mask1', data_i) data3.append(data_i) fig_test3_multiple_dq = plt.figure(figsize=(13, 6)) fig_test3_multiple_dq_ax1 = fig_test3_multiple_dq.add_subplot(1, 2, 1) fig_test3_multiple_dq_ax2 = fig_test3_multiple_dq.add_subplot(1, 2, 2) count = 0 strain_values = [] for elements in data3: fig_test3_multiple_dq_ax1.plot(x2[100:200], elements.load_g('Mask1', 'phasegM')[128, 100:200], linewidth=3, label=str(dq[count])) fig_test3_multiple_dq_ax2.plot(x2, elements.load_g('Mask1', 'deltagM')[1, 128, :], linewidth=3, label=str(dq[count])) strain_value = elements.load_g('Mask1', 'deltagM')[1, 128, 200] strain_values.append(strain_value) count += 1 fig_test3_multiple_dq_ax1.set_ylim(-np.pi, np.pi) fig_test3_multiple_dq_ax1.legend() fig_test3_multiple_dq_ax1.set_title('Raw phase') fig_test3_multiple_dq_ax2.set_title('Horizontal component of Δg') fig_test3_multiple_dq_ax2.set_ylim(-0.1, 0.1) fig_test3_multiple_dq.savefig( '/media/alex/Work/PhD/Course/CAS 741/project/STEM_Moire_GPA/Doc/TestReport/Figures/Test_3_test_results.png', dpi=300, bbox_inches='tight') print('dg = ', dg) print('strain = ', strain_values) print('error = ', np.abs(np.array(dg)-np.array(strain_values))) plt.show() # ####################################### # Test #4 in TestPlan document # ####################################### data_test_4 = data.SMGData() # STEM Moire hologram with a unique periodicity of 4 pixels along horizontal axis on half image and a periodicity of # 4.5 pixel along the same axis on the second half of the image (easy case). This leads to o a 'delta g' of # 1/4-1/4.5 = 0.0277778. x4a = np.linspace(0, 127, 128) x4b = np.linspace(128, 255, 128) y4 = np.linspace(0, 255, 256) mx4a, my4a = np.meshgrid(x4a, y4) mx4b, my4b = np.meshgrid(x4b, y4) ismh4a = np.sin(mx4a * 2 * np.pi / 4) ismh4b = np.sin(mx4b * 2 * np.pi / 4.5) ismh4 = np.concatenate((ismh4a, ismh4b), axis=1) ft_ismh4 = np.fft.fft2(ismh4) # Store data in datastructure data_test_4.store('ISMHexp', ismh4) data_test_4.store('FTISMHexp', ft_ismh4) # Mask parameter r4 = [1, 2, 4, 8, 16, 32, 64, 128, 256] center4 = (192, 128) # Create the amount of branches corresponding the the radius tested and store the mask properties for radius in r4: data_test_4.create_branch(str(radius)) circle4 = center4, radius data_test_4.store_g(str(radius), 'Mask', circle4) # Looping on the various radius into gpa for radius in r4: gpa.gpa(str(radius), data_test_4) count = 1 fig_test4 = plt.figure() fig_test4_1d = plt.figure() fig_test4_deltag_1d = plt.figure() N_sublpot = math.ceil(math.sqrt(len(r4))) print(math.ceil(math.sqrt(len(r4)))) for element in r4: fig_test4.add_subplot(N_sublpot, N_sublpot, count).imshow(data_test_4.load_g( str(element), 'phasegM'), cmap='gray', vmin=-np.pi, vmax=np.pi) fig_ax = fig_test4_1d.add_subplot(N_sublpot, N_sublpot, count) fig_ax.plot(x2, data_test_4.load_g(str(element), 'phasegM')[128, :]) fig_ax.set_ylim(-np.pi, np.pi) fig_ax_g = fig_test4_deltag_1d.add_subplot(N_sublpot, N_sublpot, count) fig_ax_g.plot(x2, data_test_4.load_g( str(element), 'deltagM')[1, 128, :]) fig_ax_g.set_ylim(-0.1, 0.1) count += 1 plt.show()
slimpotatoes/STEM_Moire_GPA
src/smhsimulation.py
# STEM Moire hologram simulation module import numpy as np import data as data import math as math from scipy.ndimage.interpolation import shift def smh_sim(datastruct): """Simulate the Fourier transform of the STEM Moire hologram based on the reference image""" # Make the Fourier transform of the STEM Moire hologram and store it. data.SMGData.store(datastruct, 'FTISMHexp', np.fft.fft2(data.SMGData.load(datastruct, 'ISMHexp'))) # Load the intermediate variables from the reference image to make the simulation. ft_ic = np.fft.fftshift(np.abs(np.fft.fft2(data.SMGData.load(datastruct, 'ICref')) ** 2)) p = data.SMGData.load(datastruct, 'p') pref = data.SMGData.load(datastruct, 'pref') n_lim = int(math.floor(p/pref)) fov = ft_ic.shape[0] tile = int(pref / p * fov) ft_ismh_sim = np.ndarray(ft_ic.shape) # Simulation by calculating the STEM Moire hologram equation. counter = 0 if n_lim % 2 != 0: initial_coordinate = (int(0.5 * (fov - n_lim * tile)), int(0.5 * (fov - n_lim * tile))) ft_ic_square = np.ndarray((len(range(0, n_lim)), len(range(0, n_lim)), tile, tile)) else: n_lim -= 1 initial_coordinate = (int(0.5 * (fov - n_lim * tile)), int(0.5 * (fov - n_lim * tile))) ft_ic_square = np.ndarray((len(range(0, n_lim)), len(range(0, n_lim)), tile, tile)) # Print statement to inform user print('Shape in pixel of reference: ', ft_ismh_sim.shape) print('Shape in pixel of the tile: ', ft_ic_square.shape) print('Tile size (in pixel): ', tile) print('Number of tiles: ', n_lim ** 2) print('Please wait the calculation can take some time !!!') for i in range(0, n_lim): for j in range(0, n_lim): ft_ismh_sim += shift(ft_ic, [int(i - 0.5 * n_lim) * tile, int(j - 0.5 * n_lim) * tile], cval=0, order=0, prefilter=False) a = int(initial_coordinate[0] + i * tile) b = a + tile c = int(initial_coordinate[0] + j * tile) d = c + tile ft_ic_square[i][j] = ft_ic[a:b, c:d] counter += 1 # print('Loop number: ', counter) print('Simulation done') # Storing data into data structure. data.SMGData.store(datastruct, 'FTISMHsim', ft_ismh_sim[ int(0.5*(ft_ismh_sim.shape[0] - tile)):int(0.5*(ft_ismh_sim.shape[0] + tile)), int(0.5*(ft_ismh_sim.shape[0] - tile)):int(0.5*(ft_ismh_sim.shape[0] + tile))]) data.SMGData.store(datastruct, 'FTISMHsimDisplay', ft_ic_square)
slimpotatoes/STEM_Moire_GPA
src/mask.py
<reponame>slimpotatoes/STEM_Moire_GPA # Mask Module import numpy as np def mask_classic(center, r, shape): """Return the mask function in the image space I defined by shape (see MIS). The classic mask takes the center of a circle center = (xc,yc) and its radius r and put 1 if the image space is inside the circle and 0 outside. In addition, the center of the mask is used to return g_0 that correspond to a first estimation of the unstrained reference.""" """Do not forget event coordinate (x,y) from matplotlib should be switched compared to numpy array indexing""" g_0 = np.array([(center[1] - 0.5 * shape[0]) / shape[0] * np.ones(shape), (center[0] - 0.5 * shape[1]) / shape[1] * np.ones(shape)]) mask = np.ndarray(shape=shape) """Do not forget event coordinate (x,y) from matplotlib should be switched compared to numpy array indexing""" for i in range(0, shape[1]): for j in range(0, shape[0]): if ((i - center[1]) ** 2 + (j - center[0]) ** 2) < (r ** 2): mask[i, j] = 1 else: mask[i, j] = 0 return mask, g_0 def mask_gaussian(center, r, shape): """Return the mask function in the image space I defined by shape (see MIS). The Gaussian mask takes the center of a circle center = (xc,yc) and its radius r to generate a 2D gaussian function centered around the circle. In addition, the center of the mask is used to return g_0 that correspond to the unstrained reference.""" """Do not forget event coordinate (x,y) from matplotlib should be switched compared to numpy array indexing""" g_0 = np.array([(center[1] - 0.5 * shape[0]) / shape[0] * np.ones(shape), (center[0] - 0.5 * shape[1]) / shape[1] * np.ones(shape)]) """Do not forget event coordinate (x,y) from matplotlib should be switched compared to numpy array indexing - r corresponds to 3 * sigma => 99% gaussian mask included in circle""" const = 1 / (2 * (r / 3) ** 2) mesh_x, mesh_y = np.meshgrid(np.arange(shape[0]), np.arange(shape[1])) delta_x = (mesh_x - center[0]) ** 2 delta_y = (mesh_y - center[1]) ** 2 mask = np.exp(-(delta_x + delta_y) * const) return mask, g_0
slimpotatoes/STEM_Moire_GPA
src/gui.py
# STEM Moire GPA GUI Module import matplotlib.pyplot as plt import matplotlib.gridspec as grid import matplotlib.cm as cm import matplotlib.colors as colors import matplotlib.artist as artist import matplotlib.patches as patch from tkinter import filedialog from matplotlib.widgets import Button from matplotlib.widgets import TextBox from matplotlib_scalebar.scalebar import ScaleBar import data as data import numpy as np import guimaskmanager as maskmanag import guirectanglemanager as rectmanag import guilinemanager as linemanag import guidisplay as display class SMGGUI(object): def __init__(self, datastruct): self.fig_GUIFlow = None self.event_input = None self.event_smhsim = None self.event_gpa = None self.event_ref = None self.event_convert = None self.event_strain = None self.fig_SMHexp = None self.ax_fig_SMHexp = None self.fig_SMHSim = None self.fig_GPA_M1 = None self.fig_GPA_M2 = None self.fig_GPA_M1_ax = None self.fig_GPA_M2_ax= None self.rectangle_M1 = None self.rectangle_M2 = None self.fig_NM = None self.fig_strain = None self.mask =dict() self.mask_selected = None self.h_1 = 0 self.h_2 = 0 self.v_1 = 0 self.v_2 = 0 self.datastruct = datastruct self.line_rot = None def guiflow(self): self.fig_GUIFlow = plt.figure(num='SMG Flow', figsize=(2, 5)) gs_button = grid.GridSpec(6, 1) self.event_input = Button(self.fig_GUIFlow.add_axes(self.fig_GUIFlow.add_subplot(gs_button[0, 0])), 'Input') self.event_smhsim = Button(self.fig_GUIFlow.add_axes(self.fig_GUIFlow.add_subplot(gs_button[1, 0])), 'SMHSim') self.event_gpa = Button(self.fig_GUIFlow.add_axes(self.fig_GUIFlow.add_subplot(gs_button[2, 0])), 'GPA') self.event_ref = Button(self.fig_GUIFlow.add_axes(self.fig_GUIFlow.add_subplot(gs_button[3, 0])), 'Ref') self.event_convert = Button(self.fig_GUIFlow.add_axes(self.fig_GUIFlow.add_subplot(gs_button[4, 0])), 'Convert') self.event_strain = Button(self.fig_GUIFlow.add_axes(self.fig_GUIFlow.add_subplot(gs_button[5, 0])), 'Strain') self.fig_GUIFlow.canvas.mpl_connect('key_press_event', self.custom_display) def guiconv(self): def collect_shift_h1(text): if text != '': if isinstance(int(text), int) == True: self.h_1 = int(text) else: return else: return def collect_shift_v1(text): if text != '': if isinstance(int(text), int) == True: self.v_1 = int(text) else: return else: return def collect_shift_h2(text): if text != '': if isinstance(int(text), int) == True: self.h_2 = int(text) else: return else: return def collect_shift_v2(text): if text != '': if isinstance(int(text), int) == True: self.v_2 = int(text) else: return else: return self.fig_NM = plt.figure(figsize=(1.2,2), num='(n,m) shift') self.fig_NM.canvas.mpl_disconnect(self.fig_NM.canvas.manager.key_press_handler_id) fig_ax_text_1 = self.fig_NM.add_axes(plt.axes([0.1, 0.8, 0.8, 0.1])) fig_ax_text_2 = self.fig_NM.add_axes(plt.axes([0.1, 0.4, 0.8, 0.1])) fig_ax_text_1.set_axis_off() fig_ax_text_1.format_coord = lambda x, y: "" fig_ax_text_1.text(0, 0.2, 'Shift related to the Red mask') fig_ax_text_2.set_axis_off() fig_ax_text_2.text(0, 0.2, 'Shift related to the Blue mask') fig_ax_text_2.format_coord = lambda x, y: "" self.textbox_h_1 = TextBox(self.fig_NM.add_axes(plt.axes([0.6, 0.65, 0.2, 0.1])), 'Horizontal shift :', initial='0', label_pad=0.2) self.textbox_v_1 = TextBox(self.fig_NM.add_axes(plt.axes([0.6, 0.55, 0.2, 0.1])), 'Vertical shift :', initial='0', label_pad=0.2) self.textbox_h_2 = TextBox(self.fig_NM.add_axes(plt.axes([0.6, 0.25, 0.2, 0.1])), 'Horizontal shift :', initial='0', label_pad=0.2) self.textbox_v_2 = TextBox(self.fig_NM.add_axes(plt.axes([0.6, 0.15, 0.2, 0.1])), 'Vertical shift :', initial='0', label_pad=0.2) id1h = self.textbox_h_1.on_text_change(collect_shift_h1) id2h = self.textbox_h_2.on_text_change(collect_shift_h2) id1v = self.textbox_v_1.on_text_change(collect_shift_v1) id2v = self.textbox_v_2.on_text_change(collect_shift_v2) def guismhexp(self, datastruct): self.fig_SMHexp = plt.figure(num='SMH and reference image') self.ax_fig_SMHexp = self.fig_SMHexp.add_subplot(1, 2, 1) self.ax_fig_SMHexp.imshow(data.SMGData.load(datastruct, 'ISMHexp'), cmap='gray') scalebar1 = ScaleBar(data.SMGData.load(datastruct, 'p') * 10 ** -9) plt.gca().add_artist(scalebar1) self.line_rot = linemanag.LineDraw(self.ax_fig_SMHexp) self.line_rot.ConnectDraw() self.fig_SMHexp.add_subplot(1, 2, 2).imshow(data.SMGData.load(datastruct, 'ICref'), cmap='gray') scalebar2 = ScaleBar(data.SMGData.load(datastruct, 'pref') * 10 ** -9) plt.gca().add_artist(scalebar2) plt.show() def guismhsim(self, datastruct): def edit_mode(event): if event.key == 'e': print('Edit mode open, please edit your masks') for circle in self.circles: circle.connect() for element in self.mask.keys(): data.SMGData.remove_branch(datastruct, element) if event.key == 'd': print('Edit mode closed, please select the mask for the GPA process') for circle in self.circles: self.mask[artist.Artist.get_gid(circle.artist)] = (circle.artist.center, circle.artist.radius) circle.disconnect_edit() for element in self.mask.keys(): data.SMGData.create_branch(datastruct, element) data.SMGData.store_g(datastruct, element, 'Mask', self.mask[element]) print(element, ' Center : ', self.mask[element][0], ' Radius : ', self.mask[element][1]) self.fig_SMHsim = plt.figure(num='SMH Simulation') self.fig_SMHsim.canvas.mpl_connect('key_press_event', edit_mode) self.fig_SMHsim_axis = self.fig_SMHsim.add_subplot(1,2,1) ftsmhexp = data.SMGData.load(datastruct, 'FTISMHexp') self.fig_SMHsim_axis.imshow(np.log1p(self.fft_display(ftsmhexp)), cmap='gray') self.fig_SMHsim.add_subplot(1, 2, 2).imshow(np.log1p(data.SMGData.load(datastruct, 'FTISMHsim')), cmap='gray') fticsquare = data.SMGData.load(datastruct, 'FTISMHsimDisplay') colormaps_Moire, colormaps_IC = self.generate_colormap_smhsim(fticsquare.shape[0] * fticsquare.shape[1]) ftsmhsim = plt.figure(num='Simulated SMH with colors') icsplit = plt.figure(num='Ic split into tiles') # A cleaner ZERO = np.zeros(fticsquare[0][0].shape) ftsmhsimaxis = ftsmhsim.add_subplot(1, 1, 1) ftsmhsimaxis.imshow(ZERO, cmap="gray", alpha=1) max_general = np.max(np.max(np.max(np.max(np.log1p(fticsquare))))) count = 0 for i in range(0, fticsquare.shape[0]): for j in range(0, fticsquare.shape[1]): # Threshold - Color on Moire FFT reconstructed Test = np.log1p(fticsquare[i, j]) mask = Test[:, :] < (0.67 * max_general) Test[mask] = 0 cmap_moire = colors.LinearSegmentedColormap.from_list('my_cmap', colormaps_Moire[count]) cmap_ic = colors.LinearSegmentedColormap.from_list('my_cmap', colormaps_IC[count]) ftsmhsimaxis.imshow(Test, cmap = cmap_moire, alpha = .7, clim=(20, 40)) # Color on split HRES FFT icsplitaxis = icsplit.add_subplot(fticsquare.shape[0], fticsquare.shape[0], count + 1) icsplitaxis.imshow(np.log1p(fticsquare[i, j]), cmap=cmap_ic, clim=(27.5, 34.5)) icsplitaxis.xaxis.set_visible(False) icsplitaxis.yaxis.set_visible(False) count += 1 smgmaskcreate = maskmanag.MaskCreator(self.fig_SMHsim_axis,ftsmhexp) circle1 = smgmaskcreate.make_circle('Mask1') circle2 = smgmaskcreate.make_circle('Mask2', colored='b',off_center=(20,20)) self.mask[circle1[0]] = circle1[1] self.mask[circle2[0]] = circle2[1] self.circles = [] for el in self.fig_SMHsim_axis.artists: smgmaskedit = maskmanag.MaskEditor(el) self.circles.append(smgmaskedit) plt.show() def guiphase(self, mask_id, datastruct): def close_window(event): if event.canvas.figure == self.fig_GPA_M1: self.fig_GPA_M1.canvas.mpl_disconnect(cid1close) self.fig_GPA_M1.canvas.mpl_disconnect(cid1draw) self.fig_GPA_M1 = None elif event.canvas.figure == self.fig_GPA_M2: self.fig_GPA_M2.canvas.mpl_disconnect(cid2close) self.fig_GPA_M2.canvas.mpl_disconnect(cid2draw) self.fig_GPA_M2 = None def change_rectangle(event): if event.canvas.figure == self.fig_GPA_M1 and self.fig_GPA_M2 != None and self.phaseref.done == 1: self.phaseref.done = 0 self.phaseref2.remove_rectangle() rectangle = self.fig_GPA_M1_ax.findobj(patch.Rectangle)[0] self.phaseref2.create_rectangle(rectangle) U = self.reference_extract(rectangle) data.SMGData.store(datastruct, 'Uref', U) elif event.canvas.figure == self.fig_GPA_M2 and self.fig_GPA_M1 != None and self.phaseref2.done == 1: self.phaseref2.done = 0 self.phaseref.remove_rectangle() rectangle = self.fig_GPA_M2_ax.findobj(patch.Rectangle)[0] self.phaseref.create_rectangle(rectangle) U = self.reference_extract(rectangle) data.SMGData.store(datastruct, 'Uref', U) else: return if mask_id == 'Mask1': if self.fig_GPA_M1 == None: self.fig_GPA_M1 = plt.figure(num='GPA - Mask Red') self.fig_GPA_M1_ax = self.fig_GPA_M1.add_subplot(1, 1, 1) phase = data.SMGData.load_g(datastruct, mask_id, 'phasegM') self.image_mask_1 = self.fig_GPA_M1_ax.imshow(phase, cmap='gray') self.rectangle_M1 = rectmanag.make_rectangle(self.fig_GPA_M1_ax, phase) self.phaseref = rectmanag.RectEditor(self.fig_GPA_M1, self.fig_GPA_M1_ax, self.rectangle_M1) self.phaseref.connect() cid1close = self.fig_GPA_M1.canvas.mpl_connect('close_event', close_window) cid1draw = self.fig_GPA_M1.canvas.mpl_connect('draw_event', change_rectangle) plt.show() else: self.fig_GPA_M1_ax.imshow(data.SMGData.load_g(datastruct, mask_id, 'phasegM'), cmap='gray') plt.draw() if mask_id == 'Mask2': if self.fig_GPA_M2 == None: self.fig_GPA_M2 = plt.figure(num='GPA - Mask Blue') self.fig_GPA_M2_ax = self.fig_GPA_M2.add_subplot(1, 1, 1) phase = data.SMGData.load_g(datastruct, mask_id, 'phasegM') self.image_mask_2 = self.fig_GPA_M2_ax.imshow(phase, cmap='gray') self.rectangle_M2 = rectmanag.make_rectangle(self.fig_GPA_M2_ax, phase) self.phaseref2 = rectmanag.RectEditor(self.fig_GPA_M2, self.fig_GPA_M2_ax, self.rectangle_M2) self.phaseref2.connect() cid2close = self.fig_GPA_M2.canvas.mpl_connect('close_event', close_window) cid2draw = self.fig_GPA_M2.canvas.mpl_connect('draw_event', change_rectangle) plt.show() else: self.fig_GPA_M2_ax.imshow(data.SMGData.load_g(datastruct, mask_id, 'phasegM'), cmap='gray') plt.draw() else: return def guistrain(self, datastruct): self.fig_strain = plt.figure(num='Strain maps') fig_strain_ax_exx = self.fig_strain.add_subplot(2, 2, 1) fig_strain_ax_eyy = self.fig_strain.add_subplot(2, 2, 2) fig_strain_ax_exy = self.fig_strain.add_subplot(2, 2, 3) fig_strain_ax_rxy = self.fig_strain.add_subplot(2, 2, 4) exx = fig_strain_ax_exx.imshow(data.SMGData.load(datastruct,'Exx'), cmap='bwr', vmin=-0.02, vmax=0.02) eyy = fig_strain_ax_eyy.imshow(data.SMGData.load(datastruct, 'Eyy'), cmap='bwr', vmin=-0.02, vmax=0.02) exy = fig_strain_ax_exy.imshow(data.SMGData.load(datastruct, 'Exy'), cmap='bwr', vmin=-0.02, vmax=0.02) rxy = fig_strain_ax_rxy.imshow(data.SMGData.load(datastruct, 'Rxy'), cmap='bwr', vmin=-0.02,vmax=0.02) fig_strain_ax_exx.set_title('εxx') fig_strain_ax_exx.xaxis.set_visible(False) fig_strain_ax_eyy.set_title('εyy') fig_strain_ax_eyy.xaxis.set_visible(False) fig_strain_ax_exy.set_title('εxy') fig_strain_ax_exy.xaxis.set_visible(False) fig_strain_ax_rxy.set_title('ωxy') fig_strain_ax_rxy.xaxis.set_visible(False) fig_strain_ax_rxy.add_artist(ScaleBar(data.SMGData.load(datastruct, 'p') * 10 ** -9)) plt.show() @staticmethod def fft_display(fft): return np.fft.fftshift(np.abs(fft ** 2)) @staticmethod def generate_colormap_smhsim(total_tiles): initial_color_map = cm.get_cmap(name="jet") coef_weighted_color = np.arange(0, total_tiles).astype(float) / (total_tiles - 1) customized_color_map_max = [] for elements in coef_weighted_color: customized_color_map_max.append(initial_color_map(elements)) customized_color_maps_Moire = [] customized_color_maps_Ic =[] for elements in customized_color_map_max: customized_color_maps_Moire.append([(0, 0, 0, 0.0), elements]) customized_color_maps_Ic.append([(0, 0, 0, 0.8), elements]) return customized_color_maps_Moire, customized_color_maps_Ic def mask_selection(self): for circle in self.circles: if circle.mask_selected is not None: self.mask_selected = circle.mask_selected print('Mask selected') return self.mask_selected def reference_extract(self, rectangle): x0, y0 = rectangle.get_xy() x1, y1 = x0 + rectangle.get_width(), y0 + rectangle.get_height() print(int(x0), int(y0), int(x1), int(y1)) return int(x0), int(y0), int(x1), int(y1) def update_phase(self, mask_id, datastruct): if mask_id == 'Mask1': phase = data.SMGData.load_g(datastruct, mask_id, 'phasegM') print('Mask 1 update') self.fig_GPA_M1_ax.imshow(phase, cmap='gray') self.fig_GPA_M1.canvas.draw() elif mask_id == 'Mask2': print('Mask 2 update') phase = data.SMGData.load_g(datastruct, mask_id, 'phasegM') self.fig_GPA_M2_ax.imshow(phase, cmap='gray') self.fig_GPA_M2.canvas.draw() else: pass @staticmethod def open_files(): file_path_smh = filedialog.askopenfilename(title="Load the STEM Moire hologram") file_path_ic = filedialog.askopenfilename(title="Load the reference image") return file_path_smh, file_path_ic # NOT DOCUMENTED AND NEED TO BE DESIGNED PROPERLY def custom_display(self, event): if event.key == '1': data_to_display = data.SMGData.load(self.datastruct,'ISMHexp') p = data.SMGData.load(self.datastruct, 'p') display.GUIDisplay(data_to_display, cal=p) if event.key == '2': data_to_display = data.SMGData.load(self.datastruct, 'ICref') p = data.SMGData.load(self.datastruct, 'pref') display.GUIDisplay(data_to_display, cal=p) if event.key == '3': data_to_display = data.SMGData.load(self.datastruct, 'FTISMHexp') display.GUIDisplay(data_to_display) if event.key == '4': data_to_display = data.SMGData.load(self.datastruct, 'FTISMHsim') display.GUIDisplay(data_to_display) if event.key == '5': data_to_display = data.SMGData.load_g(self.datastruct,'Mask1', 'PhasegM') p = data.SMGData.load(self.datastruct, 'p') display.GUIDisplay(data_to_display, cal=p) if event.key == '6': data_to_display = data.SMGData.load_g(self.datastruct, 'Mask2', 'PhasegM') p = data.SMGData.load(self.datastruct, 'p') display.GUIDisplay(data_to_display, cal=p) if event.key == '7': data_to_display = data.SMGData.load(self.datastruct,'Exx') p = data.SMGData.load(self.datastruct, 'p') display.GUIDisplay(data_to_display, cal=p) if event.key == '8': data_to_display = data.SMGData.load(self.datastruct,'Eyy') p = data.SMGData.load(self.datastruct, 'p') display.GUIDisplay(data_to_display, cal=p) if event.key == '9': data_to_display = data.SMGData.load(self.datastruct,'Exy') p = data.SMGData.load(self.datastruct, 'p') display.GUIDisplay(data_to_display, cal=p) if event.key == '0': data_to_display = data.SMGData.load(self.datastruct,'Rxy') p = data.SMGData.load(self.datastruct, 'p') display.GUIDisplay(data_to_display, cal=p)
slimpotatoes/STEM_Moire_GPA
src/guimaskmanager.py
# Module Mask Manager that is used by GUI from matplotlib.patches import Circle import matplotlib.artist as artist import math class MaskCreator(object): def __init__(self, axis, image): self.axis = axis self.image = image self.circle = None self.colored = None self.off_center = None def make_circle(self, mask_id, colored='r', off_center=(0, 0)): self.colored = colored self.off_center = off_center """Create circle gui object""" self.circle = Circle((self.image.shape[0]/2 + self.off_center[0], self.image.shape[1]/2+self.off_center[1]), self.image.shape[0]/6, color=self.colored, fill=True, alpha=0.3, linewidth=3) circle_artist = self.axis.add_artist(self.circle) circle_artist.set_gid(mask_id) self.axis.figure.canvas.draw() return Circle.get_gid(self.circle), (self.circle.center, self.circle.radius), class MaskEditor(object): def __init__(self, artist): self.artist = artist self.press = None self.mask_selected = None self.circle = None self.cidpress = None self.cidrelease = None self.cidmotion = None def connect(self): self.cidpress = self.artist.figure.canvas.mpl_connect('button_press_event', self.on_press) self.cidrelease = self.artist.figure.canvas.mpl_connect('button_release_event', self.on_release) self.cidmotion = self.artist.figure.canvas.mpl_connect('motion_notify_event', self.on_motion) def on_press(self, event): if event.inaxes != self.artist.axes: return contains, attrd = self.artist.contains(event) if not contains: self.artist.fill = True self.artist.figure.canvas.draw() self.mask_selected = None return (x0, y0) = self.artist.center self.press = x0, y0, event.xdata, event.ydata self.artist.fill = False self.artist.figure.canvas.draw() #print(artist.Artist.get_gid(self.artist)) self.mask_selected = artist.Artist.get_gid(self.artist) def on_motion(self, event): if self.press is None: return if event.inaxes != self.artist.axes: return if event.button == 1: x0, y0, xpress, ypress = self.press dx = event.xdata - xpress dy = event.ydata - ypress self.artist.center = x0 + dx, y0 + dy if event.button == 3: x0, y0, xpress, ypress = self.press r = math.sqrt((event.xdata-x0) ** 2 + (event.ydata-y0) ** 2) self.artist.set_radius(r) self.artist.figure.canvas.draw() def on_release(self, event): self.press = None self.artist.figure.canvas.draw() def disconnect(self): self.artist.figure.canvas.mpl_disconnect(self.cidpress) self.artist.figure.canvas.mpl_disconnect(self.cidrelease) self.artist.figure.canvas.mpl_disconnect(self.cidmotion) def disconnect_edit(self): self.artist.figure.canvas.mpl_disconnect(self.cidmotion)
slimpotatoes/STEM_Moire_GPA
src/test_straincalc.py
import pytest import straincalc import numpy as np import data # ---------------- # Test cases # --------------- # Test 27 from TestPlan.pdf datastruct_0 = data.SMGData() datastruct_0.create_branch('Mask1') datastruct_0.create_branch('Mask2') datastruct_0.store('p', 1) g_c_uns_1 = np.transpose(np.array([[[1, 2]]]), axes=(2, 0, 1)) g_c_uns_2 = np.transpose(np.array([[[3, 4]]]), axes=(2, 0, 1)) delta_g_1 = np.transpose(np.array([[[0, 0]]]), axes=(2, 0, 1)) delta_g_2 = np.transpose(np.array([[[0, 0]]]), axes=(2, 0, 1)) datastruct_0.store_g('Mask1', 'gCuns', g_c_uns_1) datastruct_0.store_g('Mask2', 'gCuns', g_c_uns_2) datastruct_0.store_g('Mask1', 'deltagM', delta_g_1) datastruct_0.store_g('Mask2', 'deltagM', delta_g_2) test_case_0 = dict() test_case_0['datastruct'] = datastruct_0 test_case_0['to_assert_strain'] = np.array([[[0]], [[0]], [[0]], [[0]]]) # Test 28 from TestPlan.pdf datastruct_1 = data.SMGData() datastruct_1.create_branch('Mask1') datastruct_1.create_branch('Mask2') datastruct_1.store('p', 1) g_c_uns_1 = np.transpose(np.array([[[1, 0]]]), axes=(2, 0, 1)) g_c_uns_2 = np.transpose(np.array([[[0, 1]]]), axes=(2, 0, 1)) delta_g_1 = np.transpose(np.array([[[-0.1, 0]]]), axes=(2, 0, 1)) delta_g_2 = np.transpose(np.array([[[0, 0]]]), axes=(2, 0, 1)) datastruct_1.store_g('Mask1', 'gCuns', g_c_uns_1) datastruct_1.store_g('Mask2', 'gCuns', g_c_uns_2) datastruct_1.store_g('Mask1', 'deltagM', delta_g_1) datastruct_1.store_g('Mask2', 'deltagM', delta_g_2) test_case_1 = dict() test_case_1['datastruct'] = datastruct_1 test_case_1['to_assert_strain'] = np.array([[[0]], [[1/9]], [[0]], [[0]]]) datastruct_2 = data.SMGData() datastruct_2.create_branch('Mask1') datastruct_2.create_branch('Mask2') datastruct_2.store('p', 1) g_c_uns_1 = np.transpose(np.array([[[1, 0]]]), axes=(2, 0, 1)) g_c_uns_2 = np.transpose(np.array([[[0, 1]]]), axes=(2, 0, 1)) delta_g_1 = np.transpose(np.array([[[0, 0]]]), axes=(2, 0, 1)) delta_g_2 = np.transpose(np.array([[[0, 0.1]]]), axes=(2, 0, 1)) datastruct_2.store_g('Mask1', 'gCuns', g_c_uns_1) datastruct_2.store_g('Mask2', 'gCuns', g_c_uns_2) datastruct_2.store_g('Mask1', 'deltagM', delta_g_1) datastruct_2.store_g('Mask2', 'deltagM', delta_g_2) test_case_2 = dict() test_case_2['datastruct'] = datastruct_2 test_case_2['to_assert_strain'] = np.array([[[-1/11]], [[0]], [[0]], [[0]]]) datastruct_3 = data.SMGData() datastruct_3.create_branch('Mask1') datastruct_3.create_branch('Mask2') datastruct_3.store('p', 1) g_c_uns_1 = np.transpose(np.array([[[1, 0]]]), axes=(2, 0, 1)) g_c_uns_2 = np.transpose(np.array([[[0, 1]]]), axes=(2, 0, 1)) delta_g_1 = np.transpose(np.array([[[0.1, 0]]]), axes=(2, 0, 1)) delta_g_2 = np.transpose(np.array([[[0, -0.1]]]), axes=(2, 0, 1)) datastruct_3.store_g('Mask1', 'gCuns', g_c_uns_1) datastruct_3.store_g('Mask2', 'gCuns', g_c_uns_2) datastruct_3.store_g('Mask1', 'deltagM', delta_g_1) datastruct_3.store_g('Mask2', 'deltagM', delta_g_2) test_case_3 = dict() test_case_3['datastruct'] = datastruct_3 test_case_3['to_assert_strain'] = np.array([[[1/9]], [[-1/11]], [[0]], [[0]]]) datastruct_4 = data.SMGData() datastruct_4.create_branch('Mask1') datastruct_4.create_branch('Mask2') datastruct_4.store('p', 1) g_c_uns_1 = np.transpose(np.array([[[1, 0]]]), axes=(2, 0, 1)) g_c_uns_2 = np.transpose(np.array([[[0, 1]]]), axes=(2, 0, 1)) delta_g_1 = np.transpose(np.array([[[0, 0.01]]]), axes=(2, 0, 1)) delta_g_2 = np.transpose(np.array([[[0.01, 0]]]), axes=(2, 0, 1)) datastruct_4.store_g('Mask1', 'gCuns', g_c_uns_1) datastruct_4.store_g('Mask2', 'gCuns', g_c_uns_2) datastruct_4.store_g('Mask1', 'deltagM', delta_g_1) datastruct_4.store_g('Mask2', 'deltagM', delta_g_2) test_case_4 = dict() test_case_4['datastruct'] = datastruct_4 test_case_4['to_assert_strain'] = np.array([[[0]], [[0]], [[1/0.9999*1/100]], [[0]]]) datastruct_5 = data.SMGData() datastruct_5.create_branch('Mask1') datastruct_5.create_branch('Mask2') datastruct_5.store('p', 1) g_c_uns_1 = np.transpose(np.array([[[1, 0]]]), axes=(2, 0, 1)) g_c_uns_2 = np.transpose(np.array([[[0, 1]]]), axes=(2, 0, 1)) delta_g_1 = np.transpose(np.array([[[0, -0.01]]]), axes=(2, 0, 1)) delta_g_2 = np.transpose(np.array([[[0.01, 0]]]), axes=(2, 0, 1)) datastruct_5.store_g('Mask1', 'gCuns', g_c_uns_1) datastruct_5.store_g('Mask2', 'gCuns', g_c_uns_2) datastruct_5.store_g('Mask1', 'deltagM', delta_g_1) datastruct_5.store_g('Mask2', 'deltagM', delta_g_2) test_case_5 = dict() test_case_5['datastruct'] = datastruct_5 test_case_5['to_assert_strain'] = np.array([[[0]], [[0]], [[0]], [[-1/1.0001*1/100]]]) test_cases = [test_case_0, test_case_1, test_case_2, test_case_3, test_case_4, test_case_5] @pytest.mark.parametrize("test_case", test_cases) def test_strain_calc_easy(test_case): straincalc.strain_calculation('Mask1', 'Mask2', test_case['datastruct']) results = np.array([test_case['datastruct'].load('Exx'), test_case['datastruct'].load('Eyy'), test_case['datastruct'].load('Exy'), test_case['datastruct'].load('Rxy')]) print(results) print(test_case['to_assert_strain']) assert np.all(np.isclose(results, test_case['to_assert_strain'], atol=0.001))
slimpotatoes/STEM_Moire_GPA
src/guilinemanager.py
<filename>src/guilinemanager.py import matplotlib.lines as mlines import matplotlib.pyplot as plt import numpy as np class LineDraw(object): epsilon = 20 def __init__(self, axis): self.axis = axis self.canvas = self.axis.figure.canvas self.LineStartx = None self.LineStarty = None self.LineEndx = None self.LineEndy = None self.LineCoords = np.empty((2, 2)) self.line = None self.Dragging = False self.vertex = 1 self.width = 1 self.WidthData = self.WidthDataCoords() def ConnectDraw(self): print('draw a line!') self.cidclick = self.canvas.mpl_connect('button_press_event', self.LineStart) self.cidmotion = self.canvas.mpl_connect('motion_notify_event', self.DrawLine) self.cidrelease = self.canvas.mpl_connect('button_release_event', self.LineEnd) self.ciddraw = self.canvas.mpl_connect('draw_event', self.DrawCanvas) # print(self.LineStart) def DisconnectDraw(self): self.canvas.mpl_disconnect(self.cidclick) self.canvas.mpl_disconnect(self.cidrelease) self.canvas.mpl_disconnect(self.ciddraw) self.canvas.mpl_disconnect(self.cidmotion) def ConnectMove(self): print('move your line!') self.cidendpick = self.canvas.mpl_connect('button_press_event', self.MoveLinePress) self.cidenddrag = self.canvas.mpl_connect('motion_notify_event', self.DrawLine) self.cidendrelease = self.canvas.mpl_connect('button_release_event', self.MoveLineUpdate) self.cidwidth = self.canvas.mpl_connect('scroll_event', self.ChangeWidth) self.cidenddraw = self.canvas.mpl_connect('draw_event', self.DrawCanvas) def DisconnectMove(self): self.canvas.mpl_disconnect(self.cidendpick) self.canvas.mpl_disconnect(self.cidenddrag) self.canvas.mpl_disconnect(self.cidendrelease) self.canvas.mpl_disconnect(self.cidenddraw) self.canvas.mpl_disconnect(self.cidwidth) def LineStart(self, event): if event.inaxes != self.axis: return self.LineCoords[0] = [event.xdata, event.ydata] self.line = mlines.Line2D(self.LineCoords[:, 0], self.LineCoords[:, 0], lw=self.WidthData, c='g', animated=True) self.axis.add_line(self.line) self.background = self.canvas.copy_from_bbox(self.axis.bbox) def LineEnd(self, event): if event.inaxes != self.axis: return self.LineCoords[1] = [event.xdata, event.ydata] self.CoordTransform = self.axis.transData.inverted() self.WidthData = self.WidthDataCoords() self.line = mlines.Line2D(self.LineCoords[:, 0], self.LineCoords[:, 1], lw=self.WidthData, c='g', marker='o', alpha=0.5, solid_capstyle='butt') self.axis.add_line(self.line) self.Dragging = True self.DisconnectDraw() self.ConnectMove() plt.draw() def DrawLine(self, event): if event.button != 1: return if event.inaxes is None: return if self.vertex == 0: x = (event.xdata, self.LineCoords[1, 0]) y = (event.ydata, self.LineCoords[1, 1]) elif self.vertex == 1: x = (self.LineCoords[0, 0], event.xdata) y = (self.LineCoords[0, 1], event.ydata) else: return self.line.set_data(x, y) self.canvas.restore_region(self.background) self.axis.draw_artist(self.line) self.canvas.blit(self.axis.bbox) self.canvas.draw() def DrawCanvas(self, event): self.background = self.canvas.copy_from_bbox(self.axis.bbox) if self.line is not None: self.axis.draw_artist(self.line) self.WidthData = self.WidthDataCoords() self.line.set_linewidth(self.WidthData) self.canvas.blit(self.axis.bbox) def MoveLinePress(self, event): self.vertex = self.GetPoint(event) # print(self.vertex) def MoveLineUpdate(self, event): if self.vertex is not None: self.LineCoords[self.vertex] = [event.xdata, event.ydata] self.line.set_data(self.LineCoords[:, 0], self.LineCoords[:, 1]) self.canvas.draw() def WidthDataCoords(self): diff0 = self.axis.transData.inverted().transform((1, 0)) diff1 = self.axis.transData.inverted().transform((2, 0)) diff = (diff1 - diff0) * self.width # print(diff) return diff[0] def ChangeWidth(self, event): if event.button == 'up': self.width += 1 elif event.button == 'down' and self.width > 1: self.width -= 1 else: return self.WidthData = self.WidthDataCoords() self.line.set_linewidth(self.WidthData) self.canvas.draw() def GetPoint(self, event): delta2 = np.sum((self.LineCoords - [event.xdata, event.ydata]) ** 2, axis=1) index = np.where(delta2 == np.min(delta2))[0] if delta2[index] >= self.epsilon ** 2: index = None return index
slimpotatoes/STEM_Moire_GPA
src/straincalc.py
# Strain calculation Module import numpy as np import data as data import rotatecalc def strain_calculation(mask_id_1, mask_id_2, datastruct): p = data.SMGData.load(datastruct, 'p') if p <= 0: raise Exception('Pixel size negative or zero, strain calculation cannot be performed') # Load the crystalline 3D arrays and their corresponding delta G 3D arrays and calibrate them with the pixel size p g_c_uns_1 = 1 / p * data.SMGData.load_g(datastruct, mask_id_1, 'gCuns') g_c_uns_2 = 1 / p * data.SMGData.load_g(datastruct, mask_id_2, 'gCuns') delta_g_1 = 1 / p * data.SMGData.load_g(datastruct, mask_id_1, 'deltagM') delta_g_2 = 1 / p * data.SMGData.load_g(datastruct, mask_id_2, 'deltagM') identity = np.ones(delta_g_1[0, :, :].shape) identity_image = np.array([[identity, np.zeros(identity.shape)], [np.zeros(identity.shape), identity]]) # Raw data '''g_ref = np.array([[g_c_uns_1[0, :, :], g_c_uns_2[0, :, :]], [g_c_uns_1[1, :, :], g_c_uns_2[1, :, :]]]) delta_g = np.array([[delta_g_1[0, :, :], delta_g_2[0, :, :]], [delta_g_1[1, :, :], delta_g_2[1, :, :]]])''' # Data rotated by 90 degrees R = ([0, -1],[1. 0]) '''g_ref = np.array([[g_c_uns_2[0, :, :], (-1) * g_c_uns_1[0, :, :]], [g_c_uns_2[1, :, :], (-1) * g_c_uns_1[1, :, :]]]) delta_g = np.array([[delta_g_2[0, :, :], (-1) * delta_g_1[0, :, :]], [delta_g_2[1, :, :], (-1) * delta_g_1[1, :, :]]])''' # Data rotated by 90 degrees R = ([0, 1],[-1. 0]) --- Base rotation not vector rotation '''g_ref = np.array([[(-1) * g_c_uns_2[0, :, :], g_c_uns_1[0, :, :]], [(-1) * g_c_uns_2[1, :, :], g_c_uns_1[1, :, :]]]) delta_g = np.array([[(-1) * delta_g_2[0, :, :], delta_g_1[0, :, :]], [(-1) * delta_g_2[1, :, :], delta_g_1[1, :, :]]])''' # Raw data adapted to base rotation (90 degree CCW) g_ref = np.array([[(-1) * g_c_uns_1[1, :, :], (-1) * g_c_uns_2[1, :, :]], [g_c_uns_1[0, :, :], g_c_uns_2[0, :, :]]]) delta_g = np.array([[(-1) * delta_g_1[1, :, :], (-1) * delta_g_2[1, :, :]], [delta_g_1[0, :, :], delta_g_2[0, :, :]]]) g = np.add(g_ref, delta_g) t_g_ref_pixel = np.transpose(np.transpose(g_ref, axes=[2, 3, 0, 1]), axes=[0, 1, 3, 2]) t_g_pixel = np.transpose(np.transpose(g, axes=[2, 3, 0, 1]), axes=[0, 1, 3, 2]) identity_pixel = np.transpose(identity_image, axes=[2, 3, 0, 1]) # Store intermediate values because of memory ?? LOL inv_t_g_pixel = np.linalg.inv(t_g_pixel) displacement = np.zeros(t_g_pixel.shape) for i in range(0, t_g_ref_pixel[:, :, 0, 0].shape[0]): for j in range(0, t_g_ref_pixel[:, :, 0, 0].shape[1]): displacement[i, j] = np.dot(inv_t_g_pixel[i, j], t_g_ref_pixel[i, j]) # Calculate gradient deformation tensor nabla(u) d = np.subtract(displacement, identity_pixel) # Calculate strain tensor epsilon = 0.5 * np.array(np.add(d, np.transpose(d, axes=[0, 1, 3, 2]))) omega = 0.5 * np.array(np.subtract(d, np.transpose(d, axes=[0, 1, 3, 2]))) # Put them in image format and store epsilon_image = np.transpose(epsilon, axes=[2, 3, 0, 1]) omega_image = np.transpose(omega, axes=[2, 3, 0, 1]) data.SMGData.store(datastruct, 'Exx', epsilon_image[0, 0]) data.SMGData.store(datastruct, 'Eyy', epsilon_image[1, 1]) data.SMGData.store(datastruct, 'Exy', epsilon_image[1, 0]) data.SMGData.store(datastruct, 'Rxy', omega_image[1, 0]) print('2D strain calculation done !')
victorwyee/great_expectations
great_expectations/expectations/metrics/map_metric.py
import uuid from functools import wraps from typing import Any, Callable, Dict, List, Optional, Tuple, Type, Union import numpy as np from great_expectations.core import ExpectationConfiguration from great_expectations.core.util import convert_to_json_serializable from great_expectations.exceptions.metric_exceptions import ( MetricError, MetricProviderError, ) from great_expectations.execution_engine import ExecutionEngine, PandasExecutionEngine from great_expectations.execution_engine.execution_engine import ( MetricDomainTypes, MetricFunctionTypes, MetricPartialFunctionTypes, ) from great_expectations.execution_engine.sparkdf_execution_engine import ( F, SparkDFExecutionEngine, ) from great_expectations.execution_engine.sqlalchemy_execution_engine import ( SqlAlchemyExecutionEngine, sa, ) from great_expectations.expectations.metrics.metric_provider import ( MetricProvider, metric_partial, ) from great_expectations.expectations.registry import ( get_metric_provider, register_metric, ) from great_expectations.validator.validation_graph import MetricConfiguration def column_function_partial( engine: Type[ExecutionEngine], partial_fn_type: str = None, **kwargs ): """Provides engine-specific support for authing a metric_fn with a simplified signature. A metric function that is decorated as a column_function_partial will be called with the engine-specific column type and any value_kwargs associated with the Metric for which the provider function is being declared. Args: engine: **kwargs: Returns: An annotated metric_function which will be called with a simplified signature. """ domain_type = MetricDomainTypes.COLUMN if issubclass(engine, PandasExecutionEngine): if partial_fn_type is None: partial_fn_type = MetricPartialFunctionTypes.MAP_SERIES partial_fn_type = MetricPartialFunctionTypes(partial_fn_type) if partial_fn_type != MetricPartialFunctionTypes.MAP_SERIES: raise ValueError( "PandasExecutionEngine only supports map_series for column_function_partial partial_fn_type" ) def wrapper(metric_fn: Callable): @metric_partial( engine=engine, partial_fn_type=partial_fn_type, domain_type=domain_type, **kwargs, ) @wraps(metric_fn) def inner_func( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], runtime_configuration: Dict, ): filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", False) ) ( df, compute_domain_kwargs, accessor_domain_kwargs, ) = execution_engine.get_compute_domain( domain_kwargs=metric_domain_kwargs, domain_type=domain_type ) if filter_column_isnull: df = df[df[accessor_domain_kwargs["column"]].notnull()] values = metric_fn( cls, df[accessor_domain_kwargs["column"]], **metric_value_kwargs, _metrics=metrics, ) return values, compute_domain_kwargs, accessor_domain_kwargs return inner_func return wrapper elif issubclass(engine, SqlAlchemyExecutionEngine): if partial_fn_type is None: partial_fn_type = MetricPartialFunctionTypes.MAP_FN partial_fn_type = MetricPartialFunctionTypes(partial_fn_type) if partial_fn_type not in [MetricPartialFunctionTypes.MAP_FN]: raise ValueError( "SqlAlchemyExecutionEngine only supports map_fn for column_function_partial partial_fn_type" ) def wrapper(metric_fn: Callable): @metric_partial( engine=engine, partial_fn_type=partial_fn_type, domain_type=domain_type, **kwargs, ) @wraps(metric_fn) def inner_func( cls, execution_engine: "SqlAlchemyExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], runtime_configuration: Dict, ): filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", False) ) if filter_column_isnull: compute_domain_kwargs = execution_engine.add_column_row_condition( metric_domain_kwargs ) else: # We do not copy here because if compute domain is different, it will be copied by get_compute_domain compute_domain_kwargs = metric_domain_kwargs ( selectable, compute_domain_kwargs, accessor_domain_kwargs, ) = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type=domain_type ) column_name = accessor_domain_kwargs["column"] dialect = execution_engine.dialect_module column_function = metric_fn( cls, sa.column(column_name), **metric_value_kwargs, _dialect=dialect, _table=selectable, _metrics=metrics, ) return column_function, compute_domain_kwargs, accessor_domain_kwargs return inner_func return wrapper elif issubclass(engine, SparkDFExecutionEngine): if partial_fn_type is None: partial_fn_type = MetricPartialFunctionTypes.MAP_FN partial_fn_type = MetricPartialFunctionTypes(partial_fn_type) if partial_fn_type not in [ MetricPartialFunctionTypes.MAP_FN, MetricPartialFunctionTypes.WINDOW_FN, ]: raise ValueError( "SparkDFExecutionEngine only supports map_fn and window_fn for column_function_partial partial_fn_type" ) def wrapper(metric_fn: Callable): @metric_partial( engine=engine, partial_fn_type=partial_fn_type, domain_type=domain_type, **kwargs, ) @wraps(metric_fn) def inner_func( cls, execution_engine: "SparkDFExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], runtime_configuration: Dict, ): filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", False) ) if filter_column_isnull: compute_domain_kwargs = execution_engine.add_column_row_condition( metric_domain_kwargs ) else: # We do not copy here because if compute domain is different, it will be copied by get_compute_domain compute_domain_kwargs = metric_domain_kwargs ( data, compute_domain_kwargs, accessor_domain_kwargs, ) = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type=domain_type ) column_name = accessor_domain_kwargs["column"] column_function = metric_fn( cls, column=data[column_name], **metric_value_kwargs, _metrics=metrics, _compute_domain_kwargs=compute_domain_kwargs, ) return column_function, compute_domain_kwargs, accessor_domain_kwargs return inner_func return wrapper else: raise ValueError("Unsupported engine for column_function_partial") def column_condition_partial( engine: Type[ExecutionEngine], partial_fn_type: Optional[Union[str, MetricPartialFunctionTypes]] = None, **kwargs, ): """Provides engine-specific support for authing a metric_fn with a simplified signature. A column_condition_partial must provide a map function that evalues to a boolean value; it will be used to provide supplemental metrics, such as the unexpected_value count, unexpected_values, and unexpected_rows. A metric function that is decorated as a column_condition_partial will be called with the engine-specific column type and any value_kwargs associated with the Metric for which the provider function is being declared. Args: engine: **kwargs: Returns: An annotated metric_function which will be called with a simplified signature. """ domain_type = MetricDomainTypes.COLUMN if issubclass(engine, PandasExecutionEngine): if partial_fn_type is None: partial_fn_type = MetricPartialFunctionTypes.MAP_CONDITION_SERIES partial_fn_type = MetricPartialFunctionTypes(partial_fn_type) if partial_fn_type not in [MetricPartialFunctionTypes.MAP_CONDITION_SERIES]: raise ValueError( "PandasExecutionEngine only supports map_condition_series for column_condition_partial partial_fn_type" ) def wrapper(metric_fn: Callable): @metric_partial( engine=engine, partial_fn_type=partial_fn_type, domain_type=domain_type, **kwargs, ) @wraps(metric_fn) def inner_func( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], runtime_configuration: Dict, ): filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", True) ) ( df, compute_domain_kwargs, accessor_domain_kwargs, ) = execution_engine.get_compute_domain( domain_kwargs=metric_domain_kwargs, domain_type=domain_type ) if filter_column_isnull: df = df[df[accessor_domain_kwargs["column"]].notnull()] meets_expectation_series = metric_fn( cls, df[accessor_domain_kwargs["column"]], **metric_value_kwargs, _metrics=metrics, ) return ( ~meets_expectation_series, compute_domain_kwargs, accessor_domain_kwargs, ) return inner_func return wrapper elif issubclass(engine, SqlAlchemyExecutionEngine): if partial_fn_type is None: partial_fn_type = MetricPartialFunctionTypes.MAP_CONDITION_FN partial_fn_type = MetricPartialFunctionTypes(partial_fn_type) if partial_fn_type not in [ MetricPartialFunctionTypes.MAP_CONDITION_FN, MetricPartialFunctionTypes.WINDOW_CONDITION_FN, ]: raise ValueError( "SqlAlchemyExecutionEngine only supports map_condition_fn for column_condition_partial partial_fn_type" ) def wrapper(metric_fn: Callable): @metric_partial( engine=engine, partial_fn_type=partial_fn_type, domain_type=domain_type, **kwargs, ) @wraps(metric_fn) def inner_func( cls, execution_engine: "SqlAlchemyExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], runtime_configuration: Dict, ): filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", True) ) ( selectable, compute_domain_kwargs, accessor_domain_kwargs, ) = execution_engine.get_compute_domain( metric_domain_kwargs, domain_type=domain_type ) column_name = accessor_domain_kwargs["column"] dialect = execution_engine.dialect_module sqlalchemy_engine = execution_engine.engine expected_condition = metric_fn( cls, sa.column(column_name), **metric_value_kwargs, _dialect=dialect, _table=selectable, _sqlalchemy_engine=sqlalchemy_engine, _metrics=metrics, ) if filter_column_isnull: # If we "filter" (ignore) nulls then we allow null as part of our new expected condition unexpected_condition = sa.and_( sa.not_(sa.column(column_name).is_(None)), sa.not_(expected_condition), ) else: unexpected_condition = sa.not_(expected_condition) return ( unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs, ) return inner_func return wrapper elif issubclass(engine, SparkDFExecutionEngine): if partial_fn_type is None: partial_fn_type = MetricPartialFunctionTypes.MAP_CONDITION_FN partial_fn_type = MetricPartialFunctionTypes(partial_fn_type) if partial_fn_type not in [ MetricPartialFunctionTypes.MAP_CONDITION_FN, MetricPartialFunctionTypes.WINDOW_CONDITION_FN, ]: raise ValueError( "SparkDFExecutionEngine only supports map_condition_fn and window_condition_fn for column_condition_partial partial_fn_type" ) def wrapper(metric_fn: Callable): @metric_partial( engine=engine, partial_fn_type=partial_fn_type, domain_type=domain_type, **kwargs, ) @wraps(metric_fn) def inner_func( cls, execution_engine: "SparkDFExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], runtime_configuration: Dict, ): filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", True) ) ( data, compute_domain_kwargs, accessor_domain_kwargs, ) = execution_engine.get_compute_domain( domain_kwargs=metric_domain_kwargs, domain_type=domain_type ) column_name = accessor_domain_kwargs["column"] column = data[column_name] expected_condition = metric_fn( cls, column, **metric_value_kwargs, _table=data, _metrics=metrics, _compute_domain_kwargs=compute_domain_kwargs, _accessor_domain_kwargs=accessor_domain_kwargs, ) if partial_fn_type == MetricPartialFunctionTypes.WINDOW_CONDITION_FN: if filter_column_isnull: compute_domain_kwargs = ( execution_engine.add_column_row_condition( compute_domain_kwargs, column_name=column_name ) ) unexpected_condition = ~expected_condition else: if filter_column_isnull: unexpected_condition = column.isNotNull() & ~expected_condition else: unexpected_condition = ~expected_condition return ( unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs, ) return inner_func return wrapper else: raise ValueError("Unsupported engine for column_condition_partial") def _pandas_map_condition_unexpected_count( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """Returns unexpected count for MapExpectations""" return np.count_nonzero(metrics["unexpected_condition"][0]) def _pandas_column_map_condition_values( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """Return values from the specified domain that match the map-style metric in the metrics dictionary.""" ( boolean_map_unexpected_values, compute_domain_kwargs, accessor_domain_kwargs, ) = metrics["unexpected_condition"] df, _, _ = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) ### # NOTE: 20201111 - JPC - in the map_series / map_condition_series world (pandas), we # currently handle filter_column_isnull differently than other map_fn / map_condition # cases. ### filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", False) ) if filter_column_isnull: df = df[df[accessor_domain_kwargs["column"]].notnull()] if "column" in accessor_domain_kwargs: domain_values = df[accessor_domain_kwargs["column"]] else: raise ValueError( "_pandas_column_map_condition_values requires a column in accessor_domain_kwargs" ) result_format = metric_value_kwargs["result_format"] if result_format["result_format"] == "COMPLETE": return list(domain_values[boolean_map_unexpected_values == True]) else: return list( domain_values[boolean_map_unexpected_values == True][ : result_format["partial_unexpected_count"] ] ) def _pandas_column_map_series_and_domain_values( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """Return values from the specified domain that match the map-style metric in the metrics dictionary.""" ( boolean_map_unexpected_values, compute_domain_kwargs, accessor_domain_kwargs, ) = metrics["unexpected_condition"] ( map_series, compute_domain_kwargs_2, accessor_domain_kwargs_2, ) = metrics["metric_partial_fn"] assert ( compute_domain_kwargs == compute_domain_kwargs_2 ), "map_series and condition must have the same compute domain" assert ( accessor_domain_kwargs == accessor_domain_kwargs_2 ), "map_series and condition must have the same accessor kwargs" df, _, _ = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) ### # NOTE: 20201111 - JPC - in the map_series / map_condition_series world (pandas), we # currently handle filter_column_isnull differently than other map_fn / map_condition # cases. ### filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", False) ) if filter_column_isnull: df = df[df[accessor_domain_kwargs["column"]].notnull()] if "column" in accessor_domain_kwargs: domain_values = df[accessor_domain_kwargs["column"]] else: raise ValueError( "_pandas_column_map_series_and_domain_values requires a column in accessor_domain_kwargs" ) result_format = metric_value_kwargs["result_format"] if result_format["result_format"] == "COMPLETE": return ( list(domain_values[boolean_map_unexpected_values == True]), list(map_series[boolean_map_unexpected_values == True]), ) else: return ( list( domain_values[boolean_map_unexpected_values == True][ : result_format["partial_unexpected_count"] ] ), list( map_series[boolean_map_unexpected_values == True][ : result_format["partial_unexpected_count"] ] ), ) def _pandas_map_condition_index( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): ( boolean_mapped_unexpected_values, compute_domain_kwargs, accessor_domain_kwargs, ) = metrics.get("unexpected_condition") df, _, _ = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) ### # NOTE: 20201111 - JPC - in the map_series / map_condition_series world (pandas), we # currently handle filter_column_isnull differently than other map_fn / map_condition # cases. ### filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", False) ) if filter_column_isnull: df = df[df[accessor_domain_kwargs["column"]].notnull()] data = df[accessor_domain_kwargs["column"]] result_format = metric_value_kwargs["result_format"] if result_format["result_format"] == "COMPLETE": return list(df[boolean_mapped_unexpected_values == True].index) else: return list( df[boolean_mapped_unexpected_values == True].index[ : result_format["partial_unexpected_count"] ] ) def _pandas_column_map_condition_value_counts( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """Returns respective value counts for distinct column values""" ( boolean_mapped_unexpected_values, compute_domain_kwargs, accessor_domain_kwargs, ) = metrics.get("unexpected_condition") df, _, _ = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) ### # NOTE: 20201111 - JPC - in the map_series / map_condition_series world (pandas), we # currently handle filter_column_isnull differently than other map_fn / map_condition # cases. ### filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", False) ) if filter_column_isnull: df = df[df[accessor_domain_kwargs["column"]].notnull()] data = df[accessor_domain_kwargs["column"]] if "column" in accessor_domain_kwargs: domain_values = df[accessor_domain_kwargs["column"]] else: raise ValueError( "_pandas_column_map_condition_value_counts requires a column in accessor_domain_kwargs" ) result_format = metric_value_kwargs["result_format"] value_counts = None try: value_counts = domain_values[ boolean_mapped_unexpected_values == True ].value_counts() except ValueError: try: value_counts = ( domain_values[boolean_mapped_unexpected_values == True] .apply(tuple) .value_counts() ) except ValueError: pass if not value_counts: raise MetricError("Unable to compute value counts") if result_format["result_format"] == "COMPLETE": return value_counts else: return value_counts[result_format["partial_unexpected_count"]] def _pandas_map_condition_rows( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """Return values from the specified domain (ignoring the column constraint) that match the map-style metric in the metrics dictionary.""" ( boolean_mapped_unexpected_values, compute_domain_kwargs, accessor_domain_kwargs, ) = metrics.get("unexpected_condition") df, _, _ = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) ### # NOTE: 20201111 - JPC - in the map_series / map_condition_series world (pandas), we # currently handle filter_column_isnull differently than other map_fn / map_condition # cases. ### filter_column_isnull = kwargs.get( "filter_column_isnull", getattr(cls, "filter_column_isnull", False) ) if filter_column_isnull: df = df[df[accessor_domain_kwargs["column"]].notnull()] data = df[accessor_domain_kwargs["column"]] result_format = metric_value_kwargs["result_format"] if result_format["result_format"] == "COMPLETE": return df[boolean_mapped_unexpected_values == True] else: return df[boolean_mapped_unexpected_values == True][ result_format["partial_unexpected_count"] ] def _sqlalchemy_map_condition_unexpected_count_aggregate_fn( cls, execution_engine: "SqlAlchemyExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """Returns unexpected count for MapExpectations""" unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) return ( sa.func.sum( sa.case( [(unexpected_condition, 1)], else_=0, ) ), compute_domain_kwargs, accessor_domain_kwargs, ) def _sqlalchemy_map_condition_unexpected_count_value( cls, execution_engine: "SqlAlchemyExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """Returns unexpected count for MapExpectations. This is a *value* metric, which is useful for when the unexpected_condition is a window function. """ unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) (selectable, _, _,) = execution_engine.get_compute_domain( compute_domain_kwargs, domain_type="identity" ) temp_table_name: str = f"ge_tmp_{str(uuid.uuid4())[:8]}" if execution_engine.dialect == "mssql": # mssql expects all temporary table names to have a prefix '#' temp_table_name = f"#{temp_table_name}" with execution_engine.engine.begin(): metadata: sa.MetaData = sa.MetaData(execution_engine.engine) temp_table_obj: sa.Table = sa.Table( temp_table_name, metadata, sa.Column("condition", sa.Integer, primary_key=False, nullable=False), ) temp_table_obj.create(execution_engine.engine, checkfirst=True) count_case_statement: List[sa.sql.elements.Label] = [ sa.case( [ ( unexpected_condition, 1, ) ], else_=0, ).label("condition") ] inner_case_query: sa.sql.dml.Insert = temp_table_obj.insert().from_select( count_case_statement, sa.select(count_case_statement).select_from(selectable), ) execution_engine.engine.execute(inner_case_query) unexpected_count_query: sa.Select = ( sa.select( [ sa.func.sum(sa.column("condition")).label("unexpected_count"), ] ) .select_from(temp_table_obj) .alias("UnexpectedCountSubquery") ) unexpected_count = execution_engine.engine.execute( sa.select( [ unexpected_count_query.c.unexpected_count, ] ) ).scalar() return convert_to_json_serializable(unexpected_count) def _sqlalchemy_column_map_condition_values( cls, execution_engine: "SqlAlchemyExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """ Particularly for the purpose of finding unexpected values, returns all the metric values which do not meet an expected Expectation condition for ColumnMapExpectation Expectations. """ unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) (selectable, _, _,) = execution_engine.get_compute_domain( compute_domain_kwargs, domain_type="identity" ) result_format = metric_value_kwargs["result_format"] if "column" not in accessor_domain_kwargs: raise ValueError( "_sqlalchemy_column_map_condition_values requires a column in accessor_domain_kwargs" ) query = ( sa.select( [sa.column(accessor_domain_kwargs.get("column")).label("unexpected_values")] ) .select_from(selectable) .where(unexpected_condition) ) if result_format["result_format"] != "COMPLETE": query = query.limit(result_format["partial_unexpected_count"]) return [ val.unexpected_values for val in execution_engine.engine.execute(query).fetchall() ] def _sqlalchemy_column_map_condition_value_counts( cls, execution_engine: "SqlAlchemyExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """ Returns value counts for all the metric values which do not meet an expected Expectation condition for instances of ColumnMapExpectation. """ unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) (selectable, _, _,) = execution_engine.get_compute_domain( compute_domain_kwargs, domain_type="identity" ) if "column" not in accessor_domain_kwargs: raise ValueError( "_sqlalchemy_column_map_condition_value_counts requires a column in accessor_domain_kwargs" ) column = sa.column(accessor_domain_kwargs["column"]) return execution_engine.engine.execute( sa.select([column, sa.func.count(column)]) .select_from(selectable) .where(unexpected_condition) .group_by(column) ).fetchall() def _sqlalchemy_map_condition_rows( cls, execution_engine: "SqlAlchemyExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): """ Returns all rows of the metric values which do not meet an expected Expectation condition for instances of ColumnMapExpectation. """ unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) (selectable, _, _,) = execution_engine.get_compute_domain( compute_domain_kwargs, domain_type="identity" ) result_format = metric_value_kwargs["result_format"] query = ( sa.select([sa.text("*")]).select_from(selectable).where(unexpected_condition) ) if result_format["result_format"] != "COMPLETE": query = query.limit(result_format["partial_unexpected_count"]) return execution_engine.engine.execute(query).fetchall() def _spark_map_condition_unexpected_count_aggregate_fn( cls, execution_engine: "SparkDFExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) return ( F.sum(F.when(unexpected_condition, 1).otherwise(0)), compute_domain_kwargs, accessor_domain_kwargs, ) def _spark_map_condition_unexpected_count_value( cls, execution_engine: "SparkDFExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): # fn_domain_kwargs maybe updated to reflect null filtering unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) (df, _, _) = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) data = df.withColumn("__unexpected", unexpected_condition) filtered = data.filter(F.col("__unexpected") == True).drop(F.col("__unexpected")) return filtered.count() def spark_column_map_condition_values( cls, execution_engine: "SparkDFExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) (df, _, _) = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) data = df.withColumn("__unexpected", unexpected_condition) if "column" not in accessor_domain_kwargs: raise ValueError( "spark_column_map_condition_values requires a column in accessor_domain_kwargs" ) column_name = accessor_domain_kwargs["column"] result_format = metric_value_kwargs["result_format"] filtered = data.filter(F.col("__unexpected") == True).drop(F.col("__unexpected")) if result_format["result_format"] == "COMPLETE": rows = filtered.select(F.col(column_name)).collect() else: rows = ( filtered.select(F.col(column_name)) .limit(result_format["partial_unexpected_count"]) .collect() ) return [row[column_name] for row in rows] def _spark_column_map_condition_value_counts( cls, execution_engine: "SparkDFExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) (df, _, _) = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) data = df.withColumn("__unexpected", unexpected_condition) if "column" not in accessor_domain_kwargs: raise ValueError( "spark_column_map_condition_values requires a column in accessor_domain_kwargs" ) column_name = accessor_domain_kwargs["column"] result_format = metric_value_kwargs["result_format"] filtered = data.filter(F.col("__unexpected") == True).drop(F.col("__unexpected")) value_counts = filtered.groupBy(F.col(column_name)).count() if result_format["result_format"] == "COMPLETE": rows = value_counts.collect() else: rows = value_counts.collect()[: result_format["partial_unexpected_count"]] return rows def _spark_map_condition_rows( cls, execution_engine: "PandasExecutionEngine", metric_domain_kwargs: Dict, metric_value_kwargs: Dict, metrics: Dict[str, Any], **kwargs, ): unexpected_condition, compute_domain_kwargs, accessor_domain_kwargs = metrics.get( "unexpected_condition" ) (df, _, _) = execution_engine.get_compute_domain( domain_kwargs=compute_domain_kwargs, domain_type="identity" ) data = df.withColumn("__unexpected", unexpected_condition) result_format = metric_value_kwargs["result_format"] filtered = data.filter(F.col("__unexpected") == True).drop(F.col("__unexpected")) if result_format["result_format"] == "COMPLETE": return filtered.collect() else: return filtered.limit(result_format["partial_unexpected_count"]).collect() class MapMetricProvider(MetricProvider): condition_domain_keys = ( "batch_id", "table", "row_condition", "condition_parser", ) function_domain_keys = ( "batch_id", "table", "row_condition", "condition_parser", ) condition_value_keys = tuple() function_value_keys = tuple() filter_column_isnull = True @classmethod def _register_metric_functions(cls): if not hasattr(cls, "function_metric_name") and not hasattr( cls, "condition_metric_name" ): return for attr, candidate_metric_fn in cls.__dict__.items(): if not hasattr(candidate_metric_fn, "metric_engine"): # This is not a metric continue metric_fn_type = getattr(candidate_metric_fn, "metric_fn_type") engine = candidate_metric_fn.metric_engine if not issubclass(engine, ExecutionEngine): raise ValueError( "metric functions must be defined with an Execution Engine" ) if metric_fn_type in [ MetricPartialFunctionTypes.MAP_CONDITION_SERIES, MetricPartialFunctionTypes.MAP_CONDITION_FN, MetricPartialFunctionTypes.WINDOW_CONDITION_FN, ]: if not hasattr(cls, "condition_metric_name"): raise ValueError( "A MapMetricProvider must have a metric_condition_name to have a decorated column_condition_partial method." ) condition_provider = candidate_metric_fn metric_name = cls.condition_metric_name metric_domain_keys = cls.condition_domain_keys metric_value_keys = cls.condition_value_keys metric_definition_kwargs = getattr( condition_provider, "metric_definition_kwargs", dict() ) domain_type = getattr( condition_provider, "domain_type", metric_definition_kwargs.get( "domain_type", MetricDomainTypes.TABLE ), ) if issubclass(engine, PandasExecutionEngine): register_metric( metric_name=metric_name + ".condition", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=condition_provider, metric_fn_type=metric_fn_type, ) register_metric( metric_name=metric_name + ".unexpected_count", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=_pandas_map_condition_unexpected_count, metric_fn_type=MetricFunctionTypes.VALUE, ) register_metric( metric_name=metric_name + ".unexpected_index_list", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_pandas_map_condition_index, metric_fn_type=MetricFunctionTypes.VALUE, ) register_metric( metric_name=metric_name + ".unexpected_rows", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_pandas_map_condition_rows, metric_fn_type=MetricFunctionTypes.VALUE, ) if domain_type == MetricDomainTypes.COLUMN: register_metric( metric_name=metric_name + ".unexpected_values", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_pandas_column_map_condition_values, metric_fn_type=MetricFunctionTypes.VALUE, ) register_metric( metric_name=metric_name + ".unexpected_value_counts", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_pandas_column_map_condition_value_counts, metric_fn_type=MetricFunctionTypes.VALUE, ) elif issubclass(engine, SqlAlchemyExecutionEngine): register_metric( metric_name=metric_name + ".condition", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=condition_provider, metric_fn_type=metric_fn_type, ) if metric_fn_type == MetricPartialFunctionTypes.MAP_CONDITION_FN: register_metric( metric_name=metric_name + ".unexpected_count.aggregate_fn", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=_sqlalchemy_map_condition_unexpected_count_aggregate_fn, metric_fn_type=MetricPartialFunctionTypes.AGGREGATE_FN, ) register_metric( metric_name=metric_name + ".unexpected_count", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=None, metric_fn_type=MetricFunctionTypes.VALUE, ) elif ( metric_fn_type == MetricPartialFunctionTypes.WINDOW_CONDITION_FN ): register_metric( metric_name=metric_name + ".unexpected_count", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=_sqlalchemy_map_condition_unexpected_count_value, metric_fn_type=MetricFunctionTypes.VALUE, ) register_metric( metric_name=metric_name + ".unexpected_rows", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_sqlalchemy_map_condition_rows, metric_fn_type=MetricFunctionTypes.VALUE, ) if domain_type == MetricDomainTypes.COLUMN: register_metric( metric_name=metric_name + ".unexpected_values", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_sqlalchemy_column_map_condition_values, metric_fn_type=MetricFunctionTypes.VALUE, ) register_metric( metric_name=metric_name + ".unexpected_value_counts", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_sqlalchemy_column_map_condition_value_counts, metric_fn_type=MetricFunctionTypes.VALUE, ) elif issubclass(engine, SparkDFExecutionEngine): register_metric( metric_name=metric_name + ".condition", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=condition_provider, metric_fn_type=metric_fn_type, ) if metric_fn_type == MetricPartialFunctionTypes.MAP_CONDITION_FN: register_metric( metric_name=metric_name + ".unexpected_count.aggregate_fn", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=_spark_map_condition_unexpected_count_aggregate_fn, metric_fn_type=MetricPartialFunctionTypes.AGGREGATE_FN, ) register_metric( metric_name=metric_name + ".unexpected_count", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=None, metric_fn_type=MetricFunctionTypes.VALUE, ) elif ( metric_fn_type == MetricPartialFunctionTypes.WINDOW_CONDITION_FN ): register_metric( metric_name=metric_name + ".unexpected_count", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=_spark_map_condition_unexpected_count_value, metric_fn_type=MetricFunctionTypes.VALUE, ) register_metric( metric_name=metric_name + ".unexpected_rows", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_spark_map_condition_rows, metric_fn_type=MetricFunctionTypes.VALUE, ) if domain_type == MetricDomainTypes.COLUMN: register_metric( metric_name=metric_name + ".unexpected_values", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=spark_column_map_condition_values, metric_fn_type=MetricFunctionTypes.VALUE, ) register_metric( metric_name=metric_name + ".unexpected_value_counts", metric_domain_keys=metric_domain_keys, metric_value_keys=(*metric_value_keys, "result_format"), execution_engine=engine, metric_class=cls, metric_provider=_spark_column_map_condition_value_counts, metric_fn_type=MetricFunctionTypes.VALUE, ) elif metric_fn_type in [ MetricPartialFunctionTypes.MAP_SERIES, MetricPartialFunctionTypes.MAP_FN, MetricPartialFunctionTypes.WINDOW_FN, ]: if not hasattr(cls, "function_metric_name"): raise ValueError( "A MapMetricProvider must have a function_metric_name to have a decorated column_function_partial method." ) map_function_provider = candidate_metric_fn metric_name = cls.function_metric_name metric_domain_keys = cls.function_domain_keys metric_value_keys = cls.function_value_keys register_metric( metric_name=metric_name + ".map", metric_domain_keys=metric_domain_keys, metric_value_keys=metric_value_keys, execution_engine=engine, metric_class=cls, metric_provider=map_function_provider, metric_fn_type=metric_fn_type, ) @classmethod def _get_evaluation_dependencies( cls, metric: MetricConfiguration, configuration: Optional[ExpectationConfiguration] = None, execution_engine: Optional[ExecutionEngine] = None, runtime_configuration: Optional[dict] = None, ): metric_name = metric.metric_name base_metric_value_kwargs = { k: v for k, v in metric.metric_value_kwargs.items() if k != "result_format" } dependencies = dict() metric_suffix = ".unexpected_count" if metric_name.endswith(metric_suffix): try: _ = get_metric_provider(metric_name + ".aggregate_fn", execution_engine) has_aggregate_fn = True except MetricProviderError: has_aggregate_fn = False if has_aggregate_fn: dependencies["metric_partial_fn"] = MetricConfiguration( metric_name + ".aggregate_fn", metric.metric_domain_kwargs, base_metric_value_kwargs, ) else: dependencies["unexpected_condition"] = MetricConfiguration( metric_name[: -len(metric_suffix)] + ".condition", metric.metric_domain_kwargs, base_metric_value_kwargs, ) # MapMetric uses the condition to build unexpected_count.aggregate_fn as well metric_suffix = ".unexpected_count.aggregate_fn" if metric_name.endswith(metric_suffix): dependencies["unexpected_condition"] = MetricConfiguration( metric_name[: -len(metric_suffix)] + ".condition", metric.metric_domain_kwargs, base_metric_value_kwargs, ) for metric_suffix in [ ".unexpected_values", ".unexpected_value_counts", ".unexpected_index_list", ".unexpected_rows", ]: if metric_name.endswith(metric_suffix): dependencies["unexpected_condition"] = MetricConfiguration( metric_name[: -len(metric_suffix)] + ".condition", metric.metric_domain_kwargs, base_metric_value_kwargs, ) try: _ = get_metric_provider(metric_name + ".map", execution_engine) dependencies["metric_map_fn"] = MetricConfiguration( metric_name + ".map", metric.metric_domain_kwargs, metric.metric_value_kwargs, ) except MetricProviderError: pass return dependencies class ColumnMapMetricProvider(MapMetricProvider): condition_domain_keys = ( "batch_id", "table", "column", "row_condition", "condition_parser", ) function_domain_keys = ( "batch_id", "table", "column", "row_condition", "condition_parser", ) condition_value_keys = tuple() function_value_keys = tuple()
hadi1976/python_project
codes/main.py
<reponame>hadi1976/python_project<filename>codes/main.py from typing import TextIO from pipes import TypeOfPipe import flowrate_factors import performances import plots from config import * """ Author: <NAME> Author: <NAME> The script "main.py" has the function of integrating the outputs from all the codes, through the implementation of functions. These functions are going to be called in the "GUI.py" code. """ """ First part of the code: Integrating "flowrate_factors" outputs through functions, for be referenced in the GUI. """ def consumption(): """ :return: int of the consumption depending on the selected country in (lit/sec.Capita). """ country_consumption = flowrate_factors.get_flow_discharge() return country_consumption def hourly(): """ :return: float of the flow discharge per hour in (lit/hour). """ flow_hourly = flowrate_factors.max_hourly_flow() return "%.2f" % flow_hourly def daily(): """ :return: float of the flow discharge per day in (lit/day). """ flow_daily = flowrate_factors.max_daily_flow() return "%.2f" % flow_daily """ Second part of the code: Instantiating the objects from the TypeOfPipe class. And, integrating this class outputs through functions, to be called in the GUI. """ # Reading the length, inputted by the user, from the GUI. with open(transmission_pipe_length) as file: transmission_pipe_length = float(file.read()) # Instantiating transmission_pipeline as an object from 'TypeOfPipe' class. transmission_pipeline = TypeOfPipe(flowrate_factors.max_daily_flow(), transmission_pipe_length) # Reading the length, inputted by the user, from the GUI. with open(distribution_pipe_length) as file: distribution_pipe_length = float(file.read()) # Instantiating distribution_pipeline as an object from 'TypeOfPipe' class. distribution_pipeline = TypeOfPipe(flowrate_factors.max_hourly_flow(), distribution_pipe_length) # Referencing the method 'pipe_dimensioning' from 'TypeOfPipe' class. transmission_dim = transmission_pipeline.pipe_dimensioning() distribution_dim = distribution_pipeline.pipe_dimensioning() # Integrating the pipes dimensions, for be referenced in the GUI. def diameter_transmission(): """ :return: int of the diameter of the transmission pipeline in (mm). """ return str(transmission_dim[4]) def velocity_transmission(): """ :return: float of the velocity of the transmission pipeline in (m/s). """ return str(transmission_dim[5]) def reynolds_transmission(): """ :return: float of the reynolds number of the transmission pipeline. """ return "%.2f" % transmission_dim[0] def friction_transmission(): """ :return: float of the friction factor of the transmission pipeline. """ return "%.2f" % transmission_dim[1] def slope_transmission(): """ :return: float of the slope of the transmission pipeline in (%). """ return "%.5f" % transmission_dim[2] def loss_transmission(): """ :return: float of the head loss of the transmission pipeline in (m). """ return "%.5f" % transmission_dim[3] def diameter_distribution(): """ :return: int of the diameter of the distribution pipeline in (mm). """ return str(distribution_dim[4]) def velocity_distribution(): """ :return: float of te velocity of the distribution pipeline in (m/s). """ return str(distribution_dim[5]) def reynolds_distribution(): """ :return: float of the reynolds number of the distribution pipeline. """ return "%.2f" % distribution_dim[0] def friction_distribution(): """ :return: float of the friction factor of the distribution pipeline. """ return "%.2f" % distribution_dim[1] def slope_distribution(): """ :return: float of the slope of the distribution pipeline in (%). """ return "%.5f" % distribution_dim[2] def head_distribution(): """ :return: float of the head loss of the distribution pipeline in (m). """ return "%.5f" % distribution_dim[3] """ Third part of the code: Pumps "Model 2009 1760 RPM" performances graphs: basic, parallel arrangement, series arrangement and combined arrangement. Also, the theoretical explanation for the user, about what can be seen in the plot. This two functions will be called in the GUI. """ # Conditions for pump one model 2009 1760 RPM curves performance. with open(user_elevation) as file: elevation = float(file.read()) def plot(): """ :return: The plot according thar corresponds to the user input and the fulfillment with the if/elif statements. """ # Conditions for pump one model 2009 1760 RPM curves performance. if performances.max_elevation_one >= elevation > 0 and \ performances.max_flow_one >= performances.flow_discharge > 0: return plots.plot_one() # Conditions for pump two model 8013 1760 RPM curves performance. elif performances.max_elevation_two >= elevation > 0 and \ performances.flow_discharge <= performances.max_flow_two: return plots.plot_two() # Conditions for pump two model 8013 1760 RPM parallel curves performance. elif performances.user_elevation > performances.max_elevation_two and \ float(performances.flow_discharge) < \ float(performances.max_flow_two) and\ performances.user_elevation > 0: return plots.plot_three() # Conditions for pump two model 8013 1760 RPM curves performance in series. elif performances.max_elevation_two > elevation > 0 and \ performances.flow_discharge > performances.max_flow_two: return plots.plot_four() # Conditions for pump two model 8013 1760 RPM curves performance combined. elif performances.user_elevation > performances.max_elevation_two and \ performances.flow_discharge > performances.max_flow_two and \ performances.user_elevation > 0: return plots.plot_five() def explanation(): """ :return: The explanation of the plot according thar corresponds to the user input and the fulfillment with the if/elif statements. """ # Conditions for pump one model 2009 1760 RPM curves performance. if performances.max_elevation_one >= elevation > 0 and \ performances.max_flow_one >= performances.flow_discharge > 0: return ("For an elevation of %d m and flow discharge of %d lit/sec," "is needed one pump of Model 2009 1760 RPM, with the " "performance shown in the graph." % (performances.user_elevation, performances.flow_discharge)) # Conditions for pump two model 8013 1760 RPM curves performance. elif performances.max_elevation_two >= elevation > 0 and \ performances.flow_discharge <= performances.max_flow_two: return ( "For an elevation of %d m and flow discharge of %d, " "is needed one pump of Model 8013 1760 RPM, with the " "performance shown in the graph." % (performances.user_elevation, performances.flow_discharge)) # Conditions pump two model 8013 1760 RPM curves performance in parallel. elif performances.user_elevation > performances.max_elevation_two and \ float(performances.flow_discharge) \ < float(performances.max_flow_two) and \ performances.user_elevation > 0: x = 1 for x in range(100000): # Breaks when arranged pumps elevation is higher than user one. if ( x * performances.max_elevation_two) > \ performances.user_elevation: break return ( "For an elevation of %d m and flow discharge of %d lit/sec, " "in the Model 8013 1760 RPM, %d pumps has to be arranged in" " parallel, in order to supply the demand. Please observe the " "graph shown for a better understanding." % (performances.user_elevation, performances.flow_discharge, x)) # Conditions for pump two model 8013 1760 RPM curves performance in series. elif performances.max_elevation_two > elevation > 0 and \ performances.flow_discharge > performances.max_flow_two: x = 1 for x in range(100000): # Breaks when arranged pumps flow is higher than the user one. if (x * performances.max_flow_two) > performances.flow_discharge: break return ("For an elevation of %d m and flow discharge of %d lit/sec, " "in the Model 8013 1760 RPM, %d pumps has to be arranged in " "series, in order to supply the demand. Please observe the " "graph shown for a better understanding." % (performances.user_elevation, performances.flow_discharge, x)) # Conditions for pump two model 8013 1760 RPM curves performance combined. elif performances.user_elevation > performances.max_elevation_two and \ performances.flow_discharge > performances.max_flow_two and \ performances.user_elevation > 0: x = 1 z = 1 # Iterating for finding the amount un pumps arranged in parallel. for x in range(100000): # Breaks when arranged pumps elevation is higher than user one. if ( x * performances.max_elevation_two) > \ performances.user_elevation: break for z in range(100000): # Breaks when arranged pumps flow is higher than the user one. if (z * performances.max_flow_two) > performances.flow_discharge: break return ("For an elevation of %d m and flow discharge of %d lit/sec, " "in the Model 8013 1760 RPM, %d pumps has to be arranged in " "parallel with %d series stages, in order to supply the " "demand. " "Please observe the graph " "shown for a better understanding." % (performances.user_elevation, performances.flow_discharge, x, z))
hadi1976/python_project
codes/plots.py
<filename>codes/plots.py from config import * import performances import pumps_figure """ Author: <NAME> The script "plots" as its name says, has the function of process the data input by the user, as the outputs from files as: "flowrate_factors", "pumps_figure", and "performances", in order to elaborate the graph according to the established conditions. """ def plot_one(): """ This function plots the pump one performance curves according to the best-fit elevation with the user input. Also, prints the pump model according to the elevation and flow discharge. param: a: int of the best-fit elevation in (m). :return: The best-fit plot for the pump one, according to the user inputs. """ # a is the int of the return from the function 'pump_one_performance' from # the file 'performances.py'. This elevation is in (m). a = performances.pump_one_performance() # Conditions for finding the best-fit curve performance and plot it. if a == 13: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate1.flow_range(), pumps_figure.Pump1.head_formula(), label="1.5HP(1.1kw)") elif a == 15: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate2.flow_range(), pumps_figure.Pump2.head_formula(), label="2HP(1.5kw)") elif a == 17: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate3.flow_range(), pumps_figure.Pump3.head_formula(), label="3HP(2.2kw)") elif a == 23: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate4.flow_range(), pumps_figure.Pump4.head_formula(), label="5HP(3.6kw)") elif a == 28: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate5.flow_range(), pumps_figure.Pump5.head_formula(), label="7.5HP(5.6kw)") # Plotting legend specifications. plt.legend() plt.xlabel("flow rate Lit/Sec") plt.ylabel("Head (m)") plt.title("Model 2009 1760 RPM") plt.savefig("pump3.png") plt.show() def plot_two(): """ This function plots the pump two, performances curves according to the best-fit elevation with the user input. Also, prints the pump model according to the elevation and flow discharge. param: b: int of the best-fit elevation in (m). :return: The best-fit plot for the pump two, according to the user inputs. """ # b is the int of the return from the function 'pump_two_performance' # from the file 'performances.py'. This elevation is in (m). b = performances.pump_two_performance() # Conditions for finding the best-fit curve performance and plot it. if b == 22: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate2_1.flow_range(), pumps_figure.Pump2_1.head_formula(), label="60HP(44kw)") elif b == 29: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate2_2.flow_range(), pumps_figure.Pump2_2.head_formula(), label="75HP(56kw)") elif b == 35: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate2_3.flow_range(), pumps_figure.Pump2_3.head_formula(), label="100HP(74.5kw)") elif b == 43: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate2_4.flow_range(), pumps_figure.Pump2_4.head_formula(), label="125HP(93.2kw)") elif b == 55: plt.figure(num=None, dpi=120) plt.plot(pumps_figure.flux_rate2_5.flow_range(), pumps_figure.Pump2_5.head_formula(), label="150HP(111.85kw)") # Plotting legend specifications. plt.legend() plt.xlabel("flow rate Lit/Sec") plt.ylabel("Head (m)") plt.title("Model 8013 1760 RPM") plt.savefig("pump4.png") plt.show() def plot_three(): """ This function plots the parallel arrangement, when the user input an elevation that is higher than the maximum elevation of the pump two curve performance. :return: The plot of the parallel arrangement. In case the user input an elevation higher than the maximum elevation capacity of the pump two. """ # Iterating for finding the amount un pumps arranged in parallel. for x in range(100000): # Breaks when arranged pumps elevation is higher than user one. if (x * performances.max_elevation_two) > performances.user_elevation: break equation_parallel = \ ((x + 1) * -0.0003) * \ (pumps_figure.flux_rate2_5.flow_range() ** 2) + \ ((x + 1) * 0.0521) * pumps_figure.flux_rate2_5.flow_range() \ + ((x + 1) * 53.481) # Plotting the arranged performance curves. plt.plot(pumps_figure.flux_rate2_5.flow_range(), equation_parallel, label="150HP(111.85kw)") # Plotting legend specifications. plt.legend() plt.xlabel("flow rate Lit/Sec") plt.ylabel("Head (m)") plt.title("Model 8013 1760 RPM") plt.savefig("pump5.png") plt.show() def plot_four(): """ This function plots the series arrangement, when the user input a flow discharge that is higher than the maximum flow of the pump two curve performance. :return: The plot of the series arrangement. In case the flow discharge is higher than the maximum flow capacity of the pump two. """ # Iterating for finding the amount un pumps arranged in series. for x in range(100000): # Breaks when arranged pumps flow is higher than the user one. if (x * performances.max_flow_two) > performances.flow_discharge: break equation_series = \ (-0.0003) * ((pumps_figure.flux_rate2_5.flow_range() * x) ** 2) + \ 0.0521 * (pumps_figure.flux_rate2_5.flow_range() * x) + 53.481 # Plotting the arranged performance curves. plt.plot((pumps_figure.flux_rate2_5.flow_range() * (x + 1)), equation_series, label="150HP(111.85kw)") # Plotting legend specifications. plt.legend() plt.xlabel("flow rate Lit/Sec") plt.ylabel("Head (m)") plt.title("Model 8013 1760 RPM") plt.savefig("pump6.png") plt.show() def plot_five(): """ This function plots the parallel and series arrangements, when the user input an elevation and the result of the 'flow_discharge.py' is a flow is higher than the maximum elevation and flow of the pump two curve performance. :return: The plot of the series and parallel arrangement. In case the elevation and flow discharge are higher than the maximum capacity of the pump two. """ # Iterating for finding the amount un pumps arranged in parallel. for x in range(100000): # Breaks when arranged pumps elevation is higher than the user one. if (x * performances.max_elevation_two) > performances.user_elevation: break equation_parallel_comb = \ (x * -0.0003) * (pumps_figure.flux_rate2_5.flow_range() ** 2) + \ (x * 0.0521) * pumps_figure.flux_rate2_5.flow_range() + ( x * 53.481) plt.plot(pumps_figure.flux_rate2_5.flow_range(), equation_parallel_comb, label="150HP(111.85kw)") # Iterating for finding the amount un pumps arranged in series. for z in range(100000): # Breaks when arranged pumps flow is higher than the user one. if (z * performances.max_flow_two) > performances.flow_discharge: break equation_series_comb = \ (-0.0003) * ((pumps_figure.flux_rate2_5.flow_range() * z) ** 2) + \ 0.0521 * (pumps_figure.flux_rate2_5.flow_range() * z) + 53.481 # Plotting the arranged performance curves. plt.plot((pumps_figure.flux_rate2_5.flow_range() * (z + 1)), equation_series_comb, label="150HP(111.85kw)") # Plotting legend specifications. plt.legend() plt.xlabel("flow rate Lit/Sec") plt.ylabel("Head (m)") plt.title("Model 8013 1760 RPM") plt.savefig("pump6.png") plt.show()
hadi1976/python_project
codes/diameters.py
<reponame>hadi1976/python_project<gh_stars>0 from config import* class DiameterReader: """ Author: <NAME> The class 'DiameterReader' has the function to read the 'EconomicDiameter' csv file. Making it reusable for others water distribution designing codes. """ def __init__(self, csv_file_name="EconomicDiameter.csv", delimiter=","): """ param csv_file_name: csv file with the economic diameter and velocity according to the flow discharge. param delimiter: separator in the csv file. """ self.sep = delimiter self.size_diameter_velocity = csv_file_name self.read_diameter_data(csv_file_name) def read_diameter_data(self, csv_file_name): """ Reading the csv file and creating a pandas dataframe. param: csv_file_name:csv file with the economic diameter and velocity according to the flow discharge. """ self.size_diameter_velocity = pd.read_csv(csv_file_name, header=0, sep=self.sep, usecols=[0, 1, 2], names=["Flow", "Diameter", "Velocity"] ) def __call__(self, csv_file_name, *args, **kwargs): print() """ It is used to make the object callable (as a function), so if we have an instance x that defines __call__(self, csv_file_name) we can do x(csv_file_name), which is actually a shortcut to x.__call__ (csv_file_name). """
hadi1976/python_project
codes/pipes.py
<filename>codes/pipes.py from config import * from diameters import DiameterReader class TypeOfPipe: """ Author: <NAME> The class 'TypeOfPipe' has the function of dimensioning the pipes parameters that are needed for supplying the demand in a sewer system. """ def __init__(self, flow_discharge, length): """ param: flow_discharge: float of the flow discharge (daily or hourly) in (lit/sec). param: length: int of the pipe length in (m). """ self.flow_discharge = float(flow_discharge) self.length = float(length) self.diameter = int # Variable returned in (mm), by the method two. self.velocity = float # Variable returned in (m/s), by the method two. def get_diameter_velocity_data(self): """ This method is iterating over the 'EconomicDiameter' csv file, which is in the class 'DiameterReader'. param: economic_diameter_csv: Instantiation of the DiameterReader class. param: df: 'size_diameter_velocity" dataframe. param: self.diameter: int of the diameter in (mm) read from the 'EconomicDiameters' csv. param: self.velocity: float of the velocity in (m/s) read from the 'EconomicDiameters' csv. """ # Instantiating DiameterReader class. economic_diameter_csv = DiameterReader() # Renaming the pandas dataframe "size_diameter_velocity" as "df". df = economic_diameter_csv.size_diameter_velocity # Iteration for finding the right diameter and velocity. for i in df.index: if self.flow_discharge <= df["Flow"][i]: self.diameter = df['Diameter'][i] self.velocity = df['Velocity'][i] break elif self.flow_discharge > df["Flow"][i]: self.diameter = 1000 self.velocity = 1.75 return self.diameter def pipe_dimensioning(self): """ The main function of this method is to use best-fit diameter from the previous method. Using it for dimensioning a pipe. param: diam: int of 'self.diameter' in (mm). param: reynolds: float of the Reynolds Number in the pipe. param: friction: float of the friction coefficient in the pipe. param: slope: float of the pipe distance slope in (%). param: head_loss: float of the head losses of the pipe in (m). """ diam = self.get_diameter_velocity_data() # Calculating Reynolds number of the pipe. reynolds = (1000 * ((self.flow_discharge / 1000) / ((math.pi * (diam / 1000) ** 2) / 4))) / 0.0089 # Calculating the friction factor of the pipe. friction = (1 / (-2 * math.log10(((0.0001 / (diam / 1000)) / 3.715) + (15 / reynolds)))) ** 2 # Calculating the slope of the distance traveled by the pipe. slope = friction * (((self.flow_discharge / 1000) / ((math.pi * (diam / 1000) ** 2) / 4)) / (2 * 9.82 * diam)) * 1000 # Calculating the losses in the pipe. head_loss = (slope / self.length) * 1000 # Parameters returned in a list format. return [reynolds, friction, slope, head_loss, self.diameter, self.velocity]
hadi1976/python_project
codes/performances.py
from flowrate_factors import max_daily_flow from config import* """" Author: <NAME> The script 'performances.py' has the function of evaluating the performances of the two pump capacities for the model 2009 1760 RPM and the model 8013 1760 RPM. This evaluation is according to: elevation (m) and flow discharge (lit/sec). """ with open(user_elevation) as file: user_elevation = int(file.read()) flow_discharge = max_daily_flow() # Output: flow_discharge. # Constant parameters of the pumps capacity performance. pump_one_elevation = [13, 15, 17, 23, 28] max_elevation_one = max(pump_one_elevation) pump_one_flow = [12.5, 14, 15, 16, 17.5] max_flow_one = max(pump_one_flow) pump_two_elevation = [22, 29, 35, 43, 55] max_elevation_two = max(pump_two_elevation) pump_two_flow = [155, 205, 240, 255, 270] max_flow_two = max(pump_two_flow) def pump_one_performance(): """ param: user_elevation: float of the elevation of the town in m. param: flow_discharge: float of the flow discharge in lit/sec. param: pump_one_elevation: int of the elevations of the pump one (m) regarding flow discharge (lit/sec). param: pump_one_flow: int of the flow discharge of the pump one (lit/sec) regarding elevations (m). param: max_elevation_one: int of the maximal elevation reached by pump one (m). param: max_flow_one: int of the maximal flow discharge reached by pump one (lit/sec). """ # Verifying that user inputs demand, can be supplied with the pump one. if user_elevation <= pump_one_elevation[-1] and flow_discharge <= \ pump_one_flow[-1]: # Iterating for finding the best-fit performance curve. i = 1 for i in range(len(pump_one_elevation)): if user_elevation < pump_one_elevation[i]: break j = 1 for j in range(len(pump_one_flow)): if flow_discharge < pump_one_flow[j]: break y = max(i, j) # Returning the elevation of the best-fits performing curve. return pump_one_elevation[y] def pump_two_performance(): """ param: pump_two_elevation: int of the elevations of the pump two (m) regarding flow discharge (lit/sec). param: pump_two_flow: int of the flow discharge of the pump two (lit/sec) regarding elevations (m). param: max_elevation_two: int of the maximal elevation reached by pump two (m). param: max_flow_two: int of the maximal flow discharge reached by pump two (lit/sec). """ # Verifying that user inputs demand, can be supplied with the pump two. if user_elevation <= pump_two_elevation[-1] and flow_discharge <= \ pump_two_flow[-1]: # Iterating for finding the best-fit performance curve. i = 1 for i in range(len(pump_two_elevation)): if user_elevation < pump_two_elevation[i]: break j = 1 for j in range(len(pump_two_flow)): if flow_discharge < pump_two_flow[j]: break y = max(i, j) # Returning the elevation of the best-fits performing curve. return pump_two_elevation[y]
hadi1976/python_project
codes/flowrate_factors.py
from config import * """ Author: <NAME> """ # read the country name from text file with open(country) as file: country_name = file.read() # read the number of inhabitants from text file with open(inhabitants) as file: number_inhabitants = int(file.read()) def get_flow_discharge(): """ this function read the xlsx file which contains of country names and the respective consumption value. Then it created a dictionary which the keys are country name and the value is flow discharge. after the user give us the name of the country it returns it's respective flow discharge :return Water consumption in Lit/day.capita """ dictionary_country = dict() wb = load_workbook(country_codes) ws = wb.active for row in range(2, ws.max_row + 1): if ws["B" + str(row)].value is None: pass else: dictionary_country[ws["B" + str(row)].value] = int( ws["C" + str(row)].value) return dictionary_country[country_name] def hour_factor(): """ This function give us the hourly factor which will be used in pipes design :return:hourly_factor """ if number_inhabitants <= 1000: hourly_factor = 5.66 return hourly_factor elif 1000 < number_inhabitants <= 10000: hourly_factor = 3.84 return hourly_factor elif 10000 < number_inhabitants <= 100000: hourly_factor = 2.61 return hourly_factor elif 100000 < number_inhabitants <= 1000000: hourly_factor = 1.77 return hourly_factor def day_factor(): """ This function give us the daily factor which will be used in max_daily_flow function. :return:daily_factor """ if number_inhabitants <= 1000: daily_factor = 2.32 return daily_factor elif 1000 < number_inhabitants <= 10000: daily_factor = 1.95 return daily_factor elif 10000 < number_inhabitants <= 100000: daily_factor = 1.64 return daily_factor elif 100000 < number_inhabitants <= 1000000: daily_factor = 1.38 return daily_factor def max_hourly_flow(): """ Calculation of Max hourly flow rate.This will be used in pipes design. :return: flow2 """ flow2 = ((get_flow_discharge() / 24) * number_inhabitants * hour_factor()) / 3600 return float(flow2) def max_daily_flow(): """ Calculation of Max daily flow rate. This will be used in pipes design. :return:flow1 """ flow1 = ((get_flow_discharge() * number_inhabitants * day_factor())) / 86400 return float(flow1)
hadi1976/python_project
codes/GUI.py
from config import * """ Author: <NAME> in this file I made a stand alone GUI which takes all the inputs from the user, save it in a text file, and show the results. """ class MyApp(tk.Frame): def __init__(self, master=None): tk.Frame.__init__(self, master) self.Label018 = None self.Label013 = None self.logo = None self.Label017 = None self.Label016 = None self.Label015 = None self.Label014 = None self.Label012 = None self.Label011 = None self.Label010 = None self.Label009 = None self.Label008 = None self.MyImage = None self.Label007 = None self.Label006 = None self.Label005 = None self.Label004 = None self.Label003 = None self.Label002 = None self.Label001 = None self.pop_up2 = None self.Label66 = None self.Label55 = None self.Label44 = None self.Label33 = None self.Label22 = None self.Label11 = None self.Label00 = None self.pop_up = None self.DistanceSD = None self.DistanceSS = None self.ElevationsStorage = float self.ElevationReservoir = float self.ElevationTown = float self.NumberOfInhabitants = int self.CountryName = None self.master.title("GUI project") self.master.iconbitmap("gui_pics/pipe.ico") # Set geometry: upper-left corner of the window ww = 1100 # width wh = 600 # height wx = (self.master.winfo_screenwidth() - ww) / 2 wy = (self.master.winfo_screenheight() - wh) / 2 # assign geometry self.master.geometry("%dx%d+%d+%d" % (ww, wh, wx, wy)) # assign space holders around widgets self.dx = 5 self.dy = 5 # Image logo = tk.PhotoImage(file="gui_pics/image.png") logo = logo.subsample(1, 1) # controls size self.l_img = tk.Label(master, image=logo) self.l_img.image = logo self.l_img.grid(row=11, column=1) # Label self.Label1 = tk.Label(master, text="Number of inhabitants") self.Label1.grid(column=0, row=1, padx=self.dx, pady=self.dy) self.Label2 = tk.Label(master, text="Elevation of the storage tank") self.Label2.grid(column=0, row=2, padx=self.dx, pady=self.dy) self.Label5 = tk.Label(master, text="Distance between Source and " "Storage tank") self.Label5.grid(column=0, row=5, padx=self.dx, pady=self.dy) self.Label6 = tk.Label(master, text="Distance between Storage tank " "and Distribution net") self.Label6.grid(column=0, row=6, padx=self.dx, pady=self.dy) # Entry # Number of Inhabitants self.Number_of_Inhabitants = tk.IntVar() self.Number_of_Inhabitants = tk.Entry(master, width=20) self.Number_of_Inhabitants.grid(column=1, row=1, padx=5, pady=5) # Elevation of the town self.Elevation_town = tk.IntVar() self.Elevation_town = tk.Entry(master, width=20) self.Elevation_town.grid(column=1, row=2, padx=5, pady=5) # Distance between Source and Storage tank self.Distance1 = tk.IntVar() self.Distance1 = tk.Entry(master, width=20) self.Distance1.grid(column=1, row=5, padx=5, pady=5) # Distance between Storage tank and Distribution net self.Distance2 = tk.IntVar() self.Distance2 = tk.Entry(master, width=20) self.Distance2.grid(column=1, row=6, padx=5, pady=5) # Combobox self.selected_country = tk.StringVar() self.cbox = ttk.Combobox(master, width=20, textvariable=self.selected_country) self.cbox.grid(column=0, row=0, padx=self.dx, pady=self.dy) self.cbox['state'] = 'readonly' with open(file="Countries.txt", mode="r") as ff: self.cbox['values'] = ff.read().splitlines() self.cbox.set("Country names") self.cbox_selection = self.cbox.get() # define Button to assign the values self.button_country = tk.Button(master, text="Assign", command=lambda: self.save_country_name()) self.button_country.grid(column=3, row=0, padx=self.dx, pady=self.dy) self.button1 = tk.Button(master, text="Assign", command=lambda: self.save_info1()) self.button1.grid(column=3, row=1, padx=self.dx, pady=self.dy) self.button2 = tk.Button(master, text="Assign", command=lambda: self.save_info2()) self.button2.grid(column=3, row=2, padx=self.dx, pady=self.dy) self.button5 = tk.Button(master, text="Assign", command=lambda: self.save_info5()) self.button5.grid(column=3, row=5, padx=self.dx, pady=self.dy) self.button6 = tk.Button(master, text="Assign", command=lambda: self.save_info6()) self.button6.grid(column=3, row=6, padx=self.dx, pady=self.dy) self.button7 = tk.Button(master, text="Check your Inputs", bg="red", fg="blue", command=lambda: self.check_inputs()) self.button7.grid(column=3, row=7, padx=5, pady=5) self.button8 = tk.Button(master, text="Show results", bg="pink", fg="blue", command=lambda: self.show_results()) self.button8.grid(column=3, row=8, padx=5, pady=5) # if the user wants to see the original pumps figures self.button9 = tk.Button(master, text="Show pumps", bg="black", fg="white", command=lambda: self.show_pumps()) self.button9.grid(column=3, row=9, padx=5, pady=5) self.button_quit = tk.Button(master, text="Quit", bg="yellow", fg="red", command=lambda: self.quit_gui()) self.button_quit.grid(column=3, row=10, padx=5, pady=5) def save_country_name(self): """ This method would insert the user's country in a txt file name, country.txt So when the user click on Assign button the country name wil be written in text file. """ try: with open("inputs/country.txt", "w") as file: self.CountryName = self.selected_country.get() file.write(self.CountryName) except OSError: showinfo("Error", "Please insert a country name.") def save_info1(self): """ This method would insert the user's number of inhabitants in a txt file name, Input1.txt .So when the user click on Assign button the input wil be written in text file. """ try: if type(int(self.Number_of_Inhabitants.get())) == int: if 0 < int(self.Number_of_Inhabitants.get()) < 1000000: with open("inputs/Input1.txt", "w") as file: self.NumberOfInhabitants = \ self.Number_of_Inhabitants.get() file.write(self.NumberOfInhabitants) else: showinfo("Error", "Please insert a number below one million ") except ValueError: showinfo("Error", "Please insert an integer number.") def save_info2(self): """ This method would insert the user's elevation town in a txt file name, Input2.txt .So when the user click on Assign button the input will be written in text file. """ try: if type(int(self.Elevation_town.get())) == int: if (int(self.Elevation_town.get())) > 0: with open("inputs/Input2.txt", "w") as file: self.ElevationTown = self.Elevation_town.get() file.write(self.ElevationTown) else: showinfo("Error", "Please insert a positive number.") except ValueError: showinfo("Error", "Please insert an integer number.") def save_info5(self): """ This method would insert the user's pipe length between the reservoir and tank in a txt file name, Input5.txt .So when the user click on Assign button the input will be written in text file. """ try: if type(int(self.Distance1.get())) == int: if (int(self.Distance1.get())) > 0: with open("inputs/Input5.txt", "w") as file: self.DistanceSS = self.Distance1.get() file.write(self.DistanceSS) else: showinfo("Error", "Please insert a positive number.") except ValueError: showinfo("Error", "Please insert an integer number.") def save_info6(self): """ This method would insert the user's pipe length between the tank and town in a txt file name, Input6.txt .So when the user click on Assign button the input will be written in text file. """ try: if type(int(self.Distance2.get())) == int: if (int(self.Distance2.get())) > 0: with open("inputs/Input6.txt", "w") as file: self.DistanceSD = self.Distance2.get() file.write(self.DistanceSD) else: showinfo("Error", "Please insert a positive number.") except ValueError: showinfo("Error", "Please insert an integer number") def check_inputs(self): """ In this method when the user click on the respective button(Check your Inputs), The program navigate to the text files were all the inputs are saved. And It shows the values that the user has entered. """ self.pop_up = tk.Toplevel(master=self) # Geometry of pop up window self.pop_up.geometry("400x400") self.pop_up.title("pop_up Window") # add a label for showing the selected country file1 = open("inputs/country.txt", "r") show_country = file1.read() file1.close() self.Label00 = tk.Label(self.pop_up, text="The name of the " "Country is: " + show_country) self.Label00.grid(column=0, row=1, padx=self.dx, pady=self.dy) # add a label for showing the selected country file2 = open("inputs/Input1.txt", "r") show_inhabitants = file2.read() file2.close() self.Label11 = tk.Label(self.pop_up, text="Number of " "inhabitants is: " + show_inhabitants) self.Label11.grid(column=0, row=2, padx=self.dx, pady=self.dy) # add a label for showing the Elevation town file3 = open("inputs/Input2.txt", "r") show_elevation = file3.read() file3.close() self.Label22 = tk.Label(self.pop_up, text="Elevation " "of the town is: " + show_elevation) self.Label22.grid(column=0, row=3, padx=self.dx, pady=self.dy) # add a label for showing the Distance between Reservoir and Tank file6 = open("inputs/Input5.txt", "r") show_distance1 = file6.read() file6.close() self.Label55 = tk.Label(self.pop_up, text="Distance between Source and " "Storage tank is: " + show_distance1) self.Label55.grid(column=0, row=6, padx=self.dx, pady=self.dy) # add a label for showing the Distance between Tank and Town file7 = open("inputs/Input6.txt", "r") show_distance2 = file7.read() file7.close() self.Label66 = tk.Label(self.pop_up, text="Distance between Storage" " tank and " "Distribution net is: " + show_distance2) self.Label66.grid(column=0, row=7, padx=self.dx, pady=self.dy) def show_results(self): """ In this Method when the user click on the respective button(Show results), It would import all the outputs that we want from this project, e.g. The consumption value, The hourly flow discharge, Diameter of the pipe,etc. """ from main import consumption, hourly, daily, diameter_transmission, \ velocity_transmission, reynolds_transmission, \ diameter_distribution, \ velocity_distribution, reynolds_distribution, \ friction_transmission, \ friction_distribution, slope_transmission, slope_distribution, \ loss_transmission, head_distribution, plot, explanation self.pop_up2 = tk.Toplevel(master=self) # Geometry of pop up window self.pop_up2.geometry("500x500") self.pop_up2.title("Show results") # Label for country consumption self.Label001 = tk.Label(self.pop_up2, text="The consumption value is: " + str(consumption()) + " Lit/day.Capita") self.Label001.grid(column=0, row=1, padx=self.dx, pady=self.dy) self.Label002 = tk.Label(self.pop_up2, text="The hourly flow discharge : " + hourly() + " Lit/hour") self.Label002.grid(column=0, row=2, padx=self.dx, pady=self.dy) self.Label003 = tk.Label(self.pop_up2, text="The daily flow discharge : " + daily() + " Lit/day") self.Label003.grid(column=0, row=3, padx=self.dx, pady=self.dy) # Label for asserting the transmission and distribution self.Label004 = tk.Label(self.pop_up2, text="Transmission pipeline ") self.Label004.grid(column=0, row=4, padx=self.dx, pady=self.dy) self.Label005 = tk.Label(self.pop_up2, text="Distribution pipeline ") self.Label005.grid(column=1, row=4, padx=self.dx, pady=self.dy) # Label Diameter for Transmission pipeline self.Label007 = tk.Label(self.pop_up2, text="Diameter: " + diameter_transmission() + " mm") self.Label007.grid(column=0, row=5, padx=self.dx, pady=self.dy) # Label Diameter for Distribution pipeline self.Label008 = tk.Label(self.pop_up2, text="Diameter: " + diameter_distribution() + " mm") self.Label008.grid(column=1, row=5, padx=self.dx, pady=self.dy) # Label for velocity number self.Label009 = tk.Label(self.pop_up2, text="Velocity: " + velocity_transmission() + " m/sec") self.Label009.grid(column=0, row=6, padx=self.dx, pady=self.dy) self.Label009 = tk.Label(self.pop_up2, text="Velocity: " + velocity_distribution() + " m/sec") self.Label009.grid(column=1, row=6, padx=self.dx, pady=self.dy) # Label for reynolds self.Label010 = tk.Label(self.pop_up2, text="Reynolds number: " + reynolds_transmission() + "") self.Label010.grid(column=0, row=7, padx=self.dx, pady=self.dy) self.Label011 = tk.Label(self.pop_up2, text="Reynolds number: " + reynolds_distribution() + "") self.Label011.grid(column=1, row=7, padx=self.dx, pady=self.dy) # Label Friction loss self.Label012 = tk.Label(self.pop_up2, text="Friction coefficient: " + friction_transmission() + "") self.Label012.grid(column=0, row=8, padx=self.dx, pady=self.dy) self.Label013 = tk.Label(self.pop_up2, text="Friction coefficient: " + friction_distribution() + "") self.Label013.grid(column=1, row=8, padx=self.dx, pady=self.dy) # Label Slope self.Label014 = tk.Label(self.pop_up2, text="Slope: " + slope_transmission() + "%") self.Label014.grid(column=0, row=9, padx=self.dx, pady=self.dy) self.Label015 = tk.Label(self.pop_up2, text="Slope: " + slope_distribution() + "%") self.Label015.grid(column=1, row=9, padx=self.dx, pady=self.dy) # Label Head loss self.Label016 = tk.Label(self.pop_up2, text="Head loss: " + loss_transmission() + "") self.Label016.grid(column=0, row=10, padx=self.dx, pady=self.dy) self.Label017 = tk.Label(self.pop_up2, text="Head loss: " + head_distribution() + "") self.Label017.grid(column=1, row=10, padx=self.dx, pady=self.dy) # plotting the pump that we need self.Label018 = tk.Label(self.pop_up2, text="Amount and type of " "pump required: " + explanation(), wraplength=250) self.Label018.grid(row=11, padx=self.dx, pady=self.dy) plot() @staticmethod def show_pumps(): from pumps_figure import plot_pump1, plot_pump2 plot_pump1() plot_pump2() def quit_gui(self): self.master.destroy() if __name__ == '__main__': MyApp().mainloop()
hadi1976/python_project
codes/config.py
""" Author: <NAME> The "config.py" file, is importing all the libraries and modules that are required for the correct functionality of this code.As well as, the path of the GUI inputs. Also, in case one of the libraries or modules is not properly installed or the path of the GUI cannot be found. The user will be able to realize that is an ImportError with its respective message. """ try: import os import logging import math except ImportError: logging.info("ERROR: Cannot import basic Python libraries.") try: import numpy as np import pandas as pd except ImportError: logging.info("ERROR: Cannot import SciPy libraries.") try: import matplotlib.pyplot as plt except ImportError: logging.info("ERROR: Cannot import Matplotlib libraries") try: import tkinter as tk from tkinter.messagebox import showinfo from tkinter import ttk except ImportError: logging.info("ERROR: Cannot import tkinter modules") try: from openpyxl import load_workbook except ImportError: logging.info("ERROR: Cannot import openpyxl libraries") try: p = os.path.abspath("") country = p + "/inputs/country.txt" country_codes = p + "/Country_Codes_and_Names.xlsx" inhabitants = p + "/inputs/Input1.txt" user_elevation = p + "/inputs/Input2.txt" transmission_pipe_length = p + "/inputs/Input5.txt" distribution_pipe_length = p + "/inputs/Input6.txt" except ImportError: logging.info("ERROR: Cannot import the input path") # Creating the log file. logging.basicConfig(filename="my-logfile.log", format="%(asctime)s - %(message)s", filemode="w", level=logging.DEBUG)
hadi1976/python_project
codes/pumps_figure.py
from config import * class FlowRatePump: """ Author:<NAME> This class "FlowRatePump" has 3 attributes;start, stop, name The Method is called flow_rate which gets the attributes and create a list of 500 numbers between stop and start. """ def __init__(self, start, stop, name): self.start = start self.stop = stop self.name = name def flow_range(self): """ It gets the start,stop point and generate a list of 500 numbers between them. :return: self.name """ self.name = np.linspace(self.start, self.stop, num=500) return self.name # Object of FlowRatePump class for the first pump flux_rate1 = FlowRatePump(0, 12, "flux_rate1") flux_rate2 = FlowRatePump(0, 12.5, "flux_rate2") flux_rate3 = FlowRatePump(0, 14, "flux_rate3") flux_rate4 = FlowRatePump(0, 16, "flux_rate4") flux_rate5 = FlowRatePump(0, 17.2, "flux_rate5") # Object of FlowRatePump class for the second pump flux_rate2_1 = FlowRatePump(0, 160, "flux_rate1") flux_rate2_2 = FlowRatePump(0, 200, "flux_rate2") flux_rate2_3 = FlowRatePump(0, 230, "flux_rate3") flux_rate2_4 = FlowRatePump(0, 250, "flux_rate4") flux_rate2_5 = FlowRatePump(0, 285, "flux_rate5") class HeadPump: """ author:<NAME> This class "HeadPump" create 6 attributes. It gives us the pumps' equation which is a 3rd grade equation. so the equation of our pump is; y=aX^3+bX^2+cX+d in this class our X is flux_rate1,flux_rate2,...flux_rate5. and y is our head pump. And the Method's name is head_formula which gets all the attributes and generate the respective equation. """ def __init__(self, a, b, c, d, flux_rate, name): self.a = a self.b = b self.c = c self.d = d self.name = name self.flux_rate = flux_rate def head_formula(self): """ It gets the coefficients of the formula and generate the formula :return: self.name """ self.name = self.a * (self.flux_rate ** 3) + self.b *\ (self.flux_rate ** 2) + self.c * ( self.flux_rate ** 1) + self.d * ( self.flux_rate ** 0) return self.name # Object of HeadPump class for the first pump Pump1 = HeadPump(-0.0061, -0.01, 0.015, 11.364, flux_rate1.flow_range(), "Pump1") Pump2 = HeadPump(-0.0051, 0.0121, - 0.0341, 14.619, flux_rate2.flow_range(), "Pump2") Pump3 = HeadPump(-0.0034, - 0.014, 0.1205, 18.116, flux_rate3.flow_range(), "Pump3") Pump4 = HeadPump(0, -0.0612, 0.3119, 22.021, flux_rate4.flow_range(), "Pump4") Pump5 = HeadPump(0, -0.0589, 0.3821, 26.731, flux_rate5.flow_range(), "Pump5") # Object of HeadPump class for the second pump Pump2_1 = HeadPump(0, -0.001, 0.07, 22.941, flux_rate2_1.flow_range(), "Pump1") Pump2_2 = HeadPump(0, -0.0006, 0.0648, 28.838, flux_rate2_2.flow_range(), "Pump2") Pump2_3 = HeadPump(0, -0.0006, 0.0779, 35.279, flux_rate2_3.flow_range(), "Pump3") Pump2_4 = HeadPump(0, -0.0004, 0.0521, 43.847, flux_rate2_4.flow_range(), "Pump4") Pump2_5 = HeadPump(0, -0.0003, 0.0521, 53.481, flux_rate2_5.flow_range(), "Pump5") def plot_pump1(): """ This function would plot all the pumps bg getting the flow_rate in X_axis and head pump in y_axis. """ plt.figure(num=None, dpi=120) plt.plot(flux_rate1.flow_range(), Pump1.head_formula(), label="1.5HP(1.1kw)") plt.plot(flux_rate2.flow_range(), Pump2.head_formula(), label="2HP(1.5kw)") plt.plot(flux_rate3.flow_range(), Pump3.head_formula(), label="3HP(2.2kw)") plt.plot(flux_rate4.flow_range(), Pump4.head_formula(), label="5HP(3.6kw)") plt.plot(flux_rate5.flow_range(), Pump5.head_formula(), label="7.5HP(5.6kw)") plt.legend() plt.xlabel("flow rate Lit/Sec") plt.ylabel("Head (m)") plt.title("Model 2009 1760 RPM") plt.savefig("pump1.png") plt.show() def plot_pump2(): """ This function would plot all the pumps bg getting the flow_rate in X_axis and head pump in y_axis. """ plt.figure(num=None, dpi=120) plt.plot(flux_rate2_1.flow_range(), Pump2_1.head_formula(), label="60HP(44kw)") plt.plot(flux_rate2_2.flow_range(), Pump2_2.head_formula(), label="75HP(56kw)") plt.plot(flux_rate2_3.flow_range(), Pump2_3.head_formula(), label="100HP(74.5kw)") plt.plot(flux_rate2_4.flow_range(), Pump2_4.head_formula(), label="125HP(93.2kw)") plt.plot(flux_rate2_5.flow_range(), Pump2_5.head_formula(), label="150HP(111.85kw)") plt.legend() plt.xlabel("flow rate Lit/Sec") plt.ylabel("Head (m)") plt.title("Model 8013 1760 RPM") plt.savefig("pump2.png") plt.show()
grim13b/testtone
sinewave.py
# -*- coding: utf-8 -*- __author__ = 'grim3lt.org' import numpy import struct class SineWaveCreator: def offset_header(self, fp): for i in range(44): fp.write(b'0') def write(self, fp, a, f0, fs, depth, sec): """ a = amplitude f0 = frequency fs = sampling frequency depth = bit depth sec = seconds """ if a > 1.0: return None dr = 2 ** depth / 2 - 1 for n in range(sec * fs): y = a * numpy.sin(2 * numpy.pi * f0 * n / fs) s = int(y * dr) fp.write(struct.pack('h', s)) fp.write(struct.pack('h', s)) def write_header(self, fp, fs, depth): data_chunk_size = fp.tell() fp.seek(0, 0) fp.write(b'RIFF') fp.write(struct.pack('I', data_chunk_size + 44)) # data size + format header size fp.write(b'WAVE') # WAVE format chunk Header fp.write(b'fmt ') # fmt chunk header fp.write(struct.pack('I', 16)) # size of chunk fp.write(struct.pack('H', 0x0001)) # liner pcm fp.write(struct.pack('H', 0x0002)) # channel fp.write(struct.pack('I', fs)) # sampling frequency(unit Hz) fp.write(struct.pack('I', fs * 2 * 2)) # data transfer speed fp.write(struct.pack('H', 2 * 2)) # data block size fp.write(struct.pack('H', depth)) # bit depth fp.write(b'data') fp.write(struct.pack('I', data_chunk_size)) def main(): # サンプリング周波数です。 44100 や 96000 などを入力してください。 fs = 192000 # ビット深度は 16 bit 固定です。 bitDepth = 16 # [StartFreq, EndFreq, StepFreq] album = [ [10, 200, 10], [200, 2000, 100], [2100, 5000, 100], [5100, 8000, 100], [8100, 10000, 100], [10100, 12000, 100], [12100, 15000, 100], [15100, 18000, 100], [18100, 20000, 100] ] for track in album: filename = '{0}-{1}Hz-{2}Hz.wav'.format(track[0], track[1], track[2]) with open(filename, 'w+b') as fp: sw = SineWaveCreator() sw.offset_header(fp) for freq in range(track[0], track[1], track[2]): sw.write(fp, 0.7, freq, fs, bitDepth, 3) sw.write_header(fp, fs, bitDepth) if __name__ == '__main__': main()
taki0112/GDWCT-Tensorflow
GDWCT.py
<gh_stars>10-100 from ops import * from utils import * from glob import glob import time from tensorflow.contrib.data import prefetch_to_device, shuffle_and_repeat, map_and_batch class GDWCT(object) : def __init__(self, sess, args): self.model_name = 'GDWCT' self.sess = sess self.checkpoint_dir = args.checkpoint_dir self.result_dir = args.result_dir self.log_dir = args.log_dir self.sample_dir = args.sample_dir self.dataset_name = args.dataset self.augment_flag = args.augment_flag self.epoch = args.epoch self.iteration = args.iteration self.gan_type = args.gan_type self.batch_size = args.batch_size self.print_freq = args.print_freq self.save_freq = args.save_freq self.num_style = args.num_style # for test self.guide_img = args.guide_img self.direction = args.direction self.img_h = args.img_h self.img_w = args.img_w self.img_ch = args.img_ch self.init_lr = args.lr self.decay_flag = args.decay_flag self.decay_start_epoch = args.decay_start_epoch self.decay_step_epoch = args.decay_step_epoch self.ch = args.ch self.phase = args.phase """ Weight """ self.gan_w = args.gan_w self.recon_x_w = args.recon_x_w self.recon_s_w = args.recon_s_w self.recon_c_w = args.recon_c_w self.recon_x_cyc_w = args.recon_x_cyc_w self.lambda_w = args.lambda_w self.lambda_c = args.lambda_c """ Generator """ self.n_res = args.n_res self.group_num = args.group_num self.style_dim = args.style_dim """ Discriminator """ self.n_dis = args.n_dis self.n_scale = args.n_scale self.sn = args.sn self.sample_dir = os.path.join(args.sample_dir, self.model_dir) check_folder(self.sample_dir) self.trainA_dataset = glob('./dataset/{}/*.*'.format(self.dataset_name + '/trainA')) self.trainB_dataset = glob('./dataset/{}/*.*'.format(self.dataset_name + '/trainB')) self.dataset_num = max(len(self.trainA_dataset), len(self.trainB_dataset)) print("##### Information #####") print("# gan type : ", self.gan_type) print("# dataset : ", self.dataset_name) print("# max dataset number : ", self.dataset_num) print("# batch_size : ", self.batch_size) print("# epoch : ", self.epoch) print("# iteration per epoch : ", self.iteration) print("# style in test phase : ", self.num_style) print() print("##### Generator #####") print("# residual blocks : ", self.n_res) print("# Style dimension : ", self.style_dim) print("# group number : ", self.group_num) print() print("##### Discriminator #####") print("# Discriminator layer : ", self.n_dis) print("# Multi-scale Dis : ", self.n_scale) print("# spectral norm : ", self.sn) print() ################################################################################## # Encoder and Decoders ################################################################################## def content_encoder(self, x, reuse=False, scope='content_encoder'): channel = self.ch with tf.variable_scope(scope, reuse=reuse) : x = conv(x, channel, kernel=7, stride=1, pad=3, pad_type='reflect', sn=self.sn, scope='conv_0') x = instance_norm(x, scope='ins_0') x = relu(x) for i in range(2) : x = conv(x, channel*2, kernel=4, stride=2, pad=1, pad_type='reflect', sn=self.sn, scope='conv_'+str(i+1)) x = instance_norm(x, scope='ins_'+str(i+1)) x = relu(x) channel = channel * 2 for i in range(self.n_res) : x = resblock(x, channel, sn=self.sn, scope='resblock_'+str(i)) return x def style_encoder(self, x, reuse=False, scope='style_encoder'): # use group_norm channel = self.ch with tf.variable_scope(scope, reuse=reuse) : x = conv(x, channel, kernel=7, stride=1, pad=3, pad_type='reflect', sn=self.sn, scope='conv_0') x = group_norm(x, groups=self.group_num, scope='group_norm_0') x = relu(x) for i in range(2) : x = conv(x, channel*2, kernel=4, stride=2, pad=1, pad_type='reflect', sn=self.sn, scope='conv_'+str(i+1)) x = group_norm(x, groups=self.group_num, scope='group_norm_' + str(i+1)) x = relu(x) channel = channel * 2 for i in range(2) : x = conv(x, channel, kernel=4, stride=2, pad=1, pad_type='reflect', sn=self.sn, scope='conv_'+str(i+1+2)) x = group_norm(x, groups=self.group_num, scope='group_norm_' + str(i+1+2)) x = relu(x) x = global_avg_pooling(x) # global average pooling return x def generator(self, content, style, reuse=False, scope="decoder"): channel = self.style_dim with tf.variable_scope(scope, reuse=reuse) : x = content U_list = [] # for regularization for i in range(self.n_res) : if i == 0: x, U = self.WCT(content, style, sn=self.sn, scope='front_WCT_' + str(i)) U_list.append(U) x = no_norm_resblock(x, channel, sn=self.sn, scope='no_norm_resblock_' + str(i)) x, U = self.WCT(x, style, sn=self.sn, scope='back_WCT_' + str(i)) U_list.append(U) for i in range(2) : x = up_sample_nearest(x, scale_factor=2) x = conv(x, channel//2, kernel=5, stride=1, pad=2, pad_type='reflect', sn=self.sn, scope='conv_'+str(i)) x = layer_norm(x, scope='layer_norm_' + str(i)) x = relu(x) channel = channel // 2 x = conv(x, channels=self.img_ch, kernel=7, stride=1, pad=3, pad_type='reflect', sn=self.sn, scope='G_logit') x = tanh(x) return x, U_list def WCT(self, content, style, sn=False, scope='wct'): with tf.variable_scope(scope) : mu = self.MLP(style, sn=sn, scope='MLP_mu') ct = self.MLP(style, sn=sn, scope='MLP_CT') alpha = tf.get_variable('alpha', shape=[1], initializer=tf.constant_initializer(0.6), constraint=lambda v: tf.clip_by_value(v, 0.0, 1.0)) x, U = GDWCT_block(content, ct, style_mu=mu, group_num=self.group_num) x = alpha * x + (1 - alpha) * content return x, U def MLP(self, x, sn=False, scope='MLP'): channel = self.style_dim with tf.variable_scope(scope) : for i in range(2) : x = fully_connected(x, channel, sn=sn, scope='linear_' + str(i)) x = lrelu(x, 0.01) x = fully_connected(x, channel, sn=sn, scope='logit') x = tf.reshape(x, shape=[-1, 1, channel]) return x ################################################################################## # Discriminator ################################################################################## def discriminator(self, x_init, reuse=False, scope="discriminator"): D_logit = [] with tf.variable_scope(scope, reuse=reuse) : for scale in range(self.n_scale) : channel = self.ch x = conv(x_init, channel, kernel=4, stride=2, pad=1, pad_type='reflect', sn=self.sn, scope='ms_' + str(scale) + '_conv_0') x = lrelu(x, 0.01) for i in range(1, self.n_dis): x = conv(x, channel * 2, kernel=4, stride=2, pad=1, pad_type='reflect', sn=self.sn, scope='ms_' + str(scale) +'_conv_' + str(i)) x = lrelu(x, 0.01) channel = channel * 2 x = conv(x, channels=1, kernel=1, stride=1, sn=self.sn, scope='ms_' + str(scale) + '_D_logit') D_logit.append(x) x_init = down_sample(x_init) return D_logit ################################################################################## # Model ################################################################################## def encoder_A(self, x_A, reuse=False): style_A = self.style_encoder(x_A, reuse=reuse, scope='style_encoder_A') content_A = self.content_encoder(x_A, reuse=reuse, scope='content_encoder_A') return content_A, style_A def encoder_B(self, x_B, reuse=False): style_B = self.style_encoder(x_B, reuse=reuse, scope='style_encoder_B') content_B = self.content_encoder(x_B, reuse=reuse, scope='content_encoder_B') return content_B, style_B def decoder_A(self, content_B, style_A, reuse=False): x_ba, U_style_A = self.generator(content=content_B, style=style_A, reuse=reuse, scope='decoder_A') return x_ba, U_style_A def decoder_B(self, content_A, style_B, reuse=False): x_ab, U_style_B = self.generator(content=content_A, style=style_B, reuse=reuse, scope='decoder_B') return x_ab, U_style_B def discriminate_real(self, x_A, x_B): real_A_logit = self.discriminator(x_A, scope="discriminator_A") real_B_logit = self.discriminator(x_B, scope="discriminator_B") return real_A_logit, real_B_logit def discriminate_fake(self, x_ba, x_ab): fake_A_logit = self.discriminator(x_ba, reuse=True, scope="discriminator_A") fake_B_logit = self.discriminator(x_ab, reuse=True, scope="discriminator_B") return fake_A_logit, fake_B_logit def build_model(self): self.lr = tf.placeholder(tf.float32, name='learning_rate') """ Input Image""" Image_Data_Class = ImageData(self.img_h, self.img_w, self.img_ch, self.augment_flag) trainA = tf.data.Dataset.from_tensor_slices(self.trainA_dataset) trainB = tf.data.Dataset.from_tensor_slices(self.trainB_dataset) gpu_device = '/gpu:0' trainA = trainA.\ apply(shuffle_and_repeat(self.dataset_num)). \ apply(map_and_batch(Image_Data_Class.image_processing, self.batch_size, num_parallel_batches=16, drop_remainder=True)). \ apply(prefetch_to_device(gpu_device, None)) trainB = trainB. \ apply(shuffle_and_repeat(self.dataset_num)). \ apply(map_and_batch(Image_Data_Class.image_processing, self.batch_size, num_parallel_batches=16, drop_remainder=True)). \ apply(prefetch_to_device(gpu_device, None)) # When using dataset.prefetch, use buffer_size=None to let it detect optimal buffer size trainA_iterator = trainA.make_one_shot_iterator() trainB_iterator = trainB.make_one_shot_iterator() self.domain_A = trainA_iterator.get_next() self.domain_B = trainB_iterator.get_next() """ Define Encoder, Generator, Discriminator """ # encode content_a, style_a = self.encoder_A(self.domain_A) content_b, style_b = self.encoder_B(self.domain_B) # decode (cross domain) x_ba, U_A = self.decoder_A(content_B=content_b, style_A=style_a) x_ab, U_B = self.decoder_B(content_A=content_a, style_B=style_b) # decode (within domain) x_aa, _ = self.decoder_A(content_B=content_a, style_A=style_a, reuse=True) x_bb, _ = self.decoder_B(content_A=content_b, style_B=style_b, reuse=True) # encode again content_ba, style_ba = self.encoder_A(x_ba, reuse=True) content_ab, style_ab = self.encoder_B(x_ab, reuse=True) # decode again (if needed) x_aba, _ = self.decoder_A(content_B=content_ab, style_A=style_ba, reuse=True) x_bab, _ = self.decoder_B(content_A=content_ba, style_B=style_ab, reuse=True) real_A_logit, real_B_logit = self.discriminate_real(self.domain_A, self.domain_B) fake_A_logit, fake_B_logit = self.discriminate_fake(x_ba, x_ab) """ Define Loss """ G_adv_A = self.gan_w * generator_loss(self.gan_type, fake_A_logit) G_adv_B = self.gan_w * generator_loss(self.gan_type, fake_B_logit) D_adv_A = self.gan_w * discriminator_loss(self.gan_type, real_A_logit, fake_A_logit) D_adv_B = self.gan_w * discriminator_loss(self.gan_type, real_B_logit, fake_B_logit) recon_style_A = self.recon_s_w * L1_loss(style_ba, style_a) recon_style_B = self.recon_s_w * L1_loss(style_ab, style_b) recon_content_A = self.recon_c_w * L1_loss(content_ab, content_a) recon_content_B = self.recon_c_w * L1_loss(content_ba, content_b) cyc_recon_A = self.recon_x_cyc_w * L1_loss(x_aba, self.domain_A) cyc_recon_B = self.recon_x_cyc_w * L1_loss(x_bab, self.domain_B) recon_A = self.recon_x_w * L1_loss(x_aa, self.domain_A) # reconstruction recon_B = self.recon_x_w * L1_loss(x_bb, self.domain_B) # reconstruction whitening_A, coloring_A = group_wise_regularization(deep_whitening_transform(content_a), U_A, self.group_num) whitening_B, coloring_B = group_wise_regularization(deep_whitening_transform(content_b), U_B, self.group_num) whitening_A = self.lambda_w * whitening_A whitening_B = self.lambda_w * whitening_B coloring_A = self.lambda_c * coloring_A coloring_B = self.lambda_c * coloring_B G_reg_A = regularization_loss('decoder_A') + regularization_loss('encoder_A') G_reg_B = regularization_loss('decoder_B') + regularization_loss('encoder_B') D_reg_A = regularization_loss('discriminator_A') D_reg_B = regularization_loss('discriminator_B') Generator_A_loss = G_adv_A + \ recon_A + \ recon_style_A + \ recon_content_A + \ cyc_recon_B + \ whitening_A + \ coloring_A + \ G_reg_A Generator_B_loss = G_adv_B + \ recon_B + \ recon_style_B + \ recon_content_B + \ cyc_recon_A + \ whitening_B + \ coloring_B + \ G_reg_B Discriminator_A_loss = D_adv_A + D_reg_A Discriminator_B_loss = D_adv_B + D_reg_B self.Generator_loss = Generator_A_loss + Generator_B_loss self.Discriminator_loss = Discriminator_A_loss + Discriminator_B_loss """ Training """ t_vars = tf.trainable_variables() G_vars = [var for var in t_vars if 'decoder' in var.name or 'encoder' in var.name] D_vars = [var for var in t_vars if 'discriminator' in var.name] self.G_optim = tf.train.AdamOptimizer(self.lr, beta1=0.5, beta2=0.999).minimize(self.Generator_loss, var_list=G_vars) self.D_optim = tf.train.AdamOptimizer(self.lr, beta1=0.5, beta2=0.999).minimize(self.Discriminator_loss, var_list=D_vars) """" Summary """ self.all_G_loss = tf.summary.scalar("Generator_loss", self.Generator_loss) self.all_D_loss = tf.summary.scalar("Discriminator_loss", self.Discriminator_loss) self.G_A_loss = tf.summary.scalar("G_A_loss", Generator_A_loss) self.G_B_loss = tf.summary.scalar("G_B_loss", Generator_B_loss) self.D_A_loss = tf.summary.scalar("D_A_loss", Discriminator_A_loss) self.D_B_loss = tf.summary.scalar("D_B_loss", Discriminator_B_loss) self.G_A_adv_loss = tf.summary.scalar("G_A_adv_loss", G_adv_A) self.G_A_style_loss = tf.summary.scalar("G_A_style_loss", recon_style_A) self.G_A_content_loss = tf.summary.scalar("G_A_content_loss", recon_content_A) self.G_A_cyc_loss = tf.summary.scalar("G_A_cyc_loss", cyc_recon_A) self.G_A_identity_loss = tf.summary.scalar("G_A_identity_loss", recon_A) self.G_A_whitening_loss = tf.summary.scalar("G_A_whitening_loss", whitening_A) self.G_A_coloring_loss = tf.summary.scalar("G_A_coloring_loss", coloring_A) self.G_B_adv_loss = tf.summary.scalar("G_B_adv_loss", G_adv_B) self.G_B_style_loss = tf.summary.scalar("G_B_style_loss", recon_style_B) self.G_B_content_loss = tf.summary.scalar("G_B_content_loss", recon_content_B) self.G_B_cyc_loss = tf.summary.scalar("G_B_cyc_loss", cyc_recon_B) self.G_B_identity_loss = tf.summary.scalar("G_B_identity_loss", recon_B) self.G_B_whitening_loss = tf.summary.scalar("G_B_whitening_loss", whitening_B) self.G_B_coloring_loss = tf.summary.scalar("G_B_coloring_loss", coloring_B) self.alpha_var = [] for var in tf.trainable_variables(): if 'alpha' in var.name: self.alpha_var.append(tf.summary.histogram(var.name, var)) self.alpha_var.append(tf.summary.scalar(var.name, tf.reduce_max(var))) G_summary_list = [self.G_A_adv_loss, self.G_A_style_loss, self.G_A_content_loss, self.G_A_cyc_loss, self.G_A_identity_loss, self.G_A_whitening_loss, self.G_A_coloring_loss, self.G_A_loss, self.G_B_adv_loss, self.G_B_style_loss, self.G_B_content_loss, self.G_B_cyc_loss, self.G_B_identity_loss, self.G_B_whitening_loss, self.G_B_coloring_loss, self.G_B_loss, self.all_G_loss] G_summary_list.extend(self.alpha_var) self.G_loss = tf.summary.merge(G_summary_list) self.D_loss = tf.summary.merge([self.D_A_loss, self.D_B_loss, self.all_D_loss]) """ Image """ self.fake_A = x_ba self.fake_B = x_ab self.real_A = self.domain_A self.real_B = self.domain_B """ Test """ """ Guided Image Translation """ self.content_image = tf.placeholder(tf.float32, [1, self.img_h, self.img_w, self.img_ch], name='content_image') self.style_image = tf.placeholder(tf.float32, [1, self.img_h, self.img_w, self.img_ch], name='guide_style_image') if self.direction == 'a2b' : guide_content_A, _ = self.encoder_A(self.content_image, reuse=True) _, guide_style_B = self.encoder_B(self.style_image, reuse=True) self.guide_fake_B, _ = self.decoder_B(content_A=guide_content_A, style_B=guide_style_B, reuse=True) else : guide_content_B, _ = self.encoder_B(self.content_image, reuse=True) _, guide_style_A = self.encoder_A(self.style_image, reuse=True) self.guide_fake_A, _ = self.decoder_A(content_B=guide_content_B, style_A=guide_style_A, reuse=True) def train(self): # initialize all variables tf.global_variables_initializer().run() # saver to save model self.saver = tf.train.Saver(max_to_keep=20) # summary writer self.writer = tf.summary.FileWriter(self.log_dir + '/' + self.model_dir, self.sess.graph) # restore check-point if it exits could_load, checkpoint_counter = self.load(self.checkpoint_dir) if could_load: start_epoch = (int)(checkpoint_counter / self.iteration) start_batch_id = checkpoint_counter - start_epoch * self.iteration counter = checkpoint_counter print(" [*] Load SUCCESS") else: start_epoch = 0 start_batch_id = 0 counter = 1 print(" [!] Load failed...") # loop for epoch start_time = time.time() lr = self.init_lr for epoch in range(start_epoch, self.epoch): if self.decay_flag and self.decay_start_epoch > epoch : lr = self.init_lr * pow(0.5, (epoch - self.decay_start_epoch) // self.decay_step_epoch) for idx in range(start_batch_id, self.iteration): train_feed_dict = { self.lr : lr } # Update D _, d_loss, summary_str = self.sess.run([self.D_optim, self.Discriminator_loss, self.D_loss], feed_dict = train_feed_dict) self.writer.add_summary(summary_str, counter) # Update G batch_A_images, batch_B_images, fake_A, fake_B, _, g_loss, summary_str = self.sess.run([self.real_A, self.real_B, self.fake_A, self.fake_B, self.G_optim, self.Generator_loss, self.G_loss], feed_dict = train_feed_dict) self.writer.add_summary(summary_str, counter) # display training status counter += 1 print("Epoch: [%2d] [%6d/%6d] time: %4.4f d_loss: %.8f, g_loss: %.8f" \ % (epoch, idx, self.iteration, time.time() - start_time, d_loss, g_loss)) if np.mod(idx+1, self.print_freq) == 0 : save_images(batch_A_images, [self.batch_size, 1], './{}/real_A_{:03d}_{:05d}.jpg'.format(self.sample_dir, epoch, idx+1)) # save_images(batch_B_images, [self.batch_size, 1], # './{}/real_B_{:03d}_{:05d}.jpg'.format(self.sample_dir, epoch, idx+1)) # save_images(fake_A, [self.batch_size, 1], # './{}/fake_A_{:03d}_{:05d}.jpg'.format(self.sample_dir, epoch, idx+1)) save_images(fake_B, [self.batch_size, 1], './{}/fake_B_{:03d}_{:05d}.jpg'.format(self.sample_dir, epoch, idx+1)) if np.mod(counter - 1, self.save_freq) == 0 : self.save(self.checkpoint_dir, counter) # After an epoch, start_batch_id is set to zero # non-zero value is only for the first epoch after loading pre-trained model start_batch_id = 0 # save model for final step self.save(self.checkpoint_dir, counter) @property def model_dir(self): n_dis = str(self.n_scale) + 'multi_' + str(self.n_dis) + 'dis' sn = '' if self.sn : sn = '_sn' return "{}_{}_{}_{}_{}adv_{}style_{}content_{}identity_{}cyc_{}color_{}white{}".format(self.model_name, self.dataset_name, self.gan_type, n_dis, self.gan_w, self.recon_s_w, self.recon_c_w, self.recon_x_w, self.recon_x_cyc_w, self.lambda_c, self.lambda_w, sn) def save(self, checkpoint_dir, step): checkpoint_dir = os.path.join(checkpoint_dir, self.model_dir) if not os.path.exists(checkpoint_dir): os.makedirs(checkpoint_dir) self.saver.save(self.sess, os.path.join(checkpoint_dir, self.model_name + '.model'), global_step=step) def load(self, checkpoint_dir): print(" [*] Reading checkpoints...") checkpoint_dir = os.path.join(checkpoint_dir, self.model_dir) ckpt = tf.train.get_checkpoint_state(checkpoint_dir) if ckpt and ckpt.\ model_checkpoint_path: ckpt_name = os.path.basename(ckpt.model_checkpoint_path) self.saver.restore(self.sess, os.path.join(checkpoint_dir, ckpt_name)) counter = int(ckpt_name.split('-')[-1]) print(" [*] Success to read {}".format(ckpt_name)) return True, counter else: print(" [*] Failed to find a checkpoint") return False, 0 def style_guide_test(self): tf.global_variables_initializer().run() test_A_files = glob('./dataset/{}/*.*'.format(self.dataset_name + '/testA')) test_B_files = glob('./dataset/{}/*.*'.format(self.dataset_name + '/testB')) style_file = load_test_data(self.guide_img, size_h=self.img_h, size_w=self.img_w) self.saver = tf.train.Saver() could_load, checkpoint_counter = self.load(self.checkpoint_dir) self.result_dir = os.path.join(self.result_dir, self.model_dir, 'guide') check_folder(self.result_dir) if could_load: print(" [*] Load SUCCESS") else: print(" [!] Load failed...") # write html for visual comparison index_path = os.path.join(self.result_dir, 'index.html') index = open(index_path, 'w') index.write("<html><body><table><tr>") index.write("<th>name</th><th>input</th><th>output</th></tr>") if self.direction == 'a2b' : for sample_file in test_A_files: # A -> B print('Processing A image: ' + sample_file) sample_image = load_test_data(sample_file, size_h=self.img_h, size_w=self.img_w) image_path = os.path.join(self.result_dir, '{}'.format(os.path.basename(sample_file))) fake_img = self.sess.run(self.guide_fake_B, feed_dict={self.content_image: sample_image, self.style_image : style_file}) save_images(fake_img, [1, 1], image_path) index.write("<td>%s</td>" % os.path.basename(image_path)) index.write("<td><img src='%s' width='%d' height='%d'></td>" % (sample_file if os.path.isabs(sample_file) else ( '../../..' + os.path.sep + sample_file), self.img_w, self.img_h)) index.write("<td><img src='%s' width='%d' height='%d'></td>" % (image_path if os.path.isabs(image_path) else ( '../../..' + os.path.sep + image_path), self.img_w, self.img_h)) index.write("</tr>") else : for sample_file in test_B_files: # B -> A print('Processing B image: ' + sample_file) sample_image = np.asarray(load_test_data(sample_file, size_h=self.img_h, size_w=self.img_w)) image_path = os.path.join(self.result_dir, '{}'.format(os.path.basename(sample_file))) fake_img = self.sess.run(self.guide_fake_A, feed_dict={self.content_image: sample_image, self.style_image : style_file}) save_images(fake_img, [1, 1], image_path) index.write("<td>%s</td>" % os.path.basename(image_path)) index.write("<td><img src='%s' width='%d' height='%d'></td>" % (sample_file if os.path.isabs(sample_file) else ( '../../..' + os.path.sep + sample_file), self.img_w, self.img_h)) index.write("<td><img src='%s' width='%d' height='%d'></td>" % (image_path if os.path.isabs(image_path) else ( '../../..' + os.path.sep + image_path), self.img_w, self.img_h)) index.write("</tr>") index.close()
taki0112/GDWCT-Tensorflow
ops.py
<gh_stars>10-100 import tensorflow as tf import tensorflow.contrib as tf_contrib import numpy as np from utils import pytorch_xavier_weight_factor, pytorch_kaiming_weight_factor factor, mode, uniform = pytorch_kaiming_weight_factor(a=0.0, uniform=False) weight_init = tf_contrib.layers.variance_scaling_initializer(factor=factor, mode=mode, uniform=uniform) # weight_init = tf.random_normal_initializer(mean=0.0, stddev=0.02) weight_regularizer = tf_contrib.layers.l2_regularizer(scale=0.0001) weight_regularizer_fully = tf_contrib.layers.l2_regularizer(scale=0.0001) ################################################################################## # Layer ################################################################################## def conv(x, channels, kernel=4, stride=2, pad=0, pad_type='zero', use_bias=True, sn=False, scope='conv_0'): with tf.variable_scope(scope): if scope.__contains__("discriminator"): weight_init = tf.random_normal_initializer(mean=0.0, stddev=0.02) else: weight_init = tf_contrib.layers.variance_scaling_initializer(factor=factor, mode=mode, uniform=uniform) if pad > 0: h = x.get_shape().as_list()[1] if h % stride == 0: pad = pad * 2 else: pad = max(kernel - (h % stride), 0) pad_top = pad // 2 pad_bottom = pad - pad_top pad_left = pad // 2 pad_right = pad - pad_left if pad_type == 'zero': x = tf.pad(x, [[0, 0], [pad_top, pad_bottom], [pad_left, pad_right], [0, 0]]) if pad_type == 'reflect': x = tf.pad(x, [[0, 0], [pad_top, pad_bottom], [pad_left, pad_right], [0, 0]], mode='REFLECT') if sn: w = tf.get_variable("kernel", shape=[kernel, kernel, x.get_shape()[-1], channels], initializer=weight_init, regularizer=weight_regularizer) x = tf.nn.conv2d(input=x, filter=spectral_norm(w), strides=[1, stride, stride, 1], padding='VALID') if use_bias: bias = tf.get_variable("bias", [channels], initializer=tf.constant_initializer(0.0)) x = tf.nn.bias_add(x, bias) else: x = tf.layers.conv2d(inputs=x, filters=channels, kernel_size=kernel, kernel_initializer=weight_init, kernel_regularizer=weight_regularizer, strides=stride, use_bias=use_bias) return x def fully_connected(x, units, use_bias=True, sn=False, scope='linear'): with tf.variable_scope(scope): x = flatten(x) shape = x.get_shape().as_list() channels = shape[-1] if sn: w = tf.get_variable("kernel", [channels, units], tf.float32, initializer=weight_init, regularizer=weight_regularizer_fully) if use_bias: bias = tf.get_variable("bias", [units], initializer=tf.constant_initializer(0.0)) x = tf.matmul(x, spectral_norm(w)) + bias else: x = tf.matmul(x, spectral_norm(w)) else: x = tf.layers.dense(x, units=units, kernel_initializer=weight_init, kernel_regularizer=weight_regularizer_fully, use_bias=use_bias) return x def flatten(x) : return tf.layers.flatten(x) def GDWCT_block(content, style, style_mu, group_num) : content = deep_whitening_transform(content) # [bs, h, w, ch] U, style = deep_coloring_transform(style, group_num) # [bs, 1, ch], [bs, ch, ch] bs, h, w, ch = content.get_shape().as_list() content = tf.reshape(content, shape=[bs, h*w, ch]) x = tf.matmul(content, style) + style_mu x = tf.reshape(x, shape=[bs, h, w, ch]) return x, U def deep_whitening_transform(c) : mu, _ = tf.nn.moments(c, axes=[1, 2], keep_dims=True) x = c - mu return x def deep_coloring_transform(s, group_num) : # [batch_size, 1, channel] : S : MLP^CT(s) bs, _, ch = s.get_shape().as_list() # make U l2_norm = tf.norm(s, axis=-1, keepdims=True) U = s / l2_norm # make D eye = tf.eye(num_rows=ch, num_columns=ch, batch_shape=[bs]) # [batch_size, channel, channel] D = l2_norm * eye U_block_list = [] split_num = ch // group_num for i in range(group_num) : U_ = U[:, :, i * split_num: (i + 1) * split_num] D_ = D[:, i * split_num: (i + 1) * split_num, i * split_num: (i + 1) * split_num] block_matrix = U_ * D_ * tf.transpose(U_, perm=[0, 2, 1]) operator_matrix = tf.linalg.LinearOperatorFullMatrix(block_matrix) U_block_list.append(operator_matrix) U_block_diag_matrix = tf.linalg.LinearOperatorBlockDiag(U_block_list).to_dense() return U, U_block_diag_matrix def group_wise_regularization(c_whitening, U_list, group_num) : """ Regularization """ """ whitening regularization """ bs, h, w, ch = c_whitening.get_shape().as_list() c_whitening = tf.reshape(c_whitening, shape=[bs, h * w, ch]) c_whitening = tf.matmul(tf.transpose(c_whitening, perm=[0, 2, 1]), c_whitening) # covariance of x [bs, ch, ch] bs, ch, _ = c_whitening.get_shape().as_list() # ch1 = ch2 index_matrix = make_index_matrix(bs, ch, ch, group_num) group_convariance_x = tf.where(tf.equal(index_matrix, 1.0), c_whitening, tf.zeros_like(c_whitening)) group_convariance_x = tf.linalg.set_diag(group_convariance_x, tf.ones([bs, ch])) whitening_regularization_loss = L1_loss(c_whitening, group_convariance_x) """ coloring regularization """ split_num = ch // group_num coloring_regularization_list = [] coloring_regularization_loss_list = [] for U in U_list : # [bs, 1, ch] for i in range(group_num): U_ = U[:, :, i * split_num: (i + 1) * split_num] U_TU = tf.matmul(tf.transpose(U_, perm=[0, 2, 1]), U_) # [bs, ch // group_num, ch // group_num] coloring_regularization_list.append(L1_loss(U_TU, tf.eye(num_rows=ch // group_num, num_columns=ch // group_num, batch_shape=[bs]))) coloring_regularization_loss_list.append(tf.reduce_mean(coloring_regularization_list)) coloring_regularization_loss = tf.reduce_mean(coloring_regularization_loss_list) return whitening_regularization_loss, coloring_regularization_loss def make_index_matrix(bs, ch1, ch2, group_num) : index_matrix = np.abs(np.kron(np.eye(ch1 // group_num, ch2 // group_num), np.eye(group_num, group_num) - 1)) index_matrix[index_matrix == 0] = -1 index_matrix[index_matrix == 1] = 0 index_matrix[index_matrix == -1] = 1 index_matrix = np.tile(index_matrix, [bs, 1, 1]) return index_matrix ################################################################################## # Residual-block ################################################################################## def resblock(x_init, channels, use_bias=True, sn=False, scope='resblock'): with tf.variable_scope(scope): with tf.variable_scope('res1'): x = conv(x_init, channels, kernel=3, stride=1, pad=1, pad_type='reflect', use_bias=use_bias, sn=sn) x = instance_norm(x) x = relu(x) with tf.variable_scope('res2'): x = conv(x, channels, kernel=3, stride=1, pad=1, pad_type='reflect', use_bias=use_bias, sn=sn) x = instance_norm(x) return x + x_init def no_norm_resblock(x_init, channels, use_bias=True, sn=False, scope='resblock'): with tf.variable_scope(scope): with tf.variable_scope('res1'): x = conv(x_init, channels, kernel=3, stride=1, pad=1, pad_type='reflect', use_bias=use_bias, sn=sn) x = relu(x) with tf.variable_scope('res2'): x = conv(x, channels, kernel=3, stride=1, pad=1, pad_type='reflect', use_bias=use_bias, sn=sn) return x + x_init ################################################################################## # Sampling ################################################################################## def down_sample(x) : return tf.layers.average_pooling2d(x, pool_size=3, strides=2, padding='SAME') def up_sample(x, scale_factor=2): _, h, w, _ = x.get_shape().as_list() new_size = [h * scale_factor, w * scale_factor] return tf.image.resize_nearest_neighbor(x, size=new_size) def up_sample_nearest(x, scale_factor=2): _, h, w, _ = x.get_shape().as_list() new_size = [h * scale_factor, w * scale_factor] return tf.image.resize_nearest_neighbor(x, size=new_size) def global_avg_pooling(x): gap = tf.reduce_mean(x, axis=[1, 2], keepdims=True) return gap ################################################################################## # Activation function ################################################################################## def lrelu(x, alpha=0.01): # pytorch alpha is 0.01 return tf.nn.leaky_relu(x, alpha) def relu(x): return tf.nn.relu(x) def tanh(x): return tf.tanh(x) ################################################################################## # Normalization function ################################################################################## def instance_norm(x, scope='instance_norm'): return tf_contrib.layers.instance_norm(x, epsilon=1e-05, center=True, scale=True, scope=scope) def layer_norm(x, scope='layer_norm') : return tf_contrib.layers.layer_norm(x, center=True, scale=True, scope=scope) def group_norm(x, groups=8, scope='group_norm'): return tf.contrib.layers.group_norm(x, groups=groups, epsilon=1e-05, center=True, scale=True, scope=scope) def spectral_norm(w, iteration=1): w_shape = w.shape.as_list() w = tf.reshape(w, [-1, w_shape[-1]]) u = tf.get_variable("u", [1, w_shape[-1]], initializer=tf.random_normal_initializer(), trainable=False) u_hat = u v_hat = None for i in range(iteration): """ power iteration Usually iteration = 1 will be enough """ v_ = tf.matmul(u_hat, tf.transpose(w)) v_hat = tf.nn.l2_normalize(v_) u_ = tf.matmul(v_hat, w) u_hat = tf.nn.l2_normalize(u_) u_hat = tf.stop_gradient(u_hat) v_hat = tf.stop_gradient(v_hat) sigma = tf.matmul(tf.matmul(v_hat, w), tf.transpose(u_hat)) with tf.control_dependencies([u.assign(u_hat)]): w_norm = w / sigma w_norm = tf.reshape(w_norm, w_shape) return w_norm ################################################################################## # Loss function ################################################################################## def discriminator_loss(gan_type, real, fake): n_scale = len(real) loss = [] real_loss = 0 fake_loss = 0 for i in range(n_scale) : if gan_type == 'lsgan' : real_loss = tf.reduce_mean(tf.squared_difference(real[i], 1.0)) fake_loss = tf.reduce_mean(tf.square(fake[i])) if gan_type == 'gan' : real_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.ones_like(real[i]), logits=real[i])) fake_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.zeros_like(fake[i]), logits=fake[i])) if gan_type == 'hinge': real_loss = -tf.reduce_mean(tf.minimum(real[i][-1] - 1, 0.0)) fake_loss = -tf.reduce_mean(tf.minimum(-fake[i][-1] - 1, 0.0)) loss.append(real_loss + fake_loss) return tf.reduce_sum(loss) def generator_loss(gan_type, fake): n_scale = len(fake) loss = [] fake_loss = 0 for i in range(n_scale) : if gan_type == 'lsgan' : fake_loss = tf.reduce_mean(tf.squared_difference(fake[i], 1.0)) if gan_type == 'gan' : fake_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.ones_like(fake[i]), logits=fake[i])) if gan_type == 'hinge': # fake_loss = -tf.reduce_mean(relu(fake[i][-1])) fake_loss = -tf.reduce_mean(fake[i][-1]) loss.append(fake_loss) return tf.reduce_sum(loss) def L1_loss(x, y): loss = tf.reduce_mean(tf.abs(x - y)) return loss def regularization_loss(scope_name) : """ If you want to use "Regularization" g_loss += regularization_loss('generator') d_loss += regularization_loss('discriminator') """ collection_regularization = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES) loss = [] for item in collection_regularization : if scope_name in item.name : loss.append(item) return tf.reduce_sum(loss)
altuntasmuhammet/eksisozluk-scraper
altuntas/scraper/models/__init__.py
<gh_stars>1-10 from .eksisozlukbot import Entry
altuntasmuhammet/eksisozluk-scraper
altuntas/altuntas/celery.py
<gh_stars>1-10 from __future__ import absolute_import from celery import Celery from django.conf import settings import os # set the default Django settings module for the 'celery' program. os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'altuntas.settings') app = Celery('altuntas') app.config_from_object('django.conf:settings', namespace='CELERY') app.autodiscover_tasks(lambda: settings.INSTALLED_APPS)
altuntasmuhammet/eksisozluk-scraper
altuntas/scraper/migrations/0001_initial.py
<filename>altuntas/scraper/migrations/0001_initial.py # Generated by Django 3.2.5 on 2021-07-10 17:15 from django.db import migrations, models class Migration(migrations.Migration): initial = True dependencies = [ ] operations = [ migrations.CreateModel( name='Entry', fields=[ ('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')), ('title', models.TextField()), ('content', models.TextField()), ('favourite_count', models.IntegerField()), ('author', models.TextField()), ('created_date', models.DateTimeField()), ('edited_date', models.DateTimeField(blank=True)), ('eksisozluk_entry_id', models.IntegerField(unique=True)), ('eksisozluk_author_id', models.IntegerField()), ], options={ 'verbose_name': 'Eksisozluk Entry', 'verbose_name_plural': 'Eksisozluk Entries', 'db_table': 'web_eksisozluk_entries', }, ), ]
altuntasmuhammet/eksisozluk-scraper
altuntas/scraper/migrations/0002_alter_entry_table.py
<filename>altuntas/scraper/migrations/0002_alter_entry_table.py # Generated by Django 3.2.5 on 2021-07-10 17:16 from django.db import migrations class Migration(migrations.Migration): dependencies = [ ('scraper', '0001_initial'), ] operations = [ migrations.AlterModelTable( name='entry', table='scraper_eksisozluk_entries', ), ]
altuntasmuhammet/eksisozluk-scraper
altuntas/scraper/tasks/eksisozlukbot.py
<filename>altuntas/scraper/tasks/eksisozlukbot.py from celery.decorators import task from eksisozlukbot.eksisozlukbot.spiders.eksisozluk import EksisozlukSpider from eksisozlukbot.eksisozlukbot import settings as eksisozlukbot_settings from scrapy import signals from scrapy.crawler import Crawler from scrapy.settings import Settings from twisted.internet import reactor from scrapy.crawler import CrawlerProcess from scrapy.utils.project import get_project_settings import os @task(name="Scrape EksiSozluk", track_started=True) def scrape_eksisozluk(**kwargs): # process = CrawlerProcess(settings=get_project_settings()) # process.crawl(EksisozlukSpider) # process.start() crawler_settings = Settings() crawler_settings.setmodule(eksisozlukbot_settings) #Create a crawler crawler = Crawler(EksisozlukSpider, settings=crawler_settings) # This ensures Twisted Reactor is properly shutdown crawler.signals.connect(reactor.stop, signal=signals.spider_closed) # Start crawling crawler.crawl(**kwargs) # add blocking process reactor.run()
altuntasmuhammet/eksisozluk-scraper
altuntas/scraper/models/eksisozlukbot.py
from django.conf import settings from django.db import models class Entry(models.Model): title = models.TextField() content = models.TextField() favourite_count = models.IntegerField() author = models.TextField() created_date = models.DateTimeField() edited_date = models.DateTimeField(blank=True) eksisozluk_entry_id = models.IntegerField(unique=True) eksisozluk_author_id = models.IntegerField() class Meta: app_label = "scraper" db_table = "scraper_eksisozluk_entries" verbose_name_plural = "Eksisozluk Entries" verbose_name = "Eksisozluk Entry" # abstract=True
altuntasmuhammet/eksisozluk-scraper
altuntas/altuntas/settings.py
<reponame>altuntasmuhammet/eksisozluk-scraper<filename>altuntas/altuntas/settings.py<gh_stars>1-10 """ Django settings for altuntas project. Generated by 'django-admin startproject' using Django 3.2.3. For more information on this file, see https://docs.djangoproject.com/en/3.2/topics/settings/ For the full list of settings and their values, see https://docs.djangoproject.com/en/3.2/ref/settings/ """ from pathlib import Path import os from kombu import Exchange, Queue # Build paths inside the project like this: BASE_DIR / 'subdir'. BASE_DIR = Path(__file__).resolve().parent.parent # Quick-start development settings - unsuitable for production # See https://docs.djangoproject.com/en/3.2/howto/deployment/checklist/ # SECURITY WARNING: keep the secret key used in production secret! SECRET_KEY = '<KEY>' # SECURITY WARNING: don't run with debug turned on in production! DEBUG = True ALLOWED_HOSTS = ['localhost'] # Database configuration POSTGRES_DB=os.environ['POSTGRES_DB'] POSTGRES_USER=os.environ['POSTGRES_USER'] POSTGRES_PASSWORD=os.environ['POSTGRES_PASSWORD'] POSTGRES_HOST=os.environ['POSTGRES_HOST'] POSTGRES_PORT=os.environ['POSTGRES_PORT'] # Rabbitmq configuration RABBITMQ_DEFAULT_USER = os.environ.get('RABBITMQ_DEFAULT_USER', 'guest') RABBITMQ_DEFAULT_PASS = os.environ.get('RABBITMQ_DEFAULT_PASS', 'guest') RABBITMQ_DEFAULT_HOST = os.environ.get('RABBITMQ_DEFAULT_HOST', 'rabbitmq') RABBITMQ_DEFAULT_PORT = os.environ.get('RABBITMQ_DEFAULT_PORT', 5672) RABBITMQ_DEFAULT_MGMT_PORT = os.environ.get('RABBITMQ_DEFAULT_MGMT_PORT', 15672) RABBITMQ_DEFAULT_VHOST = os.environ.get('RABBITMQ_DEFAULT_VHOST', 'vhost') # Celery configuration CELERY_BROKER_URL = f"amqp://{RABBITMQ_DEFAULT_USER}:{RABBITMQ_DEFAULT_PASS}@{RABBITMQ_DEFAULT_HOST}:{RABBITMQ_DEFAULT_PORT}/{RABBITMQ_DEFAULT_VHOST}" print("CELERY_BROKER_URL", CELERY_BROKER_URL) CELERY_RESULT_BACKEND = 'django-db' CELERY_CACHE_BACKEND = 'django-cache' CELERY_TASK_SERIALIZER = "json" CELERY_ACCEPT_CONTENT = ['application/json'] CELERY_ALWAYS_EAGER = False CELERY_ACKS_LATE = True CELERY_TASK_PUBLISH_RETRY = True CELERY_DISABLE_RATE_LIMITS = False CELERY_IMPORTS = ( 'scraper.tasks.eksisozlukbot', ) CELERY_DEFAULT_QUEUE = 'default' CELERY_QUEUES = ( Queue('default', Exchange('default'), routing_key='default'), ) # Application definition INSTALLED_APPS = [ 'bootstrap_admin', 'django.contrib.admin', 'django.contrib.auth', 'django.contrib.contenttypes', 'django.contrib.sessions', 'django.contrib.messages', 'django.contrib.staticfiles', 'django_celery_beat', 'django_celery_results', 'scraper', ] MIDDLEWARE = [ 'django.middleware.security.SecurityMiddleware', 'django.contrib.sessions.middleware.SessionMiddleware', 'django.middleware.common.CommonMiddleware', 'django.middleware.csrf.CsrfViewMiddleware', 'django.contrib.auth.middleware.AuthenticationMiddleware', 'django.contrib.messages.middleware.MessageMiddleware', 'django.middleware.clickjacking.XFrameOptionsMiddleware', ] ROOT_URLCONF = 'altuntas.urls' TEMPLATES = [ { 'BACKEND': 'django.template.backends.django.DjangoTemplates', 'DIRS': [], 'APP_DIRS': True, 'OPTIONS': { 'context_processors': [ 'django.template.context_processors.debug', 'django.template.context_processors.request', 'django.contrib.auth.context_processors.auth', 'django.contrib.messages.context_processors.messages', ], }, }, ] WSGI_APPLICATION = 'altuntas.wsgi.application' # Database # https://docs.djangoproject.com/en/3.2/ref/settings/#databases DATABASES = { 'default': { 'ENGINE': 'django.db.backends.postgresql_psycopg2', 'NAME': POSTGRES_DB, 'USER': POSTGRES_USER, 'PASSWORD':<PASSWORD>, 'HOST': POSTGRES_HOST, 'PORT': POSTGRES_PORT, } } # Password validation # https://docs.djangoproject.com/en/3.2/ref/settings/#auth-password-validators AUTH_PASSWORD_VALIDATORS = [ { 'NAME': 'django.contrib.auth.password_validation.UserAttributeSimilarityValidator', }, { 'NAME': 'django.contrib.auth.password_validation.MinimumLengthValidator', }, { 'NAME': 'django.contrib.auth.password_validation.CommonPasswordValidator', }, { 'NAME': 'django.contrib.auth.password_validation.NumericPasswordValidator', }, ] # Internationalization # https://docs.djangoproject.com/en/3.2/topics/i18n/ LANGUAGE_CODE = 'en-us' TIME_ZONE = 'UTC' USE_I18N = True USE_L10N = True USE_TZ = True # Static files (CSS, JavaScript, Images) # https://docs.djangoproject.com/en/3.2/howto/static-files/ STATIC_URL = '/static/' STATIC_ROOT = os.path.join(BASE_DIR, 'static/') # Default primary key field type # https://docs.djangoproject.com/en/3.2/ref/settings/#default-auto-field DEFAULT_AUTO_FIELD = 'django.db.models.BigAutoField'
altuntasmuhammet/eksisozluk-scraper
altuntas/eksisozlukbot/eksisozlukbot/spiders/eksisozluk.py
<gh_stars>1-10 import scrapy import time from datetime import datetime import urllib.parse EKSISOZLUK_BASEURL = "https://eksisozluk.com" DATETIME_FORMAT = "%d.%m.%Y %H:%M" START_DATE = "" END_DATE = "" # keywords = ["apache kafka"] class EksisozlukSpider(scrapy.Spider): name = 'eksisozluk' allowed_domains = ['eksisozluk.com'] # start_urls = ['http://eksisozluk.com/'] def __init__(self, keywords="", *args, **kwargs): super(EksisozlukSpider, self).__init__(*args, **kwargs) self.keywords = [k.strip() for k in keywords.split(",")] def start_requests(self): title_query_url_templated = "https://eksisozluk.com/basliklar/ara?SearchForm.Keywords={keyword}&SearchForm.Author=&SearchForm.When.From={start_date}&SearchForm.When.To={end_date}&SearchForm.NiceOnly=false&SearchForm.FavoritedOnly=false&SearchForm.SortOrder=Date&_={timestamp}" for keyword in self.keywords: encoded_keyword = urllib.parse.quote(keyword) timestamp = int(time.time() * 1000) url = title_query_url_templated.format(keyword=encoded_keyword, start_date=START_DATE, end_date=END_DATE, timestamp=timestamp) print("******URL****** - {}".format(url)) yield scrapy.Request(url=url, callback=self.parse_title_links) def parse_title_links(self, response): print() titles = response.css('div.instapaper_body ul.topic-list li') print("********TITLE********:", titles) for title in titles: title_url = EKSISOZLUK_BASEURL + title.css('a::attr(href)').get() title_name = title.css("a::text").get().strip() yield scrapy.Request(url=title_url, callback=self.parse_pages, meta={'title_url': title_url, 'title_name': title_name}) def parse_pages(self, response): page_count_str = response.css('div.pager::attr(data-pagecount)').get() page_count = int(page_count_str) if page_count_str else 1 title_url = response.meta['title_url'] title_name = response.meta['title_name'] for page_num in range(1, page_count+1): url = "{title_url}?p={page_num}".format(title_url=title_url, page_num=page_num) yield scrapy.Request(url=url, callback=self.parse_entries_per_page, meta={'title_name':title_name}) def parse_entries_per_page(self, response): entries = response.css('ul[id=entry-item-list] li') title_name = response.meta['title_name'] for entry in entries: # Content content = " ".join(entry.css("div.content *::text").getall()).strip() print("******LOG****** - Content:", content) # Title title = title_name # Favourite Count favourite_count = int(entry.attrib['data-favorite-count']) # Author Name author = entry.css("footer div.info a.entry-author::text").get().strip() # Created Date and Edited Date date = entry.css("footer div.info a.entry-date::text").get().strip() if '~' in date: created_date_str = date.split('~')[0].strip() edited_date_str = date.split('~')[1].strip() print("LOG 1", created_date_str, edited_date_str) edited_date_str = edited_date_str if len(edited_date_str)==len(created_date_str) else " ".join([created_date_str.split(' ')[0], edited_date_str]) print("LOG 2", created_date_str, edited_date_str) created_date = datetime.strptime(created_date_str, DATETIME_FORMAT) edited_date = datetime.strptime(edited_date_str, DATETIME_FORMAT) else: created_date_str = date.strip() created_date = datetime.strptime(created_date_str, DATETIME_FORMAT) edited_date = None # eksisozluk_entry_id eksisozluk_entry_id = int(entry.attrib['data-id']) # eksisozluk_author_id eksisozluk_author_id = int(entry.attrib['data-author-id']) yield { "content": content, "title": title, "favourite_count": favourite_count, "author": author, "created_date": created_date, "edited_date": edited_date, "eksisozluk_entry_id": eksisozluk_entry_id, "eksisozluk_author_id": eksisozluk_author_id }
altuntasmuhammet/eksisozluk-scraper
altuntas/scraper/admin.py
from django.contrib import admin from .models import Entry # Register your models here. class EntryAdmin(admin.ModelAdmin): list_display = ("title", "content", "favourite_count", "author", "created_date", "edited_date") admin.site.register(Entry, EntryAdmin)
altuntasmuhammet/eksisozluk-scraper
altuntas/eksisozlukbot/__init__.py
from six import iteritems from eksisozlukbot.eksisozlukbot import spiders from eksisozlukbot.eksisozlukbot import items from eksisozlukbot.eksisozlukbot import middlewares from eksisozlukbot.eksisozlukbot import pipelines from eksisozlukbot.eksisozlukbot import settings _SUBMODULES = { "spiders": spiders, "items": items, "middlewares": middlewares, "pipelines": pipelines, "settings": settings, } import sys for module_name, module in iteritems(_SUBMODULES): sys.modules["eksisozlukbot.%s" % module_name] = module
altuntasmuhammet/eksisozluk-scraper
altuntas/eksisozlukbot/eksisozlukbot/settings.py
<filename>altuntas/eksisozlukbot/eksisozlukbot/settings.py # Scrapy settings for eksisozlukbot project # # For simplicity, this file contains only settings considered important or # commonly used. You can find more settings consulting the documentation: # # https://docs.scrapy.org/en/latest/topics/settings.html # https://docs.scrapy.org/en/latest/topics/downloader-middleware.html # https://docs.scrapy.org/en/latest/topics/spider-middleware.html import os import sys import django sys.path.append( os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) ) sys.path.append( os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "..") ) sys.path.append( os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "../../") ) os.environ['DJANGO_SETTINGS_MODULE'] = 'altuntas.settings' django.setup() # from django.conf import settings BOT_NAME = 'eksisozlukbot' SPIDER_MODULES = ['eksisozlukbot.spiders'] NEWSPIDER_MODULE = 'eksisozlukbot.spiders' # Crawl responsibly by identifying yourself (and your website) on the user-agent #USER_AGENT = 'eksisozlukbot (+http://www.yourdomain.com)' # Obey robots.txt rules ROBOTSTXT_OBEY = True # Configure maximum concurrent requests performed by Scrapy (default: 16) #CONCURRENT_REQUESTS = 32 # Configure a delay for requests for the same website (default: 0) # See https://docs.scrapy.org/en/latest/topics/settings.html#download-delay # See also autothrottle settings and docs #DOWNLOAD_DELAY = 3 # The download delay setting will honor only one of: #CONCURRENT_REQUESTS_PER_DOMAIN = 16 #CONCURRENT_REQUESTS_PER_IP = 16 # Disable cookies (enabled by default) #COOKIES_ENABLED = False # Disable Telnet Console (enabled by default) #TELNETCONSOLE_ENABLED = False # Override the default request headers: #DEFAULT_REQUEST_HEADERS = { # 'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8', # 'Accept-Language': 'en', #} # Enable or disable spider middlewares # See https://docs.scrapy.org/en/latest/topics/spider-middleware.html #SPIDER_MIDDLEWARES = { # 'eksisozlukbot.middlewares.EksisozlukbotSpiderMiddleware': 543, #} # Enable or disable downloader middlewares # See https://docs.scrapy.org/en/latest/topics/downloader-middleware.html #DOWNLOADER_MIDDLEWARES = { # 'eksisozlukbot.middlewares.EksisozlukbotDownloaderMiddleware': 543, #} # Enable or disable extensions # See https://docs.scrapy.org/en/latest/topics/extensions.html #EXTENSIONS = { # 'scrapy.extensions.telnet.TelnetConsole': None, #} # Configure item pipelines # See https://docs.scrapy.org/en/latest/topics/item-pipeline.html ITEM_PIPELINES = { 'eksisozlukbot.pipelines.EksisozlukbotPipeline': 300, } # # Configure database # DATABASE = { # "drivername": "postgresql", # "host": settings.POSTGRES_HOST, # "port": settings.POSTGRES_PORT, # "username": settings.POSTGRES_USER, # "password": settings.POSTGRES_PASSWORD, # "database": settings.POSTGRES_DB, # } # Configure logging LOG_LEVEL = "INFO" # Enable and configure the AutoThrottle extension (disabled by default) # See https://docs.scrapy.org/en/latest/topics/autothrottle.html #AUTOTHROTTLE_ENABLED = True # The initial download delay #AUTOTHROTTLE_START_DELAY = 5 # The maximum download delay to be set in case of high latencies #AUTOTHROTTLE_MAX_DELAY = 60 # The average number of requests Scrapy should be sending in parallel to # each remote server #AUTOTHROTTLE_TARGET_CONCURRENCY = 1.0 # Enable showing throttling stats for every response received: #AUTOTHROTTLE_DEBUG = False # Enable and configure HTTP caching (disabled by default) # See https://docs.scrapy.org/en/latest/topics/downloader-middleware.html#httpcache-middleware-settings #HTTPCACHE_ENABLED = True #HTTPCACHE_EXPIRATION_SECS = 0 #HTTPCACHE_DIR = 'httpcache' #HTTPCACHE_IGNORE_HTTP_CODES = [] #HTTPCACHE_STORAGE = 'scrapy.extensions.httpcache.FilesystemCacheStorage'
altuntasmuhammet/eksisozluk-scraper
altuntas/eksisozlukbot/eksisozlukbot/pipelines.py
# Define your item pipelines here # # Don't forget to add your pipeline to the ITEM_PIPELINES setting # See: https://docs.scrapy.org/en/latest/topics/item-pipeline.html # useful for handling different item types with a single interface from itemadapter import ItemAdapter from scraper.models.eksisozlukbot import Entry class EksisozlukbotPipeline: def process_item(self, item, spider): """ Process the item and store to database. """ entry = Entry.objects.filter(eksisozluk_entry_id=item['eksisozluk_entry_id']) if entry: return item entry_item = Entry(**item) entry_item.save() return item
pawansingh126/yaml_editor
yaml_editor/editor.py
<filename>yaml_editor/editor.py<gh_stars>1-10 ############################################################################## # # Author : <NAME> # Email : <EMAIL> # Date : Oct 2018 # ############################################################################## import click import json import os import socket import yaml from flask import Flask, request, render_template from flask_bootstrap import Bootstrap app_path = os.path.dirname(os.path.abspath(__file__)) app = Flask('Yaml Editor!', template_folder=os.path.join(app_path, 'templates'), static_folder=os.path.join(app_path, 'static')) Bootstrap(app) @app.route('/', methods=['GET', 'POST']) def index(): """Renders index page to edit provided yaml file.""" with open(app.config['YAML_FILE_OBJ']) as file_obj: data = yaml.load(file_obj, Loader=yaml.Loader) return render_template('yaml.html', data=json.dumps(data), change_str=app.config['STRING_TO_CHANGE']) @app.route('/tree', methods=['GET', 'POST']) def tree(): """Renders tree view page to edit provided yaml file.""" with open(app.config['YAML_FILE_OBJ']) as file_obj: data = yaml.load(file_obj, Loader=yaml.Loader) return render_template('treeyaml.html', data=data, datastr=json.dumps(data), change_str=app.config['STRING_TO_CHANGE']) @app.route('/save', methods=['POST']) def save(): """Save current progress on file.""" out = request.json.get('yaml_data') with open(app.config['YAML_FILE_OBJ'], 'w') as file_obj: yaml.dump(out, file_obj, default_flow_style=False) return "Data saved successfully!" @app.route('/saveExit', methods=['POST']) def save_exit(): """Save current progress on file and shuts down the server.""" out = request.json.get('yaml_data') with open(app.config['YAML_FILE_OBJ'], 'w') as file_obj: yaml.dump(out, file_obj, default_flow_style=False) func = request.environ.get('werkzeug.server.shutdown') if func: func() return "Saved successfully, Shutting down app! You may close the tab!" @app.errorhandler(404) def page_not_found(e): """Serves 404 error.""" return '<h1>404: Page not Found!</h1>' def run(*args, **kwargs): """Starts the server.""" port = kwargs.get('port', None) if not port: port = 8161 app.run(host='0.0.0.0', port=port, debug=False) @click.command() @click.option( '--file', '-f', required=True, type=click.Path(exists=True), multiple=False, help="Path with file name to the intermediary yaml file." ) @click.option( '--port', '-p', default=8161, type=click.INT, multiple=False, help="Optional port parameter to run Flask on." ) @click.option( '--string', '-s', default='#CHANGE_ME', type=click.STRING, multiple=False, help="Text which is required to be changed on yaml file." ) def main(*args, **kwargs): print("Please go to http://{0}:{1}/ to edit your yaml file.".format( socket.gethostbyname(socket.gethostname()), kwargs['port'])) app.config['YAML_FILE_OBJ'] = kwargs['file'] app.config['STRING_TO_CHANGE'] = kwargs['string'] run(kwargs) if __name__=='__main__': """Invoked when used as a script.""" main()
ankit-vaghela30/sign-language-recognition
sign-language-recognition/image_obj_detect.py
<filename>sign-language-recognition/image_obj_detect.py #!flask/bin/python from flask import Flask, render_template, request from flask_cors import CORS, cross_origin import base64 import json # from datetime import datetime import numpy as np import sys import tensorflow as tf import os from threading import Thread from datetime import datetime import cv2 from collections import defaultdict from utils import label_map_util from protos import string_int_label_map_pb2 import multiprocessing from multiprocessing import Queue, Pool import time import datetime import argparse import os os.environ["CUDA_VISIBLE_DEVICES"]="-1" import tensorflow as tf from keras.models import load_model import sklearn import skimage from skimage.transform import resize from keras.backend import clear_session app = Flask(__name__) CORS(app, support_credentials=True) _score_thresh = 0.27 MODEL_NAME = 'hand_inference_graph' # Path to frozen detection graph. This is the actual model that is used for the object detection. PATH_TO_CKPT = MODEL_NAME + '/frozen_inference_graph.pb' # List of the strings that is used to add correct label for each box. PATH_TO_LABELS = os.path.join(MODEL_NAME, 'hand_label_map.pbtxt') # In[5]: NUM_CLASSES = 1 # load label map label_map = label_map_util.load_labelmap('hand_label_map.pbtxt') categories = label_map_util.convert_label_map_to_categories( label_map, max_num_classes=NUM_CLASSES, use_display_name=True) category_index = label_map_util.create_category_index(categories) def load_inference_graph(): # load frozen tensorflow model into memory print("> ====== loading HAND frozen graph into memory") detection_graph = tf.Graph() with detection_graph.as_default(): od_graph_def = tf.GraphDef() with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid: serialized_graph = fid.read() od_graph_def.ParseFromString(serialized_graph) tf.import_graph_def(od_graph_def, name='') sess = tf.Session(graph=detection_graph) print("> ====== Hand Inference graph loaded.") return detection_graph, sess video_source = 0 # help='Device index of the camera. num_hands=2 # help='Max number of hands to detect.') fps=1 # help='Show FPS on detection/display visualization') width=300 # help='Width of the frames in the video stream.') height=200 # help='Height of the frames in the video stream.') display=1 # help='Display the detected images using OpenCV. This reduces FPS') num_workers=4 # help='Number of workers.') queue_size=5 score_thresh = 0.2 print(">> loading frozen model for worker") print('Loading model...') global model_classification model_classification = load_model('vgg16model.h5') detection_graph, sess = load_inference_graph() # global graph2 graph = tf.get_default_graph() print('model loaded') @app.route('/upload', methods=['POST']) @cross_origin(supports_credentials=True) def upload_base64_img(): global graph with graph.as_default(): def detect_objects(image_np, detection_graph, sess): # Definite input and output Tensors for detection_graph image_tensor = detection_graph.get_tensor_by_name('image_tensor:0') # Each box represents a part of the image where a particular object was detected. detection_boxes = detection_graph.get_tensor_by_name( 'detection_boxes:0') # Each score represent how level of confidence for each of the objects. # Score is shown on the result image, together with the class label. detection_scores = detection_graph.get_tensor_by_name( 'detection_scores:0') detection_classes = detection_graph.get_tensor_by_name( 'detection_classes:0') num_detections = detection_graph.get_tensor_by_name( 'num_detections:0') image_np_expanded = np.expand_dims(image_np, axis=0) (boxes, scores, classes, num) = sess.run( [detection_boxes, detection_scores, detection_classes, num_detections], feed_dict={image_tensor: image_np_expanded}) return np.squeeze(boxes), np.squeeze(scores) def draw_box_on_image(num_hands_detect, score_thresh, scores, boxes, im_width, im_height, image_np): for i in range(num_hands_detect): if (scores[i] > score_thresh): (left, right, top, bottom) = (boxes[i][1] * im_width*0.8, boxes[i][3] * im_width*1.1, boxes[i][0] * im_height*0.8, boxes[i][2] * im_height*1.1) p1 = (int(left), int(top)) p2 = (int(right), int(bottom)) x= int(left) y= int(top) w=int(right) h=int(bottom) roi=image_np[y:y+h,x:x+w] #cv2.imwrite('image.png',roi) map_characters = {0: 'A', 1: 'B', 2: 'C', 3: 'D', 4: 'E', 5: 'F', 6: 'G', 7: 'H', 8: 'I', 9: 'J', 10: 'K', 11: 'L', 12: 'M', 13: 'N', 14: 'O', 15: 'P', 16: 'Q', 17: 'R', 18: 'S', 19: 'T', 20: 'U', 21: 'V', 22: 'W', 23: 'X', 24: 'Y', 25: 'Z', 26: 'del', 27: '-', 28: 'space'} X_ = [] imageSize=50 img_file_=roi img_file_ = skimage.transform.resize(img_file_, (imageSize, imageSize, 3)) img_arr_ = np.asarray(img_file_) X_.append(img_arr_) X_ = np.asarray(X_) y_pred = model_classification.predict(X_) Y_pred_classes = np.argmax(y_pred,axis = 1) print('Predicting the gesture') print(map_characters.get(Y_pred_classes[0])) cv2.imwrite('image.png', roi) cv2.rectangle(image_np, p1, p2, (77, 255, 9), 3, 1) def get_pred(image_path): # detection_graph, sess = load_inference_graph() sess = tf.Session(graph=detection_graph) #print("> ===== in worker loop, frame ", frame_processed) frame = cv2.imread(image_path) pred = '-' ACCURACY = 0 #frame_processed = 0 #import matplotlib.pyplot as plt #lt.imshow(frame) if (frame is not None): frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) cap_params = {} cap_params['im_height'], cap_params['im_width'] ,x = frame.shape cap_params['score_thresh'] = score_thresh # max number of hands we want to detect/track cap_params['num_hands_detect'] = num_hands # Actual detection. Variable boxes contains the bounding box cordinates for hands detected, # while scores contains the confidence for each of these boxes. # Hint: If len(boxes) > 1 , you may assume you have found atleast one hand (within your score threshold) boxes, scores = detect_objects( frame, detection_graph, sess) # draw bounding boxes im_width = cap_params['im_width'] im_height = cap_params['im_height'] image_np = frame for i in range(cap_params['num_hands_detect']): if (scores[i] > cap_params["score_thresh"]): (left, right, top, bottom) = (boxes[i][1] * im_width*0.8, boxes[i][3] * im_width*1.1, boxes[i][0] * im_height*0.8, boxes[i][2] * im_height*1.1) # (left, right, top, bottom) = (boxes[i][1] * im_width, boxes[i][3] * im_width, # boxes[i][0] * im_height, boxes[i][2] * im_height) p1 = (int(left), int(top)) p2 = (int(right), int(bottom)) x= int(left) y= int(top) w=int(right) h=int(bottom) roi=image_np[y:y+h,x:x+w] #cv2.imwrite('image.png',roi) map_characters = {0: 'A', 1: 'B', 2: 'C', 3: 'D', 4: 'E', 5: 'F', 6: 'G', 7: 'H', 8: 'I', 9: 'J', 10: 'K', 11: 'L', 12: 'M', 13: 'N', 14: 'O', 15: 'P', 16: 'Q', 17: 'R', 18: 'S', 19: 'T', 20: 'U', 21: 'V', 22: 'W', 23: 'X', 24: 'Y', 25: 'Z', 26: 'del', 27: 'nothing', 28: 'space'} X_ = [] imageSize=50 img_file_=roi img_file_ = skimage.transform.resize(img_file_, (imageSize, imageSize, 3)) img_arr_ = np.asarray(img_file_) X_.append(img_arr_) X_ = np.asarray(X_) y_pred = model_classification.predict(X_) Y_pred_classes = np.argmax(y_pred,axis = 1) ACCURACY=y_pred[0][Y_pred_classes[0]]*100 # if (ACCURACY > 80): print('Predicting the gesture') print(map_characters.get(Y_pred_classes[0])) pred = map_characters.get(Y_pred_classes[0]) cv2.imwrite('roi.jpg', roi) # cv2.rectangle(image_np, p1, p2, (77, 255, 9), 3, 1) # add frame annotated with bounding box to queue #cv2.imwrite('image_obj_det_hands.png',frame) #frame_processed += 1 # clear_session() # sess.close() return pred, ACCURACY content = request.get_json() # checking if the image is present or not. if 'image' not in content: # abort(400) # abort(Response('No Image data received')) return 'Image not received' imgdata = base64.b64decode(content['image']) # (dt, micro) = datetime.utcnow().strftime('%Y%m%d%H%M%S.%f').split('.') # dt = "%s%03d" % (dt, int(micro) / 1000) filename = 'webcamImg/some_image.jpg' with open(filename, 'wb') as f: f.write(imgdata) # print("--------------------->>>>>>", filename) letter, acc = get_pred(filename) result = { "hand_object": "http://localhost:8888/ada/roi.jpg", "letter_detected": str(letter), "accuracy": acc } return json.dumps(result) app.run(port=5000, debug=True)
ankit-vaghela30/sign-language-recognition
sign-language-recognition/gesturedetector.py
import skimage from skimage.transform import resize from keras.models import load_model import numpy as np import cv2 import os os.environ["CUDA_VISIBLE_DEVICES"]="-1" print('Loading model...') model = load_model('vgg16model.h5') print('model loaded') map_characters = {0: 'A', 1: 'B', 2: 'C', 3: 'D', 4: 'E', 5: 'F', 6: 'G', 7: 'H', 8: 'I', 9: 'J', 10: 'K', 11: 'L', 12: 'M', 13: 'N', 14: 'O', 15: 'P', 16: 'Q', 17: 'R', 18: 'S', 19: 'T', 20: 'U', 21: 'V', 22: 'W', 23: 'X', 24: 'Y', 25: 'Z', 26: 'del', 27: 'nothing', 28: 'space'} while (True): X_ = [] imageSize=50 img_file_ = cv2.imread('image.png') if img_file_ is not None: img_file_ = skimage.transform.resize(img_file_, (imageSize, imageSize, 3)) img_arr_ = np.asarray(img_file_) X_.append(img_arr_) X_ = np.asarray(X_) y_pred = model.predict(X_) Y_pred_classes = np.argmax(y_pred,axis = 1) print('Predicting the gesture') print(map_characters.get(Y_pred_classes[0]))
ankit-vaghela30/sign-language-recognition
web-service/analyze.py
<filename>web-service/analyze.py #!flask/bin/python from flask import Flask, render_template, request from flask_cors import CORS, cross_origin import base64 import json #import time from datetime import datetime app = Flask(__name__) CORS(app, support_credentials=True) @app.route('/upload', methods=['POST']) @cross_origin(supports_credentials=True) def upload_base64_img(): content = request.get_json() # checking if the image is present or not. if 'image' not in content: abort(400) abort(Response('No Image data received')) imgdata = base64.b64decode(content['image']) (dt, micro) = datetime.utcnow().strftime('%Y%m%d%H%M%S.%f').split('.') dt = "%s%03d" % (dt, int(micro) / 1000) filename = 'images/'+dt+'.jpg' with open(filename, 'wb') as f: f.write(imgdata) s = int(dt)%65 + 65 result = { "hand_object": str(base64.b64encode(imgdata)), "letter_detected": chr(s) } return json.dumps(result) app.run(port=5000)
ramonmeza/raycasting-py
tests/template/features/steps/template.py
@given('some state of an object') # type: ignore # noqa: F821 def step_impl(context) -> None: # noqa: F811 context.test_str = 'change me' @when('we change that object') # type: ignore # noqa: F821 def step_impl(context) -> None: # noqa: F811 context.test_str = 'you\'ve changed' @then('assert that the change hasn\'t broken shit') # type: ignore # noqa: F821,E501 def step_impl(context) -> None: # noqa: F811 assert context.test_str == 'you\'ve changed'
LZP4GitHub/ssd.pytorch
data/dota.py
<reponame>LZP4GitHub/ssd.pytorch """dota dataset classes original author: <NAME> https://github.com/fmassa/vision/blob/voc_dataset/torchvision/datasets/voc.py updated by: ellis brown, max degroot for dota file list is: |data/ |data/trainvalsplit/ |data/testsplit/ |data|trainvalsplit/images/ |data|trainvalsplit/labeltxt/ |data|trainvalsplit/train.txt the image folder is xxx.jpg """ from .config import HOME import os.path as osp import sys import torch import torch.utils.data as data import cv2 import numpy as np import zipfile import codecs DOTA_CLASSES = ( # always index 0 'plane', 'baseball-diamond', 'bridge', 'ground-track-field', 'small-vehicle', 'large-vehicle', 'ship', 'tennis-court', 'basketball-court', 'storage-tank', 'soccer-ball-field', 'roundabout', 'harbor', 'swimming-pool', 'helicopter') # note: if you used our download scripts, this should be right #DOTA_ROOT = osp.join(HOME, '/data/dota-split-300/') DOTA_ROOT = '/home/lzp/data/dota-split-300/' class DOTAAnnotationTransform(object): """transforms a dota annotation into a tensor of bbox coords and label index initilized with a dictionary lookup of classnames to indexes arguments: class_to_ind (dict, optional): dictionary lookup of classnames -> indexes (default: alphabetic indexing of voc's 20 classes) keep_difficult (bool, optional): keep difficult instances or not (default: false) height (int): height width (int): width """ def __init__(self, class_to_ind=None, keep_difficult=True): self.class_to_ind = class_to_ind or dict( zip(DOTA_CLASSES, range(len(DOTA_CLASSES)))) self.keep_difficult = keep_difficult def __call__(self, targets, width, height): """ arguments: target (annotation) : the target annotation to be made usable will be an et.element returns: a list containing lists of bounding boxes [bbox coords, class name] """ res = [] for target in targets: difficult = (int(target[9]) != 0) #0 -> easy if not self.keep_difficult and difficult: continue name = target[8].lower().strip() pts = ['x1', 'y1', 'x2', 'y2','x3', 'y3', 'x4', 'y4'] bndbox = [] # for i,pt in enumerate(pts): # if (i == 0 or i == 1 or i == 6 or i == 7): # # cur_pt = max(float(target[i]) - 1, 0) # cur_pt = cur_pt / width if i % 2 == 0 else cur_pt / height # bndbox.append(cur_pt) xmin, xmax, ymin, ymax = min(target[0], min(target[2], min(target[4], target[6]))), \ max(target[0], max(target[2], max(target[4], target[6]))), \ min(target[1], min(target[3], min(target[5], target[7]))), \ max(target[1], max(target[3], max(target[5], target[7]))) cur_pt_1 = max(float(xmin) - 1.0, 0.0) cur_pt_1 = 1.0*cur_pt_1 / width bndbox.append(cur_pt_1) cur_pt_2 = max(float(ymin) - 1.0, 0.0) cur_pt_2 = 1.0*cur_pt_2 / height bndbox.append(cur_pt_2) cur_pt_3 = max(float(xmax) - 1.0, 0.0) cur_pt_3 = 1.0*cur_pt_3 / width bndbox.append(cur_pt_3) cur_pt_4 = max(float(ymax) - 1.0, 0.0) cur_pt_4 = 1.0*cur_pt_4 / height bndbox.append(cur_pt_4) label_idx = self.class_to_ind[name] bndbox.append(label_idx) res += [bndbox] # [xmin, ymin, xmax, ymax, label_ind] return res class DOTADetection(data.Dataset): """voc detection dataset object input is image, target is annotation arguments: root (string): filepath to vocdevkit folder. image_set (string): imageset to use (eg. 'train', 'val', 'test') transform (callable, optional): transformation to perform on the input image target_transform (callable, optional): transformation to perform on the target `annotation` (eg: take in caption string, return tensor of word indices) dataset_name (string, optional): which dataset to load (default: 'voc2007') """ def __init__(self, root, image_sets=['trainvalsplit'], transform=None, target_transform=DOTAAnnotationTransform(), dataset_name='DOTA', train_test="train"): self.root = root self.image_set = image_sets self.transform = transform self.target_transform = target_transform self.name = dataset_name train_rootpath = osp.join(self.root, 'trainvalsplit') self._annopath = osp.join(train_rootpath, 'labelTxt', '%s.txt') self._imgpath = osp.join(train_rootpath, 'images', '%s.png') self.ids = list() if train_test == "train": for line in open(osp.join(train_rootpath, 'train11.txt')): self.ids.append((train_rootpath, line.strip())) else: for line in open(osp.join(train_rootpath, 'test.txt')): print(line) self.ids.append((train_rootpath, line.strip())) # for line in open(osp.join(train_rootpath, 'train11.txt')): # self.ids.append((train_rootpath, line.strip())) def __getitem__(self, index): im, gt, h, w = self.pull_item(index) return im, gt def __len__(self): return len(self.ids) def pull_item(self, index): img_id = self.ids[index] # print(img_id) anno_file_path = (self._annopath % img_id[1]) # print(anno_file_path) img = cv2.imread(self._imgpath % img_id[1]) # print("Image Path") # print(self._imgpath % img_id[1]) height, width, channels = img.shape # print("Image Channel:") # print(height,width,channels) #target = et.parse(self._annopath % img_id).getroot() f = codecs.open(anno_file_path,"r") targets = f.readlines() f.close() # print("GT is:") # print(targets) targets = [target.strip().strip('\n').split(' ') for target in targets] # print("targets11111111.size") # print(np.shape(targets)) if self.target_transform is not None: targets = self.target_transform(targets, width, height) # print("After target_transform GT is:") # print(targets) if self.transform is not None: targets = np.array(targets) # print("After transform GT is:") # print(targets) # if targets.size < 8: # avoide the null .txt # # return torch.from_numpy(img).permute(2, 0, 1)., [], height, width # return torch.from_numpy(img).permute(2, 0, 1).float(), [], height, width img, boxes, labels = self.transform(img, targets[:, 0:4], targets[:,4]) # to rgb img = img[:, :, (2, 1, 0)] # img = img.transpose(2, 0, 1) targets = np.hstack((boxes, np.expand_dims(labels, axis=1))) # return torch.from_numpy(img).permute(2, 0, 1), targets, height, width return torch.from_numpy(img).permute(2, 0, 1), targets, height, width # .float() # return torch.from_numpy(img), targets, height, width def pull_image(self, index): '''returns the original image object at index in pil form note: not using self.__getitem__(), as any transformations passed in could mess up this functionality. argument: index (int): index of img to show return: pil img ''' img_id = self.ids[index] return cv2.imread(self._imgpath % img_id[1], cv2.IMREAD_COLOR) #.float() def pull_anno(self, index): '''returns the original annotation of image at index note: not using self.__getitem__(), as any transformations passed in could mess up this functionality. argument: index (int): index of img to get annotation of return: list: [img_id, [(label, bbox coords),...]] eg: ('001718', [('dog', (96, 13, 438, 332))]) ''' img_id = self.ids[index] #anno = et.parse(self._annopath % img_id).getroot() f = codecs.open(self._annopath % img_id[1],"r") anno = f.readlines() f.close() anno = [target.strip().split(' ') for target in anno] gt = self.target_transform(anno, 1, 1) return img_id[1], gt def pull_tensor(self, index): '''returns the original image at an index in tensor form note: not using self.__getitem__(), as any transformations passed in could mess up this functionality. argument: index (int): index of img to show return: tensorized version of img, squeezed ''' return torch.Tensor(self.pull_image(index)).unsqueeze_(0)
LZP4GitHub/ssd.pytorch
test.py
<reponame>LZP4GitHub/ssd.pytorch from __future__ import print_function import sys import os import argparse import torch import torch.nn as nn import torch.backends.cudnn as cudnn import torchvision.transforms as transforms from torch.autograd import Variable # from data import VOC_ROOT, VOC_CLASSES as labelmap from PIL import Image # from data import DOTA_ROOT, VOC_ROOT, VOC_CLASSES as labelmap from data import DOTA_ROOT,DOTA_CLASSES as labelmap from data import DOTAAnnotationTransform, DOTADetection, BaseTransform, DOTA_CLASSES import torch.utils.data as data from ssd import build_ssd import numpy as np parser = argparse.ArgumentParser(description='Single Shot MultiBox Detection') parser.add_argument('--trained_model', default='weights/ssd300_DOTA_66800.pth', type=str, help='Trained state_dict file path to open') parser.add_argument('--save_folder', default='eval/', type=str, help='Dir to save results') parser.add_argument('--visual_threshold', default=0.1, type=float, help='Final confidence threshold') parser.add_argument('--cuda', default=True, type=bool, help='Use cuda to train model') parser.add_argument('--dota_root', default=DOTA_ROOT, help='Location of VOC root directory') parser.add_argument('-f', default=None, type=str, help="Dummy arg so we can load in Jupyter Notebooks") args = parser.parse_args() if args.cuda and torch.cuda.is_available(): torch.set_default_tensor_type('torch.cuda.FloatTensor') else: torch.set_default_tensor_type('torch.FloatTensor') if not os.path.exists(args.save_folder): os.mkdir(args.save_folder) def test_net(save_folder, net, cuda, testset, transform, thresh): # dump predictions and assoc. ground truth to text file for now filename = save_folder+'test1.txt' num_images = len(testset) for i in range(num_images): print('Testing image {:d}/{:d}....'.format(i+1, num_images)) img = testset.pull_image(i) img_id, annotation = testset.pull_anno(i) x = torch.from_numpy(transform(img)[0]).permute(2, 0, 1) x = Variable(x.unsqueeze(0)) filename_i = os.path.join('/home/lzp/Code/ssd_DOTA_rectangle/',save_folder , img_id+'.txt') # with open(filename_i, mode='w') as f: # f.write('\nGROUND TRUTH FOR: '+img_id+'\n') # for box in annotation: # f.write('label: '+' || '.join(str(b) for b in box)+'\n') if cuda: x = x.cuda() y = net(x) # forward pass detections = y.data print("detections:",detections.size()) # scale each detection back up to the image scale = np.array([img.shape[1], img.shape[0], img.shape[1], img.shape[0]]) pred_num = 0 detections = detections.cpu().numpy() index_top100 = np.unravel_index(np.argsort(detections[0, :, :, 0].ravel())[-100:], detections[0, :, :, 0].shape) for i in range(100): score = detections[0, index_top100[0][i], index_top100[1][i], 0] label_name = labelmap[ index_top100[0][i] -1] pt = (detections[0, index_top100[0][i], index_top100[1][i], 1:]*scale) #.cpu().numpy() coords = (max(pt[0],0), max(pt[1],0), max(pt[2],0), max(pt[3],0)) pred_num += 1 print("filename_i",filename_i) with open(filename_i, mode='a') as f: print(filename_i) f.write(( ' '.join(str(c) for c in coords ))+' '+ label_name + ' '+ str(score).strip().strip(')').strip('tensor(') + '\n') def test_dota(): # load net num_classes = len(DOTA_CLASSES) + 1 # +1 background net = build_ssd('test', 300, num_classes) # initialize SSD net.load_state_dict(torch.load(args.trained_model)) net.eval() print('Finished loading model!') # load data testset = DOTADetection(root = args.dota_root,train_test ="test") if args.cuda: net = net.cuda() cudnn.benchmark = True # evaluation test_net(args.save_folder, net, args.cuda, testset, BaseTransform(net.size, (104, 117, 123)), thresh=args.visual_threshold) if __name__ == '__main__': test_dota()
jancervenka/acf
acf/core/computation.py
<gh_stars>0 #!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> import logging import functools import multiprocessing as mp from typing import Any, Callable, Iterable, List, Optional, Tuple import numpy as np import pandas as pd from .preprocessing import create_user_item_matrix from .utils import ( cast_numeric_greater_than_zero, check_columns_in_dataframe, check_feedback_column_numeric, get_index_position, drop_warn_na, ) DEFAULT_REG_LAMBDA = 0.1 DEFAULT_ALPHA = 40 DEFAULT_N_FACTORS = 10 DEFAULT_N_ITER = 20 DEFAULT_N_JOBS = 1 # TODO: sparse R dataframe/matrix? # TODO: X/Y init - which distributions? # TODO: _compute_factors_1d, _get_least_square_sum to algebra.py module? class Engine: """ Collaborative filtering algorithm based on matrix factorization by alternating least squares. Designed for implicit feedback datasets. The class exposes two public methods, `fit` to train the model and `predict` to produce the recommendations. Example: ``` import acf import pandas as pd # assuming the data are in the following format: # | user_id | item_id | feedback | # |---------|---------|----------| # | 2491 | 129 | 2 | interactions = pd.read_csv('interactions.csv') engine = acf.Engine(reg_lambda=1, alpha=35, n_factors=2) engine.fit(interactions, user_column='user_id', item_column='item_id', feedback_column='feedback', n_iter=20, n_jobs=4) # get the best 20 recommendation for given user prediction = engine.predict(user=2491, top_n=20) # to print training loss value at every iteration print(engine.loss) ``` """ def __init__( self, reg_lambda: float = DEFAULT_REG_LAMBDA, alpha: float = DEFAULT_ALPHA, n_factors: int = DEFAULT_N_FACTORS, random_state: Optional[int] = None, ): """ Initializes the class with model hyperparameters and `random_state`. Args: reg_lambda: regularization parameter alpha: confidence control parameter n_factors: number of latent factors random_state: RNG seed """ self._reg_lambda = cast_numeric_greater_than_zero( reg_lambda, "reg_lambda", float ) self._alpha = cast_numeric_greater_than_zero(alpha, "alpha", float) self._n_factors = cast_numeric_greater_than_zero(n_factors, "n_factors", int) if random_state: self._rng = np.random.RandomState(random_state) else: self._rng = np.random.RandomState() self._X = self._Y = None self._user_index = self._item_index = None self._loss = [] @staticmethod def _compute_factors_1d( feedback_1d: np.array, other_factors: np.array, other_factors_small: np.array, reg_lambda: float, alpha: float, ) -> np.array: """ Computes a 1-dimensional factor array for either one user or one item. When computing user factors, `feedback_1` is an array of interactions `r_u` between user `u` and every item, `other_factors` is an item factor matrix `Y` and `other_factors_small` is `Y ^ T * Y`. When computing the user factors `x_u` for user `u`, the function solves a linear system ``` (Y^T * Y + Y^T * (C^u - I) * Y + λ * I) * x_u = Y^T * C^U * p_u ``` where `Y` is an item factor matrix, `C^u` is a diagonal matrix containing feedback `r_u` of user `u` for each item `i` (the feedback is transformed to `c_ii = 1 + alpha * r_ui`), `p_u` is a binary vector of preferences of user `u` for each item, and `λ` is regularization lambda. The computation is symmetric for item factors. Args: feedback_1d: one R user-item matrix row or column other_factors: either X or Y factor matrix other_factors_small: either X^T * X or Y^T * Y matrix :reg_lambda: regularization parameter alpha: confidence control parameter Returns: 1-dimensional factor array for one user/item """ p_ = (feedback_1d > 0).astype("uint8") C_ = np.diag(1 + alpha * feedback_1d) size, f = other_factors.shape M = other_factors_small + np.linalg.multi_dot( [other_factors.T, C_ - np.eye(size), other_factors] ) return np.linalg.solve( M + reg_lambda * np.eye(f), np.linalg.multi_dot([other_factors.T, C_, p_]) ) def _create_worker( self, other_factors: np.array, other_factors_small: np.array ) -> Callable: """ Creates a callable worker from `Engine._compute_factors_1d` for application over `R` rows/columns in multiprocessing map. The worker takes only one argument (`R` column/row), other parameters are set in this function using `functools.partial` and are identical for all given rows/columns. Args: other_factors: either X or Y factor matrix other_factors_small: either X^T * X or Y^T * Y matrix Returns: callable worker """ return functools.partial( self._compute_factors_1d, other_factors=other_factors, other_factors_small=other_factors_small, reg_lambda=self._reg_lambda, alpha=self._alpha, ) @staticmethod def _feedback_1d_generator(R: pd.DataFrame, axis: int) -> Iterable[np.array]: """ Creates a generator from `R` rows or columns. Each item in the generator is a numpy array. Args: R: user-item feedback matrix axis: axis=1 for rows, axis=0 for columns Returns: R row/column generator """ for _, r in R.iterrows() if axis else R.iteritems(): yield r.values def _compute_factors( self, pool, R: pd.DataFrame, other_factors: np.array, axis: int ) -> np.array: """ Applies `Engine._compute_factors_1d` over `R` rows/columns to compute factors for all user/items. Args: pool: multiprocessing pool R: user-item feedback matrix other_factors: either X or Y factor matrix axis: axis=1 for rows, axis=0 for columns Returns: either X or Y factor matrix """ # TODO: axis can be infered from other_factors shape other_factors_small = np.dot(other_factors.T, other_factors) worker = self._create_worker(other_factors, other_factors_small) all_feedback_1d = self._feedback_1d_generator(R, axis) return np.vstack(pool.map(worker, all_feedback_1d)) def _initialize_factors(self, m: int, n: int) -> Tuple[np.array]: """ Initialzes `X`, `Y` matrices with random values. Args: m: number of users n: number of items Returns: tuple of X, Y factor matrices """ return (self._rng.rand(m, self._n_factors), self._rng.rand(n, self._n_factors)) def _get_loss(self, X: np.array, Y: np.array, R: np.array) -> np.float: """ Computes least squares sum for given `X`, `Y`, `R`. The loss is defined as `loss := Σ_ui (C * (P - X * Y^T)^2)` where `C` is confidence matrix, `P` is preference matrix, and `X`, `Y` are factor matrices. Args: X: user factors matrix Y: item factors matrix R: user-item feedback matrix Returns: loss value """ P = (R > 0).astype("uint8") C = 1 + self._alpha * R return np.sum(C * (P - np.dot(X, Y.T)) ** 2) @staticmethod def _log_iteration(iteration: int, loss: np.float) -> None: """ Logs least squares `loss` for given `iteration`. Args: iteration: training iterataion loss: value to log """ logging.info(f"Iteration {iteration:04d} Loss Σ = {loss:.5f}") def _run_iterations(self, pool, R: pd.DataFrame, n_iter: int) -> Tuple[np.array]: """ Initializes `X`, `Y` matrices and runs alternating least squares iterations. Args: pool: multiprocessing pool R: user-item feedback matrix n_iter: number of iterations Returns: tuple of X, Y factor matrices """ X, Y = self._initialize_factors(*R.shape) for i in range(n_iter): X = self._compute_factors(pool, R, Y, axis=1) # rows Y = self._compute_factors(pool, R, X, axis=0) # columns self._loss.append(self._get_loss(X, Y, R.values)) self._log_iteration(i, self._loss[-1]) return X, Y def fit( self, interactions: pd.DataFrame, user_column: str, item_column: str, feedback_column: str, n_iter: int = DEFAULT_N_ITER, n_jobs: int = DEFAULT_N_JOBS, ) -> None: """ Fits the model by factorizing `interactions` into latent factors. Dataframe `interactions` contains implicit feedbacks values for user-item pairs in three columns `user_column`, `item_column`, `feedback_column`. ``` | user_column | item_column | feedback_column | |-------------|-------------|--------------------| | user_id_0 | item_id_0 | 0_1_feedback_value | ``` Args: interactions: user-item feedback pairs user_column: name of the column containing user ids item_column: name of the column containing item ids feedback_column: name of the column containing feedbacks values n_iter: number of alternating least squares iterations n_jobs: number of multiprocessing jobs """ check_columns_in_dataframe( interactions, (user_column, item_column, feedback_column) ) check_feedback_column_numeric(interactions, feedback_column) interactions = drop_warn_na( interactions[[user_column, item_column, feedback_column]] ) n_iter = cast_numeric_greater_than_zero(n_iter, "n_iter", int) R = create_user_item_matrix( interactions=interactions, user_column=user_column, item_column=item_column, feedback_column=feedback_column, ) self._user_index, self._item_index = R.index, R.columns pool = mp.Pool(n_jobs) try: self._X, self._Y = self._run_iterations(pool, R, n_iter) finally: pool.close() def predict(self, user: Any, top_n: Optional[int] = None) -> pd.Series: """ Computes recommendations for given `user`. Args: user: target of the recommendations top_n: if not `None`, selects only the best n items Returns: series with predicted score for each item """ row = get_index_position(self._user_index, user) prediction = pd.Series( np.dot(self._Y, self._X[row, :]), index=self._item_index, name=user ) if top_n: return prediction.nlargest(top_n) else: return prediction @property def user_factors(self) -> pd.DataFrame: """ User factors property. Returns: user factors as a dataframe """ return pd.DataFrame(self._X, index=self._user_index) @property def item_factors(self) -> pd.DataFrame: """ Item factors property. Returns: item factors as a dataframe """ return pd.DataFrame(self._Y, index=self._item_index) @property def loss(self) -> List[np.float]: """ Training loss proprety. Returns: training loss by iteration """ return self._loss
jancervenka/acf
acf/tests/test_utils.py
#!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> import unittest import numpy as np import pandas as pd from ..core import utils class CheckColumnsInDataFrameTest(unittest.TestCase): """ Tests `utils.check_columns_in_data_frame`. """ def setUp(self) -> None: """ Sets up the tests. """ self._test_df = pd.DataFrame({"x": [1], "y": [1], "z": [1]}) def test_all_columns_in_dataframe(self) -> None: """ Tests that no exception is raised if a column is present. """ tests = (tuple(), ("x", "y"), ("x",), ("x", "y", "z")) for columns in tests: utils.check_columns_in_dataframe(self._test_df, columns) def test_column_not_in_data_frame(self) -> None: """ Tests that an exception is raised if a column is not present. """ for columns in (("a"), ("x", "a")): with self.assertRaises(ValueError): utils.check_columns_in_dataframe(self._test_df, columns) class CheckFeedbackColumnNumericTest(unittest.TestCase): """ Tests `utils.check_feedback_column_numeric`. """ def setUp(self) -> None: """ Sets up the tests. """ self._test_df = pd.DataFrame({"x": ["a", "b"], "y": [1, 3]}) def test_feedback_column_is_numeric(self) -> None: """ Tests that no exception is raised if the feedback column is numeric. """ utils.check_feedback_column_numeric(self._test_df, "y") def test_feedback_column_not_numeric(self) -> None: """ Tests that a exception is raised if the feedback column is not numeric. """ with self.assertRaises(ValueError): utils.check_feedback_column_numeric(self._test_df, "x") class DropWarnNaTest(unittest.TestCase): """ Tests `utils.test_drop_warn_na`. """ def test_na_present(self) -> None: """ Tests that rows are dropped and warning is raised when NA value is present. """ test_df = pd.DataFrame({"x": [1.0, 2.0, np.nan]}) expected = pd.DataFrame({"x": [1.0, 2.0]}) with self.assertWarns(UserWarning): result = utils.drop_warn_na(test_df) pd.testing.assert_frame_equal(result, expected) @staticmethod def test_na_not_present() -> None: """ Tests that original dataframe is returned when no NA is present. """ test_df = expected = pd.DataFrame({"x": [1, 2]}) expected = pd.DataFrame({"x": [1, 2]}) result = utils.drop_warn_na(test_df) pd.testing.assert_frame_equal(result, expected) class CastNumericGreaterThanZeroTest(unittest.TestCase): """ Tests `utils.test_drop_warn_na`. """ def test_value_converted(self) -> None: """ Tests that compatible `value` are correctly cast to `required_type`. """ tests = ((1, "test", int, 1), (2.3, "test", float, 2.3), (1.2, "test", int, 1)) for value, value_name, required_type, expected in tests: result = utils.cast_numeric_greater_than_zero( value, value_name, required_type ) self.assertEqual(result, expected) def test_value_not_numeric(self) -> None: """ Test that an exception is raised when a non-numeric `value` cannot be cast to `required_type`. """ with self.assertRaises(ValueError): _ = utils.cast_numeric_greater_than_zero("test", "test", float) def test_value_not_greater_than_zero(self) -> None: """ Test that an exception is raised when a numeric `value` is not greater than zero. """ with self.assertRaises(ValueError): _ = utils.cast_numeric_greater_than_zero(-1, "test", float) class GetIndexPositionTest(unittest.TestCase): """ Tests `utils.get_index_position`. """ def test_index_value_not_unique(self) -> None: """ Tests that an exception is raised when `index_value` is not unique in `index`. """ test_index = pd.Index([1, 3, 2, 2]) test_index_value = 2 with self.assertRaises(ValueError): _ = utils.get_index_position(test_index, test_index_value) def test_index_value_not_found(self) -> None: """ Tests that an exception is raised when `index_value` is not present in `index`. """ test_index = pd.Index([1, 3, 2]) test_index_value = 4 with self.assertRaises(ValueError): _ = utils.get_index_position(test_index, test_index_value) def test_index_value_found(self) -> None: """ Tests that correct row positon is returned. """ test_index = pd.Index([1, 3, 2, 4]) test_index_value = 3 result = utils.get_index_position(test_index, test_index_value) expected = 1 self.assertEqual(result, expected) if __name__ == "__main__": unittest.main()
jancervenka/acf
acf/__init__.py
<reponame>jancervenka/acf<filename>acf/__init__.py #!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> from .core.computation import Engine from .core import metrics from .version import __version__ __doc__ = """ Lightweight recommender engine for implicit feedback datasets. The package implements a collaborative filtering algorithm as described in "Collaborative Filtering for Implicit Feedback Datasets" paper by <NAME>, <NAME>, <NAME> (https://doi.org/10.1109/ICDM.2008.22). """
jancervenka/acf
acf/tests/test_metrics.py
#!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> import unittest from unittest import mock import pandas as pd from ..core.metrics import mean_rank class MeanRankTest(unittest.TestCase): """ Tests `metrics.mean_rank`. """ def test_mean_rank_correct_value(self) -> None: """ Tests that correct mean rank value is computed based on `interactions` and `X`, `Y` factors. """ engine = mock.Mock() engine.user_factors = pd.DataFrame( {0: [0.5, 0.1, 0.9], 1: [0.1, 0.2, 0.5]}, index=["u1", "u2", "u3"] ) engine.item_factors = pd.DataFrame( {0: [0.1, 0.2, 0.3, 0.3], 1: [0.5, 0.9, 0.9, 0.7]}, index=["i1", "i2", "i3", "i4"], ) test_interactions = pd.DataFrame( { "user_id": ["u1", "u1", "u1", "u2", "u2", "u3", "u3", "u3", "u3"], "item_id": ["i1", "i2", "i4", "i2", "i3", "i1", "i2", "i3", "i4"], "feedback": [4, 2, 1, 5, 10, 9, 8, 12, 8], } ) # R = # 4 2 0 1 # 0 5 10 0 # 9 8 12 8 # X * Y^T = # 1.00 0.75 0.25 0.50 # 0.11 0.20 0.21 0.17 # ą0.34 0.63 0.72 0.62 # X * Y^T ranks # 1.00 0.75 0.25 0.50 # 1.00 0.50 0.25 0.75 # 1.00 0.50 0.25 0.75 result = mean_rank( interactions=test_interactions, user_column="user_id", item_column="item_id", feedback_column="feedback", engine=engine, ) expected = 33 / 59 self.assertEqual(result, expected) if __name__ == "__main__": unittest.main()
jancervenka/acf
acf/tests/test_preprocessing.py
<gh_stars>0 #!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> import unittest import numpy as np import pandas as pd from ..core.preprocessing import create_user_item_matrix class CreateUserItemMatrixTest(unittest.TestCase): """ Tests `preprocessing.create_user_item_matrix`. """ @staticmethod def test_create_user_item_matrix() -> None: """ Tests that the created user-item matrix is correct and that duplicate user-item pairs are correctly aggregated. """ test_interactions = pd.DataFrame( { "id_item": ["A", "B", "C", "A", "B", "A", "D", "D"], "id_user": [1, 1, 1, 2, 2, 3, 4, 4], "feedback": [6, 2, 1, 5, 4, 2, 1, 2], } ) result = create_user_item_matrix( test_interactions, "id_user", "id_item", "feedback" ) expected_matrix = np.array( [[6, 2, 1, 0], [5, 4, 0, 0], [2, 0, 0, 0], [0, 0, 0, 3]], dtype="float64" ) expected = pd.DataFrame( expected_matrix, index=pd.Index([1, 2, 3, 4], name="id_user"), columns=pd.Index(["A", "B", "C", "D"], name="id_item"), ) pd.testing.assert_frame_equal(result, expected) if __name__ == "__main__": unittest.main()
jancervenka/acf
acf/core/preprocessing.py
<filename>acf/core/preprocessing.py #!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> import pandas as pd def create_user_item_matrix( interactions: pd.DataFrame, user_column: str, item_column: str, feedback_column: str ) -> pd.DataFrame: """ Creates `R` from `interactions` dataframe by pivoting it from long to wide format. Args: interactions: user-item feedback pairs user_column: name of the column containing user ids item_column: name of the column containing item ids feedback_column: name of the column containing feedbacks values Returns: R user-item matrix """ # handles user-item duplicates by suming the feedback return interactions.pivot_table( index=user_column, columns=item_column, aggfunc="sum", values=feedback_column ).fillna(0)
jancervenka/acf
setup.py
#!/usr/bin/env python # -*- coding: utf-8 -*- # 2021, <NAME> import setuptools import pathlib VERSION = "0.2.3" def run_setup(): """ Runs the package setup. """ this_directory = pathlib.Path(__file__).parent setup_params = { "name": "acf", "version": VERSION, "description": "Lightweight recommender engine", "author": "<NAME>", "author_email": "<EMAIL>", "long_description": (this_directory / "README.md").read_text(), "long_description_content_type": "text/markdown", "packages": ["acf", "acf.core", "acf.tests"], "python_requires": ">=3.7", "install_requires": ["pandas>=1.0", "numpy>=1.16"], } setuptools.setup(**setup_params) if __name__ == "__main__": run_setup()
jancervenka/acf
acf/tests/test_computation.py
#!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> import multiprocessing as mp import unittest from typing import Iterable import numpy as np import pandas as pd from ..core.computation import Engine class EngineTest(unittest.TestCase): """ Tests `computation.Engine` class. """ def setUp(self) -> None: """ Sets up the tests. """ self._engine = Engine(random_state=0) def test_initialize_factors(self) -> None: """ Tests that `Engine._initialize_factors` produce factor matrices `X` and `Y` with random values in correct shape. """ test_m, test_n = (15, 25) X_expected, Y_expected = self._engine._initialize_factors(test_m, test_n) for expected, size in ((X_expected, test_m), (Y_expected, test_n)): self.assertTupleEqual((size, self._engine._n_factors), expected.shape) def test_feedback_1d_generator(self) -> None: """ Tests that `Engine._feedback_1d_generator` produces correct row/column arrays from `R` based on the `axis` argument. """ def assert_collection_of_numpy_equal( x: Iterable[np.array], y: Iterable[np.array] ) -> None: """ Asserts that collections `x` and `y` containing numpy arrays are identical. """ x, y = list(x), list(y) assert len(x) == len(x) for array_x, array_y in zip(x, y): np.testing.assert_array_equal(array_x, array_y) test_R = pd.DataFrame({"A": [1, 3], "B": [6, 7]}) tests = ( ([np.array([1, 3]), np.array([6, 7])], 0), ([np.array([1, 6]), np.array([3, 7])], 1), ) for expected, axis in tests: result = list(self._engine._feedback_1d_generator(test_R, axis)) assert_collection_of_numpy_equal(result, expected) def test_compute_factors_1d(self) -> None: """ Tests that `Engine._compute_factors` computes correct factor row for one user/item. """ np.random.seed(0) test_feedback_1d = np.array([5, 4, 0, 0, 0, 4, 0, 0, 0, 0]) test_other_factors = np.random.rand(10, 5) test_other_factors_small = np.dot(test_other_factors.T, test_other_factors) result = self._engine._compute_factors_1d( feedback_1d=test_feedback_1d, other_factors=test_other_factors, other_factors_small=test_other_factors_small, reg_lambda=1, alpha=40, ) expected = np.array([0.2940538, 0.52162197, 0.82403064, -0.2290155, 0.17128187]) np.testing.assert_array_almost_equal(result, expected, decimal=7) def test_compute_factors(self) -> None: """ Tests that `Engine._compute_factors` computes correct factor rows for every user/item. """ np.random.seed(0) R_test = pd.DataFrame( [ [3, 2, 0, 0, 0, 0, 0, 0, 1, 2], [0, 2, 3, 0, 0, 0, 0, 2, 3, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 2], [0, 1, 0, 2, 2, 0, 0, 0, 2, 2], [1, 0, 0, 0, 2, 0, 3, 0, 0, 0], [1, 0, 0, 0, 2, 3, 0, 3, 0, 0], [0, 0, 3, 0, 2, 0, 0, 0, 0, 0], [0, 0, 1, 3, 3, 0, 0, 0, 0, 0], [0, 0, 0, 0, 1, 0, 0, 0, 0, 0], ] ) test_n = R_test.shape[-1] # axis=1 test for user factors, other_factors matrix is for items test_axis = 1 test_other_factors = np.random.rand(test_n, 3) pool = mp.Pool(1) try: result = self._engine._compute_factors( pool=pool, R=R_test, other_factors=test_other_factors, axis=test_axis ) finally: pool.close() # shape (9, 3): 3 factors for 9 users expected = np.array( [ [0.35764017, 1.03013196, 0.27789359], [1.21552231, 1.36885415, -0.71711505], [1.5414733, -0.52604296, -0.47427519], [0.24321584, 1.0214088, 0.40020838], [0.04003445, 1.06238047, 0.06082416], [-0.51951887, 1.15240956, 1.12119426], [-0.79891536, 1.55973535, -0.05766712], [-1.09061448, 1.76283214, -0.04719709], [0.27658886, 0.94346013, -0.90276183], ] ) np.testing.assert_array_almost_equal(result, expected, decimal=7) @staticmethod def test_predict() -> None: """ Tests that `Engine.predict` computes correct recommendation for given `user`. """ engine = Engine() engine._X = np.array([[0.50, 0.90, 0.30], [0.05, 0.94, 0.81]]) engine._Y = np.array( [[0.20, 0.12, 0.80], [0.50, 0.97, 0.03], [0.75, 0.02, 0.15]] ) # user 20 likes factors 1 and 2 # item 4 really belongs to genre 1 and item 3 belongs to genre 2 # these two items will be recommended # # user 20 doesnt care about factor 0 where item 5 belongs # small recommendation for item 5 engine._user_index = pd.Index([10, 20], name="user_id") engine._item_index = pd.Index([3, 4, 5], name="item_id") result = engine.predict(user=20) result_top_2 = engine.predict(user=20, top_n=2) expected = pd.Series( [0.7708, 0.9611, 0.1778], index=engine._item_index, name=20 ) # calling .nlargest on series will sort the values (descending) expected_top_2 = expected[[True, True, False]].sort_values(ascending=False) pd.testing.assert_series_equal(result, expected) pd.testing.assert_series_equal(result_top_2, expected_top_2) def test_get_loss(self) -> None: """ Tests that `Engine._get_loss` computes correct least squares loss. """ R_test = np.array([[1, 0, 0, 4], [2, 1, 0, 0], [0, 0, 0, 1]]) X_test = np.array([[0.3, 0.1], [0.9, 0.4], [0.1, 0.5]]) Y_test = np.array([[0.4, 0.1], [0.9, 0.5], [0.9, 0.9], [0.6, 0.5]]) result = self._engine._get_loss(X_test, Y_test, R_test) expected = 177.73779 self.assertAlmostEqual(result, expected, places=4) if __name__ == "__main__": unittest.main()
jancervenka/acf
acf/core/utils.py
<reponame>jancervenka/acf #!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> import warnings from typing import Tuple, Any import numpy as np import pandas as pd def check_columns_in_dataframe(df: pd.DataFrame, columns: Tuple[str]) -> None: """ Raises an exception if any column name in `columns` is not present in `df` dataframe. Args: df: dataframe to check columns: tuple of columns that required to be present """ for col in columns: if col not in df.columns: raise ValueError(f"Column {col} is not in the dataframe.") def check_feedback_column_numeric( interactions: pd.DataFrame, feedback_column: str ) -> None: """ Raises an exception if `feedback_column` does not contain numeric values. Args: interactions: user-item interaction dataframe feedback_column: name of the column containing feedback values """ if not pd.api.types.is_numeric_dtype(interactions[feedback_column]): raise ValueError(f'Column "{feedback_column}" must be numeric.') def drop_warn_na(df: pd.DataFrame) -> pd.DataFrame: """ Drops any rows with NA values and raises a warning. Args: df: dataframe to check Returns: df without the NA values """ n_0 = len(df) df = df.dropna(how="any") if len(df) < n_0: warnings.warn( f"NA values found in the dataframe," f" {n_0 - len(df)} rows removed." ) return df def cast_numeric_greater_than_zero( value: Any, value_name: str, required_type: type ) -> None: """ Checks that `value` is greater than zero and casts it to `required_type`. Raises an exception `value` not greater than zero. Args: value: numeric value to check value_name: name to be included in the error message required_type: target type of the value Returns: value as required type """ if not isinstance(value, required_type): value = required_type(value) if value <= 0: raise ValueError(f"Value {value_name} must be greater than zero.") return value def get_index_position(index: pd.Index, index_value: Any) -> np.int64: """ Finds position of `index_value` in `index` array. This functions is used to find row number of given `user_id` or `item_id`. Raises an exception if `index` is not unique or the value is not found. Args: index: array of indeces index_value: value to find Returns: position of index_value as an integer """ pos = np.where(index == index_value)[0] if len(pos) > 1: raise ValueError("Index is not unique.") if len(pos) == 0: raise ValueError(f"index_value = {index_value} not found in the index.") return pos[0]
jancervenka/acf
acf/core/metrics.py
#!/usr/bin/env python # -*- coding: utf-8 -*- # 2020, <NAME> import numpy as np import pandas as pd from . import computation from .preprocessing import create_user_item_matrix def mean_rank( interactions: pd.DataFrame, user_column: str, item_column: str, feedback_column: str, engine: computation.Engine, ) -> np.float: """ Computes mean rank metric using real `interactions` as ground truth and predictions produced by `engine`. The metric is defined as ``` mean_rank := Σ_ui r_ui * rank_ui / Σ_ui r_ui ``` where `r_ui` is the feedback value between user `u` and item `i` and `rank_ui` is the recommendation rank of item `i` for user `u`. `rank_ui` is computed by inverse row-wise percentile ranking of values in `X * Y^T` prediction matrix. Value `r_ui = 0` means that item `i` is the first to be recommended for user `u`, `r_uj = 1` is the last to be recommended. The metric is a mean rank value weighted by `R` feedbacks. Args: interactions: user-item feedback pairs user_column: name of the column containing user ids item_column: name of the column containing item ids feedback_column: name of the column containing feedbacks values engine: trained model Returns: computed metric """ # TODO: column/NA checks R = create_user_item_matrix( interactions=interactions, user_column=user_column, item_column=item_column, feedback_column=feedback_column, ) user_ids, item_ids = R.index.tolist(), R.columns.tolist() # prediction = X * Y ^ T # ranks = apply rank row wise on prediction ranks = ( engine.user_factors.loc[user_ids, :] .dot(engine.item_factors.loc[item_ids, :].T) .rank(pct=True, ascending=False, axis=1) ) # multiply ranks with R element-wise, make a sum and divide by R sum return R.mul(ranks).sum().sum() / R.sum().sum()
Andre-Vitorino/House-Predict
api/house/House.py
import pandas as pd import inflection import math import datetime import pickle import numpy as np class House(object): def __init__(self): self.year_built_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/year_built_scaler.pkl', 'rb')) self.ms_sub_class_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/ms_sub_class_scaler.pkl', 'rb')) self.lot_frontage_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/lot_frontage_scaler.pkl', 'rb')) self.lot_area_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/lot_area_scaler.pkl', 'rb')) self.overall_cond_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/overall_cond_scaler.pkl', 'rb')) self.year_remod_add_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/year_remod_add_scaler.pkl', 'rb')) self.mas_vnr_area_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/mas_vnr_area_scaler.pkl', 'rb')) self.bsmt_fin_sf1_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_fin_sf1_scaler.pkl', 'rb')) self.bsmt_fin_sf2_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_fin_sf2_scaler.pkl', 'rb')) self.bsmt_unf_sf_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_unf_sf_scaler.pkl', 'rb')) self.total_bsmt_sf_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/total_bsmt_sf_scaler.pkl', 'rb')) self.st_flr_sf_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/1st_flr_sf_scaler.pkl', 'rb')) self.nd_flr_sf_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/2nd_flr_sf_scaler.pkl', 'rb')) self.low_qual_fin_sf_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/low_qual_fin_sf_scaler.pkl', 'rb')) self.gr_liv_area_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/gr_liv_area_scaler.pkl', 'rb')) self.bsmt_full_bath_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_full_bath_scaler.pkl', 'rb')) self.bsmt_half_bath_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_half_bath_scaler.pkl', 'rb')) self.full_bath_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/full_bath_scaler.pkl', 'rb')) self.half_bath_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/half_bath_scaler.pkl', 'rb')) self.bsmt_full_bath_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_full_bath_scaler.pkl', 'rb')) self.bedroom_abv_gr_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bedroom_abv_gr_scaler.pkl', 'rb')) self.kitchen_qual_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/kitchen_qual_scaler.pkl', 'rb')) self.kitchen_abv_gr_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/kitchen_abv_gr_scaler.pkl', 'rb')) self.tot_rms_abv_grd_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/tot_rms_abv_grd_scaler.pkl', 'rb')) self.fireplaces_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/fireplaces_scaler.pkl', 'rb')) self.garage_yr_blt_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/garage_yr_blt_scaler.pkl', 'rb')) self.garage_cars_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/garage_cars_scaler.pkl', 'rb')) self.garage_area_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/garage_area_scaler.pkl', 'rb')) self.wood_deck_sf_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/wood_deck_sf_scaler.pkl', 'rb')) self.open_porch_sf_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/open_porch_sf_scaler.pkl', 'rb')) self.enclosed_porch_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/enclosed_porch_scaler.pkl', 'rb')) self.ssn_porch_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/3_ssn_porch_scaler.pkl', 'rb')) self.screen_porch_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/screen_porch_scaler.pkl', 'rb')) self.pool_area_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/pool_area_scaler.pkl', 'rb')) self.misc_val_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/misc_val_scaler.pkl', 'rb')) self.ms_zoning_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/ms_zoning_scaler.pkl', 'rb')) self.street_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/street_scaler.pkl', 'rb')) self.lot_shape_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/lot_shape_scaler.pkl', 'rb')) self.land_contour_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/land_contour_scaler.pkl', 'rb')) self.utilities_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/utilities_scaler.pkl', 'rb')) self.lot_config_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/lot_config_scaler.pkl', 'rb')) self.land_slope_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/land_slope_scaler.pkl', 'rb')) self.neighborhood_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/neighborhood_scaler.pkl', 'rb')) self.condition1_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/condition1_scaler.pkl', 'rb')) self.condition2_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/condition2_scaler.pkl', 'rb')) self.bldg_type_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bldg_type_scaler.pkl', 'rb')) self.house_style_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/house_style_scaler.pkl', 'rb')) self.overall_qual_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/overall_qual_scaler.pkl', 'rb')) self.roof_style_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/roof_style_scaler.pkl', 'rb')) self.roof_matl_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/roof_matl_scaler.pkl', 'rb')) self.exterior1st_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/exterior1st_scaler.pkl', 'rb')) self.exterior2nd_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/exterior2nd_scaler.pkl', 'rb')) #self.mas_vnr_type_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/mas_vnr_type_scaler.pkl', 'rb')) self.exter_qual_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/exter_qual_scaler.pkl', 'rb')) self.exter_cond_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/exter_cond_scaler.pkl', 'rb')) self.foundation_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/foundation_scaler.pkl', 'rb')) self.bsmt_qual_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_qual_scaler.pkl', 'rb')) self.bsmt_cond_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_cond_scaler.pkl', 'rb')) self.bsmt_exposure_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_exposure_scaler.pkl', 'rb')) self.bsmt_fin_type1_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_fin_type1_scaler.pkl', 'rb')) self.bsmt_fin_type2_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/bsmt_fin_type2_scaler.pkl', 'rb')) self.heating_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/heating_scaler.pkl', 'rb')) self.heating_qc_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/heating_qc_scaler.pkl', 'rb')) self.central_air_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/central_air_scaler.pkl', 'rb')) self.electrical_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/electrical_scaler.pkl', 'rb')) self.functional_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/functional_scaler.pkl', 'rb')) self.garage_type_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/garage_type_scaler.pkl', 'rb')) self.garage_finish_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/garage_finish_scaler.pkl', 'rb')) self.garage_qual_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/garage_qual_scaler.pkl', 'rb')) self.garage_cond_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/garage_cond_scaler.pkl', 'rb')) self.paved_drive_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/paved_drive_scaler.pkl', 'rb')) self.sale_type_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/sale_type_scaler.pkl', 'rb')) self.sale_condition_scaler = pickle.load(open('/home/andre/repos/House-Prices/parameter/sale_condition_scaler.pkl', 'rb')) def data_cleaning(self, df1): columns_old = ['Id', 'MSSubClass', 'MSZoning', 'LotFrontage', 'LotArea', 'Street','Alley', 'LotShape', 'LandContour', 'Utilities', 'LotConfig', 'LandSlope', 'Neighborhood', 'Condition1', 'Condition2', 'BldgType','HouseStyle', 'OverallQual', 'OverallCond', 'YearBuilt', 'YearRemodAdd', 'RoofStyle', 'RoofMatl', 'Exterior1st', 'Exterior2nd', 'MasVnrType','MasVnrArea', 'ExterQual', 'ExterCond', 'Foundation', 'BsmtQual', 'BsmtCond', 'BsmtExposure', 'BsmtFinType1', 'BsmtFinSF1','BsmtFinType2', 'BsmtFinSF2', 'BsmtUnfSF', 'TotalBsmtSF', 'Heating', 'HeatingQC', 'CentralAir', 'Electrical', '1stFlrSF', '2ndFlrSF','LowQualFinSF', 'GrLivArea', 'BsmtFullBath', 'BsmtHalfBath', 'FullBath', 'HalfBath', 'BedroomAbvGr', 'KitchenAbvGr', 'KitchenQual','TotRmsAbvGrd', 'Functional', 'Fireplaces', 'FireplaceQu', 'GarageType', 'GarageYrBlt', 'GarageFinish', 'GarageCars', 'GarageArea', 'GarageQual','GarageCond', 'PavedDrive', 'WoodDeckSF', 'OpenPorchSF', 'EnclosedPorch', '3SsnPorch', 'ScreenPorch', 'PoolArea', 'PoolQC','Fence', 'MiscFeature', 'MiscVal', 'MoSold', 'YrSold', 'SaleType', 'SaleCondition'] #alterando o formato dos nomes das colunas snakecase = lambda x: inflection.underscore(x) cols_new = list(map(snakecase,columns_old)) df1.columns = cols_new # dropando as colunas que não tem como arrumar df1 = df1.drop(['alley', 'pool_qc', 'fence', 'misc_feature', 'fireplace_qu'], axis=1) #lot_frontage df1['lot_frontage'].fillna(60, inplace=True) #bsmt_qual df1['bsmt_qual'].fillna('TA', inplace=True) #bsmt_cond df1['bsmt_cond'].fillna('TA', inplace=True) #bsmt_exposure df1['bsmt_exposure'].fillna('NO', inplace=True) #bsmt_fin_type1 df1['bsmt_fin_type1'].fillna('Unf', inplace=True) #bsmt_fin_type2 df1['bsmt_fin_type2'].fillna('Unf', inplace=True) #mas_vnr_type #df1['mas_vnr_type'].fillna('None', inplace=True) df1.drop('mas_vnr_type', axis=1) #mas_vnr_area df1['mas_vnr_area'].fillna(0, inplace=True) #electrical df1['electrical'].fillna('SBrkr', inplace=True) #garage_type df1['garage_type'].fillna('Attchd', inplace=True) #garage_yr_blt(CONVERTER PARA YEAR) df1['garage_yr_blt'].fillna(2005, inplace=True) #garage_finish df1['garage_finish'].fillna('Unf', inplace=True) #garage_qual df1['garage_qual'].fillna('TA', inplace=True) #garage_cond df1['garage_cond'].fillna('TA', inplace=True) # Arrumando os formatos das colunas, conforme o necessário #alterando o tipo da colunas abaixo de float para int # Os valores de anos serão, por padrão, INT #garage_yr_blt df1['garage_yr_blt'] = df1['garage_yr_blt'].astype(int) #mas_vnr_area df1['mas_vnr_area'] = df1['mas_vnr_area'].astype(int) #lot_frontage df1['lot_frontage'] = df1['lot_frontage'].astype(int) #alterando as colunas com valores numéricos para o tipo INT(JÁ FEITO) # alterando o tipo da coluna de INT para STR df1['overall_qual'] = df1['overall_qual'].astype(str) return df1 def feature_engineer(self, df2): # Fazendo a criação de novas variáveis que sejam úteis para o modelo df2['sale_date'] = df2['mo_sold'].astype(str) + df2['yr_sold'].astype(str) df2['sale_date'] = pd.to_datetime(df2['sale_date'], format='%m%Y') df2 = df2.drop_duplicates() # Selecionando colunas para excluir df2 = df2.drop(['mo_sold', 'yr_sold'], axis=1) return df2 def data_preparation(self,df1, df5): df5['year_built'] = self.year_built_scaler.fit_transform(df5[['year_built']].values) df5['ms_sub_class'] = self.ms_sub_class_scaler.fit_transform(df5[['ms_sub_class']].values) df5['lot_frontage'] = self.lot_frontage_scaler.fit_transform(df5[['lot_frontage']].values) df5['lot_area'] = self.lot_area_scaler.fit_transform(df5[['lot_area']].values) df5['overall_cond'] = self.overall_cond_scaler.fit_transform(df5[['overall_cond']].values) df5['year_remod_add'] = self.year_remod_add_scaler.fit_transform(df5[['year_remod_add']].values) df5['mas_vnr_area'] = self.mas_vnr_area_scaler.fit_transform(df5[['mas_vnr_area']].values) df5['bsmt_fin_sf1'] = self.bsmt_fin_sf1_scaler.fit_transform(df5[['bsmt_fin_sf1']].values) df5['bsmt_fin_sf2'] = self.bsmt_fin_sf2_scaler.fit_transform(df5[['bsmt_fin_sf2']].values) df5['bsmt_unf_sf'] = self.bsmt_unf_sf_scaler.fit_transform(df5[['bsmt_unf_sf']].values) df5['total_bsmt_sf'] = self.total_bsmt_sf_scaler.fit_transform(df5[['total_bsmt_sf']].values) df5['1st_flr_sf'] = self.st_flr_sf_scaler.fit_transform(df5[['1st_flr_sf']].values) df5['2nd_flr_sf'] = self.nd_flr_sf_scaler.fit_transform(df5[['2nd_flr_sf']].values) df5['low_qual_fin_sf'] = self.low_qual_fin_sf_scaler.fit_transform(df5[['low_qual_fin_sf']].values) df5['gr_liv_area'] = self.gr_liv_area_scaler.fit_transform(df5[['gr_liv_area']].values) df5['bsmt_full_bath'] = self.bsmt_full_bath_scaler.fit_transform(df5[['bsmt_full_bath']].values) df5['bsmt_half_bath'] = self.bsmt_half_bath_scaler.fit_transform(df5[['bsmt_half_bath']].values) df5['full_bath'] = self.full_bath_scaler.fit_transform(df5[['full_bath']].values) df5['half_bath'] = self.half_bath_scaler.fit_transform(df5[['half_bath']].values) df5['bedroom_abv_gr'] = self.bedroom_abv_gr_scaler.fit_transform(df5[['bedroom_abv_gr']].values) df5['kitchen_abv_gr'] = self.kitchen_abv_gr_scaler.fit_transform(df5[['kitchen_abv_gr']].values) df5['tot_rms_abv_grd'] = self.tot_rms_abv_grd_scaler.fit_transform(df5[['tot_rms_abv_grd']].values) df5['fireplaces'] = self.fireplaces_scaler.fit_transform(df5[['fireplaces']].values) df5['garage_yr_blt'] = self.garage_yr_blt_scaler.fit_transform(df5[['garage_yr_blt']].values) df5['garage_cars'] = self.garage_cars_scaler.fit_transform(df5[['garage_cars']].values) df5['garage_area'] = self.garage_area_scaler.fit_transform(df5[['garage_area']].values) df5['wood_deck_sf'] = self.wood_deck_sf_scaler.fit_transform(df5[['wood_deck_sf']].values) df5['open_porch_sf'] = self.open_porch_sf_scaler.fit_transform(df5[['open_porch_sf']].values) df5['enclosed_porch'] = self.enclosed_porch_scaler.fit_transform(df5[['enclosed_porch']].values) df5['3_ssn_porch'] = self.ssn_porch_scaler.fit_transform(df5[['3_ssn_porch']].values) df5['screen_porch'] = self.screen_porch_scaler.fit_transform(df5[['screen_porch']].values) df5['pool_area'] = self.pool_area_scaler.fit_transform(df5[['pool_area']].values) df5['misc_val'] = self.misc_val_scaler.fit_transform(df5[['misc_val']].values) #Label Encoder df5['ms_zoning'] = self.ms_zoning_scaler.fit_transform(df5['ms_zoning']) df5['street'] = self.street_scaler.fit_transform(df5['street']) df5['lot_shape'] = self.lot_shape_scaler.fit_transform(df5['lot_shape']) df5['land_contour'] = self.land_contour_scaler.fit_transform(df5['land_contour']) df5['utilities'] = self.utilities_scaler.fit_transform(df5['utilities']) df5['lot_config'] = self.lot_config_scaler.fit_transform(df5['lot_config']) df5['land_slope'] = self.land_slope_scaler.fit_transform(df5['land_slope']) df5['neighborhood'] = self.neighborhood_scaler.fit_transform(df5['neighborhood']) df5['condition1'] = self.condition1_scaler.fit_transform(df5['condition1']) df5['condition2'] = self.condition2_scaler.fit_transform(df5['condition2']) df5['bldg_type'] = self.bldg_type_scaler.fit_transform(df5['bldg_type']) df5['house_style'] = self.house_style_scaler.fit_transform(df5['house_style']) df5['overall_qual'] = self.overall_qual_scaler.fit_transform(df5['overall_qual']) df5['roof_style'] = self.roof_style_scaler.fit_transform(df5['roof_style']) df5['roof_matl'] = self.roof_matl_scaler.fit_transform(df5['roof_matl']) df5['exterior1st'] = self.exterior1st_scaler.fit_transform(df5['exterior1st']) df5['exterior2nd'] = self.exterior2nd_scaler.fit_transform(df5['exterior2nd']) #df5['mas_vnr_type'] = self.mas_vnr_type_scaler.fit_transform(df5['mas_vnr_type']) df5['exter_qual'] = self.exter_qual_scaler.fit_transform(df5['exter_qual']) df5['exter_cond'] = self.exter_cond_scaler.fit_transform(df5['exter_cond']) df5['foundation'] = self.foundation_scaler.fit_transform(df5['foundation']) df5['bsmt_qual'] = self.bsmt_qual_scaler.fit_transform(df5['bsmt_qual']) df5['bsmt_cond'] = self.bsmt_cond_scaler.fit_transform(df5['bsmt_cond']) df5['bsmt_exposure'] = self.bsmt_exposure_scaler.fit_transform(df5['bsmt_exposure']) df5['bsmt_fin_type1'] = self.bsmt_fin_type1_scaler.fit_transform(df5['bsmt_fin_type1']) df5['bsmt_fin_type2'] = self.bsmt_fin_type2_scaler.fit_transform(df5['bsmt_fin_type2']) df5['heating'] = self.heating_scaler.fit_transform(df5['heating']) df5['heating_qc'] = self.heating_qc_scaler.fit_transform(df5['heating_qc']) df5['central_air'] = self.central_air_scaler.fit_transform(df5['central_air']) df5['electrical'] = self.electrical_scaler.fit_transform(df5['electrical']) df5['kitchen_qual'] = self.kitchen_qual_scaler.fit_transform(df5['kitchen_qual']) df5['functional'] = self.functional_scaler.fit_transform(df5['functional']) df5['garage_type'] = self.garage_type_scaler.fit_transform(df5['garage_type']) df5['garage_finish'] = self.garage_finish_scaler.fit_transform(df5['garage_finish']) df5['garage_qual'] = self.garage_qual_scaler.fit_transform(df5['garage_qual']) df5['garage_cond'] = self.garage_cond_scaler.fit_transform(df5['garage_cond']) df5['paved_drive'] = self.paved_drive_scaler.fit_transform(df5['paved_drive']) df5['sale_type'] = self.sale_type_scaler.fit_transform(df5['sale_type']) df5['sale_condition'] = self.sale_condition_scaler.fit_transform(df5['sale_condition']) #VARIÁVEIS CÍCLICAS df5['sale_date_sin'] = df1['mo_sold'].apply(lambda x: np.sin(x *(2. * np.pi /30))) df5['sale_date_cos'] = df1['mo_sold'].apply(lambda x: np.cos(x *(2. * np.pi /30))) cols_selected = ['ms_zoning','lot_area','neighborhood','overall_qual','overall_cond','year_built','year_remod_add','exter_qual','bsmt_qual','bsmt_fin_sf1', 'bsmt_unf_sf','total_bsmt_sf','1st_flr_sf','2nd_flr_sf','gr_liv_area','full_bath','garage_finish','garage_cars','garage_area'] return df5[cols_selected] def get_prediction(self, model, original_data, test_data): pred = model.predict(test_data) original_data['prediction'] = np.expm1(pred) return original_data.to_json(orient='records', date_format='iso')
Andre-Vitorino/House-Predict
api/handler.py
<gh_stars>0 import pandas as pd from flask import Flask, request, Response from house.House import House import pickle # carregando modelo model = pickle.load(open('/home/andre/repos/House-Prices/model/house_model.pkl', 'rb')) app = Flask(__name__) @app.route('/house/predict', methods=['POST']) def house_predict(): test_json = request.get_json() if test_json: #tem dados if isinstance(test_json, dict): # exemplo único test_raw = pd.Dataframe(test_json, index=[0]) else: test_raw = pd.DataFrame(test_json, columns=test_json[0].keys()) # instanciando classe House pipeline = House() # limpeza dos dados df1 = pipeline.data_cleaning(test_raw) # atributo dos dados df2 = pipeline.feature_engineer(df1) # preparação dos dados df3 = pipeline.data_preparation(df1, df2) # predição dos dados df_response = pipeline.get_prediction(model, test_raw, df3) return df_response else: return Response('{}', status=200, mimetype='application/json') print(test_json) if __name__ == '__main__': app.run('127.0.0.1')
lukemassa/socialmixingmatrix
socialmixing.py
#!/usr/bin/env python3 """ Groups social mixing data into age ranges """ __author__ = "<NAME>" __license__ = "MIT" __email__ = "<EMAIL>" import csv POP_ESTIMATE_COLUMN = "CENSUS2010POP" # Which column to use from the census data class AgeMixing(): """ Container for social mixing data, functions to manipulate it """ def __init__(self): self.social_mixing_matrix = self.load_social_mixing_matrix() self.age_proportions = self.load_age_proportions() self.scaled_matrix = self.get_scaled_matrix(self.social_mixing_matrix) @staticmethod def _age_group_to_tuple(age_group): """ Given an age group like "0-5" return (0, 5) """ lower, upper = age_group.split('-') return (int(lower), int(upper)) def _is_group_in_group(self, group1, group2): """ Is group1 (i.e. "5-10") in group2 (i.e. "0-15") """ total_min, total_max = self._age_group_to_tuple(group2) my_min, my_max = self._age_group_to_tuple(group1) return my_min >= total_min and my_max <= total_max def _proportion_in_group(self, group): """ What proportion of the population is in this group (i.e. "0-5") """ ret = 0 lower, upper = self._age_group_to_tuple(group) for i in range(lower, upper+1): ret+=self.age_proportions[i] return ret def load_social_mixing_matrix(self): """ Load the social mixing matrix into a dict of dicts """ ret = {} table = [] with open('Age-Mixing.csv', newline='') as csvfile: reader = csv.DictReader(csvfile) for row in reader: table.append(row) group_i = row["Age group"] for group_j, value in row.items(): if group_j == "Age group": continue if group_i not in ret: ret[group_i] = {} # group_i comes from the first column of this row, # group_j is the column for each cell ret[group_i][group_j] = float(value) return ret def load_age_proportions(self): """ Load age proportion data into a map from age to what proportion of the pop is that age """ ret = {} with open("US-Age-Sex-Distribution.csv", newline='') as csvfile: reader = csv.DictReader(csvfile) for row in reader: if row["SEX"] == "0": # 0 == both sexes age = int(row["AGE"]) # There's a row with age 999 that has the sum of all the data if age == 999: total = float(row[POP_ESTIMATE_COLUMN]) else: ret[age] = float(row[POP_ESTIMATE_COLUMN]) # Scale by the total for k, v in ret.items(): ret[k] = v / total return ret def get_scaled_matrix(self, social_mixing_matrix): """ Get the matrix, but scaled by proportions of population in both group a and b This amounts to multiplying each cell by the proportion of its row * its column * itself So if mixing from 0-5 to 10-15, had value of 2, and pop is 5% 0-5 and 8% 10-15, then new value is 2*.05*.08 """ ret = {} for group_i, row in social_mixing_matrix.items(): for group_j, value in row.items(): if group_i not in ret: ret[group_i] = {} ret[group_i][group_j] = self._proportion_in_group(group_i) * self._proportion_in_group(group_j) * value return ret def get_cell_for_new_matrix(self, new_group_i, new_group_j): """ Find value of a cell in the new matrix Sums up all the cells covered by the new groupings """ # This is inefficient, just go through all the cells and see if they fit ret = 0 for group_i in self.scaled_matrix: for group_j in self.scaled_matrix: if self._is_group_in_group(group_i, new_group_i) and self._is_group_in_group(group_j, new_group_j): ret+= self.scaled_matrix[group_i][group_j] return ret def get_new_matrix(self, new_groups): """ Take the new groups, and the scaled matrix, and return the new matrix """ ret = {} for group_i in new_groups: for group_j in new_groups: if group_i not in ret: ret[group_i] = {} ret[group_i][group_j] = self.get_cell_for_new_matrix(group_i, group_j) return ret def format_matrix(self, matrix): """ Given a matrix, format it as a string (essentially reverse of load_social_mixing_matrix) """ ret = [] toprow = [] groups = sorted(matrix.keys(), key=lambda x: self._age_group_to_tuple(x)[0]) toprow.append("Age group") for group in groups: toprow.append(group) ret.append(",".join(toprow)) for group_i in groups: row = [] row.append(group_i) for group_j in groups: row.append(str(matrix[group_i][group_j])) ret.append(",".join(row)) return '\n'.join(ret) def main(): a = AgeMixing() matrix = a.get_new_matrix(("0-20", "21-64", "65-100")) print(a.format_matrix(matrix)) if __name__ == "__main__": main()
rezaeiii/VDSR
main.py
<reponame>rezaeiii/VDSR import os, argparse from vdsr import VDSR # Arguments parser = argparse.ArgumentParser(description='TensorFlow implementation of VDSR') # Select GPU parser.add_argument('--gpu-id', type=int, default=0) parser.add_argument('--test', action='store_true', default=False) parser.add_argument('--model-path', type=str, default='checkpoint/VDSR_pretrained') # Training settings parser.add_argument('--epoch', type=int, default=10, help='Number of epoch, default: 60') parser.add_argument('--batch-size', type=int, default=128, help='Mini-batch size, default: 128') parser.add_argument('--learning-rate', type=float, default=1e-4, help='Learning rate, default: 0.0001') # Network setting parser.add_argument('--layer-depth', type=int, default=20, help='Depth of the network, default: 20') # Test setttings parser.add_argument('--scale', type=int, default=3, help='Up-scale factor, only for test. default: 3') parser.add_argument('--print-interval', type=int, default=100) parser.add_argument('--eval-interval', type=int, default=200) # Directory path parser.add_argument('--train-dataset', type=str, default='10') parser.add_argument('--train-dataset-path', type=str, default='Train') parser.add_argument('--valid-dataset', type=str, default='Set5') parser.add_argument('--test-dataset', type=str, default='Set5') parser.add_argument('--test-dataset-path', type=str, default='Test') parser.add_argument('--checkpoint-path', type=str, default='checkpoint/VDSR') parser.add_argument('--result-dir', type=str, default='result') args = parser.parse_args() os.environ['CUDA_VISIBLE_DEVICES'] = str(args.gpu_id) os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' def main(): if not args.test: checkpoint_path = os.path.join(args.checkpoint_path) if not os.path.exists(args.checkpoint_path): os.makedirs(args.checkpoint_path) vdsr = VDSR(args) if args.test: vdsr.test() else: vdsr.train() if __name__ == '__main__': main()
rezaeiii/VDSR
utils.py
<reponame>rezaeiii/VDSR<filename>utils.py import tensorflow as tf import numpy as np import math from PIL import Image from tqdm import tqdm import os import h5py # Read image def imread(fname): return Image.open(fname) # Save image def imsave(image, path, fname): image = image * 255. image = Image.fromarray(image.astype('uint8'), mode='YCbCr') image = image.convert('RGB') return image.save(os.path.join(path, fname)) # Save ground truth image, bicubic interpolated image and srcnn image def save_result(path, gt, bicubic, srcnn, i): imsave(gt, path, str(i)+ '_gt.png') imsave(bicubic, path, str(i) + '_bicubic.png') imsave(srcnn, path, str(i) + '_vdsr.png') # Return true if the h5 sub-images file is exists def exist_train_data(datasetname): return os.path.exists('{}.h5'.format(datasetname)) # Concatenate Y and CrCb channel def concat_ycrcb(y, crcb): return np.concatenate((y, crcb), axis=2) def psnr(gt, sr, shave=0, max_val=1.): diff = gt[shave:-shave, shave:-shave] - sr[shave:-shave, shave:-shave] diff = diff.flatten() rmse = math.sqrt(np.mean(diff ** 2)) return 20 * math.log10(max_val / rmse) def prepare_data(path, scale, is_valid=False): dir_path = os.path.join(os.getcwd(), path) path_gt = os.path.join(dir_path, 'gt') path_lr = os.path.join(dir_path, 'bicubic_{:d}x'.format(scale)) # fnames = ['baby_GT.bmp, bird_GT.bmp, ...'] fnames = os.listdir(path_gt) inputs = [] labels = [] count = 0 for fname in tqdm(fnames, desc='[*] Generating dataset ... '): count += 1 _input = imread(os.path.join(path_lr, fname)) _label = imread(os.path.join(path_gt, fname)) _input = np.array(_input) / 255. _label = np.array(_label) / 255. _label = _label[:_label.shape[0] - np.mod(_label.shape[0], scale), :_label.shape[1] - np.mod(_label.shape[1], scale)] #_label = _label[:_label.shape[0]//scale, :_label.shape[1]//scale] if is_valid: h, w, _ = _input.shape _input_y = _input[:, :, 0] _label_y = _label[:, :, 0] _input_y = _input_y.reshape([1, h, w, 1]) _label_y = _label_y.reshape([1, h, w, 1]) inputs.append(_input_y) labels.append(_label_y) else: inputs.append(_input) labels.append(_label) if is_valid: print('[*] Successfully prepared {:d} valid images !'.format(count)) else: print('[*] Successfully prepared {:d} test images !'.format(count)) return inputs, labels
rezaeiii/VDSR
dataset.py
<filename>dataset.py<gh_stars>1-10 import tensorflow as tf import numpy as np import threading, time import h5py class Dataset(): def __init__(self, data_path, batch_size): self.data_path = data_path # Patch size for training self.input_size = 41 self.label_size = 41 self.batch_size = batch_size self.queue_size = 3000 self.open_h5py_file() self.make_queue() def open_h5py_file(self): self.h5py_file = h5py.File('{}.h5'.format(self.data_path), 'r') self.data_size = self.h5py_file['data'].shape[0] self.data_index = self.data_size // self.batch_size def make_queue(self): self.input_t = tf.placeholder(tf.float32, [None, self.input_size, self.input_size, 1]) self.label_t = tf.placeholder(tf.float32, [None, self.label_size, self.label_size, 1]) queue = tf.RandomShuffleQueue( capacity=self.queue_size, min_after_dequeue=self.batch_size, dtypes=(tf.float32, tf.float32), shapes=((self.input_size, self.input_size, 1), (self.label_size, self.label_size, 1)), name = 'random_shuffle_queue' ) self.enqueue_many = queue.enqueue_many([self.input_t, self.label_t]) self.dequeue_many = queue.dequeue_many(self.batch_size) def start_enqueue_deamon(self, sess): def enqueue_thread(sess): while (True): for (input_t, label_t) in self.generator(): sess.run([self.enqueue_many], feed_dict={ self.input_t: input_t, self.label_t: label_t }) time.sleep(0.0001) thread_number = 1 threads = [] for i in range(thread_number): t = threading.Thread(target=enqueue_thread, args=(sess,), daemon=True) t.start() threads.append(t) return threads def generator(self): for i in range(self.data_index): input_t = self.h5py_file['data'][i * self.batch_size : (i+1) * self.batch_size] label_t = self.h5py_file['label'][i * self.batch_size : (i+1) * self.batch_size] yield (input_t, label_t)
mikaelengstrom/django-react-templatetags
django_react_templatetags/tests/test_manager.py
# -*- coding: utf-8 -*- from django.conf import global_settings from django.template import Context, Template from django.test import TestCase, modify_settings, override_settings import responses from django_react_templatetags.templatetags.react import _get_tag_manager, ReactTagManager class TestReactTagManager(ReactTagManager): def render(self, context): return 'Test' class ReactIncludeComponentTest(TestCase): def setUp(self): self.mocked_context = Context({'REACT_COMPONENTS': []}) def test_tag_manager_not_overridden(self): "Test that the default ReactTagManager is used by default" self.assertEqual(_get_tag_manager(), ReactTagManager) @override_settings( REACT_RENDER_TAG_MANAGER="django_react_templatetags.tests.test_manager.TestReactTagManager" ) @responses.activate def test_tag_manager_overridden(self): "Test that the TestReactTagManager is actually used" self.assertEqual(_get_tag_manager(), TestReactTagManager) responses.add(responses.POST, 'http://react-service.dev', body='Foo Bar', status=200) out = Template( "{% load react %}" "{% react_render component=\"Component\" %}" ).render(self.mocked_context) self.assertEqual('Test', out)
mikaelengstrom/django-react-templatetags
django_react_templatetags/tests/test_ssr.py
<reponame>mikaelengstrom/django-react-templatetags import json try: from unittest import mock except ImportError: import mock from django.urls import reverse from django.template import Context, Template from django.test import TestCase, override_settings import responses from django_react_templatetags.tests.demosite.models import ( Person, MovieWithContext ) @override_settings( REACT_RENDER_HOST='http://react-service.dev/', ) class SSRTemplateTest(TestCase): def setUp(self): self.mocked_context = Context({'REACT_COMPONENTS': []}) @responses.activate def test_verify_404(self): "The SSR rendering falls back to client side rendering if 404" responses.add(responses.POST, 'http://react-service.dev/', json={'error': 'not found'}, status=404) out = Template( "{% load react %}" "{% react_render component=\"Component\" %}" ).render(self.mocked_context) self.assertTrue('<div id="Component_' in out) @responses.activate def test_verify_rendition(self): "The SSR returns inner html" responses.add(responses.POST, 'http://react-service.dev', body='<h1>Title</h1>', status=200) out = Template( "{% load react %}" "{% react_render component=\"Component\" %}" ).render(self.mocked_context) self.assertTrue('<h1>Title</h1>' in out) @responses.activate def test_request_body(self): "The SSR request sends the props in a expected way" responses.add(responses.POST, 'http://react-service.dev', body='<h1>Title</h1>', status=200) person = Person(first_name='Tom', last_name='Waits') self.mocked_context["person"] = person self.mocked_context["component_data"] = {'album': 'Real gone'} Template( "{% load react %}" "{% react_render component=\"Component\" prop_person=person data=component_data %}" # NOQA ).render(self.mocked_context) request_body = { 'componentName': 'Component', 'props': { 'album': 'Real gone', 'person': { 'first_name': 'Tom', 'last_name': 'Waits' } }, 'context': {} } self.assertTrue( json.loads(responses.calls[0].request.body) == request_body ) @responses.activate def test_request_body_context(self): "The SSR request sends the props in a expected way with context" responses.add(responses.POST, 'http://react-service.dev', body='<h1>Title</h1>', status=200) movie = MovieWithContext(title='Office space', year=1991) self.mocked_context["movie"] = movie self.mocked_context["search_term"] = 'Stapler' Template( "{% load react %}" "{% react_render component=\"Component\" prop_movie=movie %}" ).render(self.mocked_context) request_body = { 'componentName': 'Component', 'props': { 'movie': { 'title': 'Office space', 'year': 1991, 'search_term': 'Stapler', } }, 'context': {} } self.assertEquals( json.loads(responses.calls[0].request.body), request_body ) @responses.activate def test_request_body_with_ssr_context(self): "The SSR request appends the 'ssr_context' in an expected way" responses.add(responses.POST, 'http://react-service.dev', body='<h1>Title</h1>', status=200) self.mocked_context["ssr_ctx"] = {"location": "http://localhost"} Template( "{% load react %}" "{% react_render component=\"Component\" ssr_context=ssr_ctx %}" ).render(self.mocked_context) request_body = { 'componentName': 'Component', 'props': {}, 'context': {'location': "http://localhost"} } self.assertEquals( json.loads(responses.calls[0].request.body), request_body ) @responses.activate def test_default_headers(self): "The SSR uses default headers with json as conten type" responses.add(responses.POST, 'http://react-service.dev', body='Foo Bar', status=200) Template( "{% load react %}" "{% react_render component=\"Component\" %}" ).render(self.mocked_context) self.assertTrue(len(responses.calls) == 1) self.assertEquals( responses.calls[0].request.headers['Content-type'], 'application/json' ) self.assertEquals( responses.calls[0].request.headers['Accept'], 'text/plain' ) @override_settings( REACT_RENDER_HEADERS={ 'Authorization': 'Basic 123' } ) @responses.activate def test_custom_headers(self): "The SSR uses custom headers if present" responses.add(responses.POST, 'http://react-service.dev', body='Foo Bar', status=200) Template( "{% load react %}" "{% react_render component=\"Component\" %}" ).render(self.mocked_context) self.assertTrue(len(responses.calls) == 1) self.assertEquals( responses.calls[0].request.headers['Authorization'], 'Basic 123' ) @responses.activate def test_hydrate_if_ssr_present(self): "Makes sure ReactDOM.hydrate is used when SSR is active" responses.add(responses.POST, 'http://react-service.dev', body='Foo Bar', status=200) out = Template( "{% load react %}" "{% react_render component=\"Component\" %}" "{% react_print %}" ).render(self.mocked_context) self.assertTrue('ReactDOM.hydrate(' in out) @override_settings( REACT_RENDER_HOST='http://react-service.dev/', ) class SSRViewTest(TestCase): @mock.patch("django_react_templatetags.ssr.load_or_empty") def test_that_disable_ssr_header_disables_ssr(self, mocked_func): self.client.get( reverse('static_react_view'), HTTP_X_DISABLE_SSR='1', ) self.assertEqual(mocked_func.call_count, 0)
mikaelengstrom/django-react-templatetags
django_react_templatetags/templatetags/react.py
<filename>django_react_templatetags/templatetags/react.py # -*- coding: utf-8 -*- """ This module contains tags for including react components into templates. """ import uuid import importlib import json from django import template from django.conf import settings from django.template import Node from django_react_templatetags import ssr from django_react_templatetags.encoders import json_encoder_cls_factory register = template.Library() CONTEXT_KEY = "REACT_COMPONENTS" CONTEXT_PROCESSOR = 'django_react_templatetags.context_processors.react_context_processor' # NOQA DEFAULT_SSR_HEADERS = { 'Content-type': 'application/json', 'Accept': 'text/plain', } def get_uuid(): return uuid.uuid4().hex def has_ssr(request): if request and request.META.get("HTTP_X_DISABLE_SSR"): return False return hasattr(settings, 'REACT_RENDER_HOST') and \ settings.REACT_RENDER_HOST def get_ssr_headers(): if not hasattr(settings, 'REACT_RENDER_HEADERS'): return DEFAULT_SSR_HEADERS return settings.REACT_RENDER_HEADERS def has_context_processor(): try: status = CONTEXT_PROCESSOR in settings.TEMPLATES[0]['OPTIONS']['context_processors'] # NOQA except Exception as e: # NOQA status = False return status def load_from_ssr(component, ssr_context=None): return ssr.load_or_empty( component, headers=get_ssr_headers(), ssr_context=ssr_context, ) class ReactTagManager(Node): """ Handles the printing of react placeholders and queueing, is invoked by react_render. """ def __init__(self, identifier, component, data=None, css_class=None, props=None, ssr_context=None): component_prefix = "" if hasattr(settings, "REACT_COMPONENT_PREFIX"): component_prefix = settings.REACT_COMPONENT_PREFIX self.identifier = identifier self.component = component self.component_prefix = component_prefix self.data = data self.css_class = css_class self.props = props self.ssr_context = ssr_context def render(self, context): if not has_context_processor(): raise Exception('"react_context_processor must be added to TEMPLATE_CONTEXT_PROCESSORS"') # NOQA qualified_component_name = self.get_qualified_name(context) identifier = self.get_identifier(context, qualified_component_name) component_props = self.get_component_props(context) component = { 'identifier': identifier, 'name': qualified_component_name, 'json': self.props_to_json(component_props, context), } components = context.get(CONTEXT_KEY, []) components.append(component) context[CONTEXT_KEY] = components placeholder_attr = ( ('id', identifier), ('class', self.resolve_template_variable(self.css_class, context)), ) placeholder_attr = [x for x in placeholder_attr if x[1] is not None] component_html = "" if has_ssr(context.get("request", None)): component_html = load_from_ssr(component, ssr_context=self.get_ssr_context(context)) return self.render_placeholder(placeholder_attr, component_html) def get_qualified_name(self, context): component_name = self.resolve_template_variable(self.component, context) return '{}{}'.format(self.component_prefix, component_name) def get_identifier(self, context, qualified_component_name): identifier = self.resolve_template_variable(self.identifier, context) if identifier: return identifier return '{}_{}'.format(qualified_component_name, get_uuid()) def get_component_props(self, context): resolved_data = self.resolve_template_variable_else_none(self.data, context) resolved_data = resolved_data if resolved_data else {} for prop in self.props: data = self.resolve_template_variable_else_none( self.props[prop], context, ) resolved_data[prop] = data return resolved_data def get_ssr_context(self, context): if not self.ssr_context: return {} return self.resolve_template_variable(self.ssr_context, context) @staticmethod def resolve_template_variable(value, context): if isinstance(value, template.Variable): return value.resolve(context) return value @staticmethod def resolve_template_variable_else_none(value, context): try: data = value.resolve(context) except template.VariableDoesNotExist: data = None except AttributeError: data = None return data @staticmethod def props_to_json(resolved_data, context): cls = json_encoder_cls_factory(context) return json.dumps(resolved_data, cls=cls) @staticmethod def render_placeholder(attributes, component_html=''): attr_pairs = map(lambda x: '{}="{}"'.format(*x), attributes) return u'<div {}>{}</div>'.format( " ".join(attr_pairs), component_html, ) @register.tag def react_render(parser, token): """ Renders a react placeholder and adds it to the global render queue. Example: {% react_render component="ListRestaurants" data=restaurants %} """ values = _prepare_args(parser, token) tag_manager = _get_tag_manager() return tag_manager(**values) def _prepare_args(parses, token): """ Normalize token arguments that can be passed along to node renderer """ values = { "identifier": None, "css_class": None, "data": None, "props": {}, } key_mapping = { "id": "identifier", "class": "css_class", "props": "data", } args = token.split_contents() method = args[0] for arg in args[1:]: key, value = arg.split(r'=',) key = key_mapping.get(key, key) is_standalone_prop = key.startswith('prop_') if is_standalone_prop: key = key[5:] value = template.Variable(value) if is_standalone_prop: values['props'][key] = value else: values[key] = value assert "component" in values, "{} is missing component value".format(method) # NOQA return values def _get_tag_manager(): """ Loads a custom React Tag Manager if provided in Django Settings. """ class_path = getattr(settings, 'REACT_RENDER_TAG_MANAGER', '') if not class_path: return ReactTagManager module_path, class_name = class_path.rsplit('.', 1) module = importlib.import_module(module_path) return getattr(module, class_name) @register.inclusion_tag('react_print.html', takes_context=True) def react_print(context): """ Generates ReactDOM.hydate calls based on REACT_COMPONENT queue, this needs to be run after react has been loaded. The queue will be cleared after beeing called. Example: {% react_print %} """ components = context[CONTEXT_KEY] context[CONTEXT_KEY] = [] new_context = context.__copy__() new_context['ssr_available'] = has_ssr( context.get("request", None) ) new_context['components'] = components return new_context
dhruvyad/MultimodalGame
model_symmetric.py
import os import sys import json import time import numpy as np import random import h5py import copy import functools import logging import pickle import torch from torch.autograd import Variable as _Variable import torch.nn.functional as F import torch.nn as nn import torch.optim as optim from torch.nn.parameter import Parameter import torchvision.datasets as dset import torchvision.models as models import torchvision.transforms as transforms from torchvision.utils import save_image from sklearn.metrics import confusion_matrix from agents import Agent from analyze_messages import convert_tensor_to_string from dataset_loader import load_shapeworld_dataset from community_util import sample_agents, build_train_matrix, build_eval_list, get_msg_pairs from misc import build_mask from misc import calculate_average_message from misc import calculate_entropy, calculate_average_entropy, check_entropy from misc import count_distinct_messages from misc import read_log_load from misc import recursively_set_device, torch_save, torch_load, torch_load_communities from misc import FileLogger from misc import VisdomLogger as Logger from misc import xavier_normal import gflags FLAGS = gflags.FLAGS SHAPES = ['circle', 'cross', 'ellipse', 'pentagon', 'rectangle', 'semicircle', 'square', 'triangle'] COLORS = ['blue', 'cyan', 'gray', 'green', 'magenta', 'red', 'yellow'] OOD_EXAMPLES = ['square_red', 'triangle_green', 'circle_blue', 'rectangle_yellow', 'cross_magenta', 'ellipse_cyan'] MAX_EXAMPLES_TO_SAVE = 200 def Variable(*args, **kwargs): var = _Variable(*args, **kwargs) if FLAGS.cuda: var = var.cuda() return var def loglikelihood(log_prob, target): """ Args: log softmax scores (N, C) where N is the batch size and C is the number of classes Output: log likelihood (N) """ return log_prob.gather(1, target) def store_exemplar_batch(data, data_type, logger, flogger): '''Writes MAX_EXAMPLES_TO_SAVE examples in the data to file for debugging data: dictionary containing data and results data = {"masked_im_1": [], "masked_im_2": [], "msg_1": [], "msg_2": [], "p": [], "target": [], "caption": [], "shapes": [], "colors": [], "texts": [], } data_type: flag giving the name of the data to be stored. e.g. "correct", "incorrect" ''' debuglogger.info(f'Num {data_type}: {len(data["masked_im_1"])}') debuglogger.info("Writing exemplar batch to file...") assert len(data["masked_im_1"]) == len(data["masked_im_2"]) == len(data["p"]) == len(data["caption"]) == len(data["shapes"]) == len(data["colors"]) == len(data["texts"]) num_examples = min(len(data["shapes"]), MAX_EXAMPLES_TO_SAVE) path = FLAGS.log_path prefix = FLAGS.experiment_name + "_" + data_type if not os.path.exists(path + "/" + prefix): os.makedirs(path + "/" + prefix) # Save images masked_im_1 = torch.stack(data["masked_im_1"][:num_examples], dim=0) debuglogger.debug(f'Masked im 1: {type(masked_im_1)}') debuglogger.debug(f'Masked im 1: {masked_im_1.size()}') save_image(masked_im_1, path + '/' + prefix + '/im1.png', nrow=16, pad_value=0.5) masked_im_2 = torch.stack(data["masked_im_2"][:num_examples], dim=0) save_image(masked_im_2, path + '/' + prefix + '/im2.png', nrow=16, pad_value=0.5) # Save other relevant info keys = ['p', 'caption', 'shapes', 'colors'] for k in keys: filename = path + '/' + prefix + '/' + k + '.txt' with open(filename, "w") as wf: for i in range(num_examples): wf.write(f'Example {i+1}: {data[k][i]}\n') # Write texts filename = path + '/' + prefix + '/texts.txt' with open(filename, "w") as wf: for i in range(num_examples): s = "" for t in data["texts"][i]: s += t + ", " wf.write(f'Example {i+1}: {s}\n') # Print average and std p np_p = np.array(data["p"]) debuglogger.info(f'p: mean: {np.mean(np_p)} std: {np.std(np_p)}') def calc_message_mean_and_std(m_store): '''Calculate the mean and std deviation of messages per agent per shape, color and shape-color combination''' for k in m_store: msgs = m_store[k]["message"] msgs = torch.stack(msgs, dim=0) debuglogger.debug(f'Key: {k}, Count: {m_store[k]["count"]}, Messages: {msgs.size()}') mean = torch.mean(msgs, dim=0).cpu() std = torch.std(msgs, dim=0).cpu() m_store[k]["mean"] = mean m_store[k]["std"] = std return m_store def log_message_stats(message_stats, logger, flogger, data_type, epoch, step, i_batch): ''' Helper function to write the message stats to file and log them to stdout Logs the mean and std deviation per set of messages per shape, per color and per shape-color for each message set. Additionally logs the distances between the mean message for each agent type per shape, color and shape-color''' debuglogger.info('Logging message stats') shape_colors = [] for s in SHAPES: for c in COLORS: shape_colors.append(str(s) + "_" + str(c)) # log shape stats for s in SHAPES: num = 0 if s in message_stats[0]["shape"]: num = message_stats[0]["shape"][s]["count"] means = [] stds = [] for i, m in enumerate(message_stats): if s in message_stats[i]["shape"]: assert num == message_stats[i]["shape"][s]["count"] m = message_stats[i]["shape"][s]["mean"] st = message_stats[i]["shape"][s]["std"] means.append(m) stds.append(st) dists = [] assert len(means) != 1 for i in range(len(means)): for j in range(i + 1, len(means)): d = torch.dist(means[i], means[j]) dists.append((i, j, d)) if i == len(means) - 2: break logger.log(key=data_type + ": " + s + " message stats: count: ", val=num, step=step) for i in range(len(means)): logger.log(key=data_type + ": " + s + " message stats: Agent " + str(i) + ": mean: ", val=means[i], step=step) logger.log(key=data_type + ": " + s + " message stats: Agent " + str(i) + ": std: ", val=stds[i], step=step) flogger.Log("Epoch: {} Step: {} Batch: {} {} message stats: shape {}: count: {}, agent {}: mean: {}, std: {}".format( epoch, step, i_batch, data_type, s, num, i, means[i], stds[i])) for i in range(len(dists)): logger.log(key=data_type + ": " + s + " message stats: distances: [" + str(dists[i][0]) + ":" + str(dists[i][1]) + "]: ", val=dists[i][2], step=step) flogger.Log("Epoch: {} Step: {} Batch: {} {} message stats: shape {}: dists: {}".format(epoch, step, i_batch, data_type, s, dists)) # log color stats for s in COLORS: num = 0 if s in message_stats[0]["color"]: num = message_stats[0]["color"][s]["count"] means = [] stds = [] for i, m in enumerate(message_stats): if s in message_stats[i]["color"]: assert num == message_stats[i]["color"][s]["count"] m = message_stats[i]["color"][s]["mean"] st = message_stats[i]["color"][s]["std"] means.append(m) stds.append(st) dists = [] assert len(means) != 1 for i in range(len(means)): for j in range(i + 1, len(means)): d = torch.dist(means[i], means[j]) dists.append((i, j, d)) if i == len(means) - 2: break logger.log(key=data_type + ": " + s + " message stats: count: ", val=num, step=step) for i in range(len(means)): logger.log(key=data_type + ": " + s + " message stats: Agent " + str(i) + ": mean: ", val=means[i], step=step) logger.log(key=data_type + ": " + s + " message stats: Agent " + str(i) + ": std: ", val=stds[i], step=step) flogger.Log("Epoch: {} Step: {} Batch: {} {} message stats: color {}: count: {}, agent {}: mean: {}, std: {}".format( epoch, step, i_batch, data_type, s, num, i, means[i], stds[i])) for i in range(len(dists)): logger.log(key=data_type + ": " + s + " message stats: distances: [" + str(dists[i][0]) + ":" + str(dists[i][1]) + "]: ", val=dists[i][2], step=step) flogger.Log("Epoch: {} Step: {} Batch: {} {} message stats: color {}: dists: {}".format(epoch, step, i_batch, data_type, s, dists)) # log shape - color stats for s in shape_colors: num = 0 if s in message_stats[0]["shape_color"]: num = message_stats[0]["shape_color"][s]["count"] means = [] stds = [] for i, m in enumerate(message_stats): if s in message_stats[i]["shape_color"]: assert num == message_stats[i]["shape_color"][s]["count"] m = message_stats[i]["shape_color"][s]["mean"] st = message_stats[i]["shape_color"][s]["std"] means.append(m) stds.append(st) dists = [] assert len(means) != 1 for i in range(len(means)): for j in range(i + 1, len(means)): d = torch.dist(means[i], means[j]) dists.append((i, j, d)) if i == len(means) - 2: break logger.log(key=data_type + ": " + s + " message stats: count: ", val=num, step=step) for i in range(len(means)): logger.log(key=data_type + ": " + s + " message stats: Agent " + str(i) + ": mean: ", val=means[i], step=step) logger.log(key=data_type + ": " + s + " message stats: Agent " + str(i) + ": std: ", val=stds[i], step=step) flogger.Log("Epoch: {} Step: {} Batch: {} {} message stats: shape_color {}: count: {}, agent {}: mean: {}, std: {}".format(epoch, step, i_batch, data_type, s, num, i, means[i], stds[i])) for i in range(len(dists)): logger.log(key=data_type + ": " + s + " message stats: distances: [" + str(dists[i][0]) + ":" + str(dists[i][1]) + "]: ", val=dists[i][2], step=step) flogger.Log("Epoch: {} Step: {} Batch: {} {} message stats: shape_color {}: dists: {}".format(epoch, step, i_batch, data_type, s, dists)) path = FLAGS.log_path + "/" + FLAGS.experiment_name + "_" + data_type + "_message_stats.pkl" pickle.dump(message_stats, open(path, "wb")) debuglogger.info(f'Saved message stats to log file') def run_analyze_messages(data, data_type, logger, flogger, epoch, step, i_batch): '''Calculates the mean and std deviation per set of messages per shape, per color and per shape-color for each message set. Additionally caculates the distances between the mean message for each agent type per shape, color and shape-color data: dictionary containing log of data_type examples data_type: flag explaining the type of data e.g. "correct", "incorrect" Each message list should have the same length and the shape and colors lists Also saves the messages and analysis to file ''' message_stats = [] messages = [data["msg_1"], data["msg_2"]] shapes = data["shapes"] colors = data["colors"] for m_set in messages: assert len(m_set) == len(shapes) assert len(m_set) == len(colors) d = {"shape": {}, "color": {}, "shape_color": {} } message_stats.append(d) debuglogger.info(f'Messages: {len(messages[0])}, {len(messages[0][0])}') for i, m_set in enumerate(messages): s_store = message_stats[i]["shape"] c_store = message_stats[i]["color"] s_c_store = message_stats[i]["shape_color"] # Collect all messages j = 0 for m, s, c in zip(m_set, shapes, colors): if s in s_store: # Potentially multiple exchanges for m_i in m: s_store[s]["count"] += 1 s_store[s]["message"].append(m_i.data) else: s_store[s] = {} s_store[s]["count"] = 1 s_store[s]["message"] = [m[0].data] if len(m) > 1: for m_i in m[1:]: s_store[s]["count"] += 1 s_store[s]["message"].append(m_i.data) if c in c_store: # Potentially multiple exchanges for m_i in m: c_store[c]["count"] += 1 c_store[c]["message"].append(m_i.data) else: c_store[c] = {} c_store[c]["count"] = 1 c_store[c]["message"] = [m[0].data] if len(m) > 1: for m_i in m[1:]: c_store[c]["count"] += 1 c_store[c]["message"].append(m_i.data) s_c = str(s) + "_" + str(c) if s_c in s_c_store: # Potentially multiple exchanges for m_i in m: s_c_store[s_c]["count"] += 1 s_c_store[s_c]["message"].append(m_i.data) else: s_c_store[s_c] = {} s_c_store[s_c]["count"] = 1 s_c_store[s_c]["message"] = [m[0].data] if len(m) > 1: for m_i in m[1:]: s_c_store[s_c]["count"] += 1 s_c_store[s_c]["message"].append(m_i.data) if j == 5: debuglogger.debug(f's_store: {s_store}') debuglogger.debug(f'c_store: {c_store}') debuglogger.debug(f's_c_store: {s_c_store}') j += 1 # Calculate and log mean and std_dev s_store = calc_message_mean_and_std(s_store) c_store = calc_message_mean_and_std(c_store) s_c_store = calc_message_mean_and_std(s_c_store) log_message_stats(message_stats, logger, flogger, data_type, epoch, step, i_batch) def add_data_point(batch, i, data_store, messages_1, messages_2, probs_1, probs_2): '''Adds the relevant data from a batch to a data store to analyze later''' # Storing images creates a huge slowdown and no need to store them # data_store["masked_im_1"].append(batch["masked_im_1"][i]) # data_store["masked_im_2"].append(batch["masked_im_2"][i]) data_store["p"].append(batch["p"][i]) data_store["target"].append(batch["target"][i]) data_store["caption"].append(batch["caption_str"][i]) data_store["shapes"].append(batch["shapes"][i]) data_store["colors"].append(batch["colors"][i]) data_store["texts"].append(batch["texts_str"][i]) # Add messages, probs and entropy from each exchange m_1 = [] # message p_1 = [] # message probability m_1_ent = [] # entropy per message m_1_str = [] # message as a string for exchange, prob in zip(messages_1, probs_1): m = exchange[i].data.cpu() p = prob[i].data.cpu() m_1.append(m) p_1.append(p) m_1_ent.append(calculate_entropy(p)) m_1_str.append(convert_tensor_to_string(m)) data_store["msg_1"].append(m_1) data_store["probs_1"].append(p_1) data_store["msg_1_str"].append(m_1_str) data_store["msg_1_ent"].append(m_1_ent) m_2 = [] # message p_2 = [] # message probability m_2_ent = [] # entropy per message m_2_str = [] # message as a string for exchange, prob in zip(messages_2, probs_2): m = exchange[i].data.cpu() p = prob[i].data.cpu() m_2.append(m) p_2.append(p) m_2_ent.append(calculate_entropy(p)) m_2_str.append(convert_tensor_to_string(m)) data_store["msg_2"].append(m_2) data_store["probs_2"].append(p_2) data_store["msg_2_str"].append(m_2_str) data_store["msg_2_ent"].append(m_2_ent) return data_store def save_messages_and_stats(correct, incorrect, agent_tag): '''Saves all messages and message probs between two agents, along with relevant tags such as the shape, color and caption, and whether the message was correct or incorrect Data is stored as a list of dicts and pickled. Each dict corresponds to one exchange. The dicts have the following keys - msg_1: message(s) sent from agent 1 - msg_2: message(s) sent from agent 2 - probs_1: message 1 probs - probs_2: message 2 probs - caption: the correct caption - shape: the shape in the caption - color: the color in the caption - correct: boolean, whether both agents solved the task after communication ''' message_data = [] num_correct = len(correct["caption"]) num_incorrect = len(incorrect["caption"]) for i in range(num_correct): elem = {} elem["caption"] = correct["caption"][i] elem["msg_1"] = correct["msg_1"][i] elem["msg_2"] = correct["msg_2"][i] elem["probs_1"] = correct["probs_1"][i] elem["probs_2"] = correct["probs_2"][i] elem["shape"] = correct["shapes"][i] elem["color"] = correct["colors"][i] elem["correct"] = True message_data.append(elem) for i in range(num_incorrect): elem = {} elem["caption"] = incorrect["caption"][i] elem["msg_1"] = incorrect["msg_1"][i] elem["msg_2"] = incorrect["msg_2"][i] elem["probs_1"] = incorrect["probs_1"][i] elem["probs_2"] = incorrect["probs_2"][i] elem["shape"] = incorrect["shapes"][i] elem["color"] = incorrect["colors"][i] elem["correct"] = False message_data.append(elem) debuglogger.info(f'Saving messages...') debuglogger.info(f'{num_correct} correct, {num_incorrect} incorrect, {num_correct + num_incorrect} total') path = FLAGS.log_path + "/" + FLAGS.experiment_name + "_" + agent_tag + "_message_stats.pkl" pickle.dump(message_data, open(path, "wb")) debuglogger.info(f'Messages saved') def get_similarity(dataset_path, in_domain_eval, agent1, agent2, a1_group, a2_group, a1_idx, a2_idx, agent_codes_1, agent_codes_2, logger, flogger): '''Computes the similarity between two language groups. Can also be used to compute self similarity Args: agent1: first agent agent2: second agent a1_group: which group agent 1 belongs to (1 or 2) a2_group: which group agent 2 belongs to (1 or 2) a1_idx: index of first agent in the codes of the agent's group a2_idx: index of second agent in the codes of the agent's group agent_codes_1: average codes for each shape and color (from correct, non blank answers) sent by agents in group 1 (one set for each agent) agent_codes_2: average codes for each shape and color (from correct, non blank answers) sent by agents in group 2 (one set for each agent) ''' # Log agent details debuglogger.info(f'Getting similarity for agents: [{a1_idx}/{a2_idx}], group: [{a1_group}/{a2_group}], length codes: [{len(agent_codes_1)}/{len(agent_codes_2)}]') # Keep track of labels true_labels = [] pred_labels_1_nc = [] pred_labels_1_com = [] pred_labels_2_nc = [] pred_labels_2_com = [] # Keep track of number of correct observations total = 0 total_correct_nc = 0 total_correct_com = 0 atleast1_correct_nc = 0 atleast1_correct_com = 0 # Keep track of score when messages are changed test_language_similarity = {"total": 0, "correct": [], "agent_w_changed_msg": [], "shape": [], "color": [], "orig_shape": [], "orig_color": [], "originally_correct": []} detail_language_similarity = [] # Load development images if in_domain_eval: eval_mode = "train" debuglogger.info("Evaluating on in domain validation set") else: eval_mode = FLAGS.dataset_eval_mode debuglogger.info("Evaluating on out of domain validation set") dev_loader = load_shapeworld_dataset(dataset_path, FLAGS.glove_path, eval_mode, FLAGS.dataset_size_dev, FLAGS.dataset_type, FLAGS.dataset_name, FLAGS.batch_size_dev, FLAGS.random_seed, FLAGS.shuffle_dev, FLAGS.img_feat, FLAGS.cuda, truncate_final_batch=False) _batch_counter = 0 for batch in dev_loader: _batch_counter += 1 debuglogger.debug(f'Batch {_batch_counter}') target = batch["target"] im_feats_1 = batch["im_feats_1"] im_feats_2 = batch["im_feats_2"] p = batch["p"] desc = Variable(batch["texts_vec"]) _batch_size = target.size(0) true_labels.append(target.cpu().numpy().reshape(-1)) # GPU support if FLAGS.cuda: im_feats_1 = im_feats_1.cuda() im_feats_2 = im_feats_2.cuda() target = target.cuda() desc = desc.cuda() data = {"im_feats_1": im_feats_1, "im_feats_2": im_feats_2, "p": p} exchange_args = dict() exchange_args["data"] = data exchange_args["target"] = target exchange_args["desc"] = desc exchange_args["train"] = False exchange_args["break_early"] = not FLAGS.fixed_exchange s, message_1, message_2, y_all, r = exchange( agent1, agent2, exchange_args) s_masks_1, s_feats_1, s_probs_1 = s[0] s_masks_2, s_feats_2, s_probs_2 = s[1] feats_1, probs_1 = message_1 feats_2, probs_2 = message_2 y_nc = y_all[0] y = y_all[1] # Mask loss if dynamic exchange length if FLAGS.fixed_exchange: binary_s_masks = None binary_agent1_masks = None binary_agent2_masks = None bas_agent1_masks = None bas_agent2_masks = None y1_masks = None y2_masks = None outp_1 = y[0][-1] outp_2 = y[1][-1] else: # TODO pass # Before communication predictions # Obtain predictions, loss and stats agent 1 (dist_1_nc, maxdist_1_nc, argmax_1_nc, ent_1_nc, nll_loss_1_nc, logs_1_nc) = get_classification_loss_and_stats(y_nc[0], target) # Obtain predictions, loss and stats agent 2 (dist_2_nc, maxdist_2_nc, argmax_2_nc, ent_2_nc, nll_loss_2_nc, logs_2_nc) = get_classification_loss_and_stats(y_nc[1], target) # After communication predictions # Obtain predictions, loss and stats agent 1 (dist_1, maxdist_1, argmax_1, ent_1, nll_loss_1_com, logs_1) = get_classification_loss_and_stats(outp_1, target) # Obtain predictions, loss and stats agent 2 (dist_2, maxdist_2, argmax_2, ent_2, nll_loss_2_com, logs_2) = get_classification_loss_and_stats(outp_2, target) # Store top 1 prediction for confusion matrix pred_labels_1_nc.append(argmax_1_nc.cpu().numpy()) pred_labels_1_com.append(argmax_1.cpu().numpy()) pred_labels_2_nc.append(argmax_2_nc.cpu().numpy()) pred_labels_2_com.append(argmax_2.cpu().numpy()) # Calculate number of correct observations for different types accuracy_1_nc, correct_1_nc, top_1_1_nc = calculate_accuracy( dist_1_nc, target, FLAGS.batch_size_dev, FLAGS.top_k_dev) accuracy_1, correct_1, top_1_1 = calculate_accuracy( dist_1, target, FLAGS.batch_size_dev, FLAGS.top_k_dev) accuracy_2_nc, correct_2_nc, top_1_2_nc = calculate_accuracy( dist_2_nc, target, FLAGS.batch_size_dev, FLAGS.top_k_dev) accuracy_2, correct_2, top_1_2 = calculate_accuracy( dist_2, target, FLAGS.batch_size_dev, FLAGS.top_k_dev) batch_correct_nc = correct_1_nc.float() + correct_2_nc.float() batch_correct_com = correct_1.float() + correct_2.float() batch_correct_top_1_nc = top_1_1_nc.float() + top_1_2_nc.float() batch_correct_top_1_com = top_1_1.float() + top_1_2.float() # Update accuracy counts total += float(_batch_size) total_correct_nc += (batch_correct_nc == 2).sum() total_correct_com += (batch_correct_com == 2).sum() atleast1_correct_nc += (batch_correct_nc > 0).sum() atleast1_correct_com += (batch_correct_com > 0).sum() # Get correct indices correct_indices_nc = (batch_correct_nc == 2) correct_indices_com = (batch_correct_com == 2) # Test compositionality for _ in range(_batch_size): # Construct batch of size 1 data = {"im_feats_1": im_feats_1[_].unsqueeze(0), "im_feats_2": im_feats_2[_].unsqueeze(0), "p": p[_]} exchange_args = dict() exchange_args["data"] = data exchange_args["desc"] = desc[_].unsqueeze(0) exchange_args["train"] = False exchange_args["break_early"] = not FLAGS.fixed_exchange exchange_args["test_language_similarity"] = True # Construct candidate example to change message to debuglogger.info(f'Agent with sight: {batch["non_blank_partition"][_]}') # Only select examples where one agent is blind if batch['non_blank_partition'][_] != 0: change_agent = batch['non_blank_partition'][_] texts = batch["texts_str"][_] s = batch["shapes"][_] c = batch["colors"][_] debuglogger.info(f'i: {_}, caption: {batch["caption_str"][_]}, original target: {target[_]}, Correct? {correct_1[_]}/{correct_2[_]}') debuglogger.debug(f'i: {_}, texts: {texts}') debuglogger.debug(f'i: {_}, texts_shapes: {batch["texts_shapes"][_]}') debuglogger.debug(f'i: {_}, texts_colors: {batch["texts_colors"][_]}') for _t, t in enumerate(texts): # Only select examples that are different to the current target if _t != target[_]: st = batch["texts_shapes"][_][_t] ct = batch["texts_colors"][_][_t] # Only select examples where there is a different of either the shape or the color, not both if (st == s and ct != c) or (st != s and ct == c): exchange_args["target"] = _t exchange_args["change_agent"] = change_agent if ct != c: exchange_args["subtract"] = c exchange_args["add"] = ct else: exchange_args["subtract"] = s exchange_args["add"] = st # Discard examples which involve adding or subtracting "None" (no code for this) if exchange_args["subtract"] is None or exchange_args["add"] is None: debuglogger.info(f'Skipping example due to None add or subtract...') continue debuglogger.info(f'i: {_} t: {_t}, subtracting: {exchange_args["subtract"]}, adding: {exchange_args["add"]}, change agent: {exchange_args["change_agent"]}') # Set up to play the game and store results for all permutations example_stats = {'subtract': {'name': exchange_args["subtract"], 'total': 0, 'correct': 0}, 'add': {'name': exchange_args["add"], 'total': 0, 'correct': 0}, 'own_correct': 0, 'originally_correct': 0, 'correct_permutations': 0, 'total_permutations': 0, 'total_own_codes': 0} # Fix who goes first for all permutations in an example exchange_args["use_given_who_goes_first"] = True if random.random() < 0.5: exchange_args["given_who_goes_first"] = 1 else: exchange_args["given_who_goes_first"] = 2 # Flip the change agent since the agents receiving the image feats will be flipped change_agent = 1 if change_agent == 2 else 2 exchange_args["change_agent"] = change_agent debuglogger.info(f'Given who goes first: {exchange_args["given_who_goes_first"]}, change agent: {exchange_args["change_agent"]}') ''' ===========================================================''' # First play game with the change agent's own codes own_codes_flag = True debuglogger.debug(f'Playing game with own codes: {own_codes_flag}') if change_agent == 1: if a1_group == 1: exchange_args["agent_subtract_dict"] = agent_codes_1[a1_idx] exchange_args["agent_add_dict"] = agent_codes_1[a1_idx] else: exchange_args["agent_subtract_dict"] = agent_codes_2[a1_idx] exchange_args["agent_add_dict"] = agent_codes_2[a1_idx] else: if a2_group == 1: exchange_args["agent_subtract_dict"] = agent_codes_1[a2_idx] exchange_args["agent_add_dict"] = agent_codes_1[a2_idx] else: exchange_args["agent_subtract_dict"] = agent_codes_2[a2_idx] exchange_args["agent_add_dict"] = agent_codes_2[a2_idx] # Play game, corrupting message _s, message_1, message_2, y_all, r = exchange( agent1, agent2, exchange_args) s_masks_1, s_feats_1, s_probs_1 = _s[0] s_masks_2, s_feats_2, s_probs_2 = _s[1] feats_1, probs_1 = message_1 feats_2, probs_2 = message_2 y_nc = y_all[0] y = y_all[1] # We only care about after communication predictions when measuring the peformance score = None new_target = torch.zeros(1).fill_(_t).long() debuglogger.debug(f'Old target: {target[_]}') na, argmax_y1 = torch.max(y[0][-1], 1) na, argmax_y2 = torch.max(y[1][-1], 1) debuglogger.debug(f'y1 logits: {y[0][-1].data}, y2 logits: {y[1][-1].data}') debuglogger.debug(f'y1: {argmax_y1.data[0]}, y2: {argmax_y2.data[0]}, new_target: {new_target[0]}') if FLAGS.cuda: new_target = new_target.cuda() if change_agent == 1: # Calculate score for agent 2 (dist_2_change, na, na, na, na, na) = get_classification_loss_and_stats(y[1][-1], new_target) debuglogger.debug(f'dist: {dist_2_change.data}') na, na, top_1_2_change = calculate_accuracy( dist_2_change, new_target, 1, FLAGS.top_k_dev) score = top_1_2_change else: # Calculate score for agent 1 (dist_1_change, na, na, na, na, na) = get_classification_loss_and_stats(y[0][-1], new_target) debuglogger.debug(f'dist: {dist_1_change.data}') na, na, top_1_1_change = calculate_accuracy( dist_1_change, new_target, 1, FLAGS.top_k_dev) score = top_1_1_change debuglogger.debug(f'i: {_}_{_t}: New caption: {t}, new target: {_t}, change_agent: {change_agent}, correct: {score[0]}, originally correct: {correct_1[_]}/{correct_2[_]}') # Store results test_language_similarity["total"] += 1 test_language_similarity["orig_shape"].append(s) test_language_similarity["orig_color"].append(c) if score[0] == 1: test_language_similarity["correct"].append(1) else: test_language_similarity["correct"].append(0) if change_agent == 2: # The other agent had their message changed test_language_similarity["originally_correct"].append(correct_1[_]) test_language_similarity["agent_w_changed_msg"].append(1) else: test_language_similarity["originally_correct"].append(correct_2[_]) test_language_similarity["agent_w_changed_msg"].append(2) if ct != c: test_language_similarity["shape"].append(None) test_language_similarity["color"].append(ct) else: test_language_similarity["shape"].append(st) test_language_similarity["color"].append(None) # Track detailed results example_stats['total_permutations'] += 1 example_stats['subtract']["total"] += 1 example_stats['add']["total"] += 1 if score[0] == 1: example_stats["subtract"]["correct"] += 1 example_stats["add"]["correct"] += 1 example_stats["correct_permutations"] += 1 if change_agent == 2: example_stats['originally_correct'] = correct_1[_] else: example_stats['originally_correct'] = correct_2[_] if own_codes_flag: example_stats['total_own_codes'] += 1 if score[0] == 1: example_stats['own_correct'] += 1 ''' ===========================================================''' # Now play the game with all permutations of codes for _g1 in range(len(agent_codes_1)): for _g2 in range(len(agent_codes_2)): # Track if agent is playing with its own codes own_codes_flag = False if FLAGS.self_similarity and (change_agent == 1) and (_g1 == _g2 == a1_idx): own_codes_flag = True elif FLAGS.self_similarity and (change_agent == 2) and (_g1 == _g2 == a2_idx): own_codes_flag = True debuglogger.debug(f'Playing game with own codes: {own_codes_flag}') ''' ===========================================================''' # First play game with codes from group 1 subtracting and code from group 2 adding exchange_args["agent_subtract_dict"] = agent_codes_1[_g1] exchange_args["agent_add_dict"] = agent_codes_2[_g2] # Play game, corrupting message _s, message_1, message_2, y_all, r = exchange( agent1, agent2, exchange_args) s_masks_1, s_feats_1, s_probs_1 = _s[0] s_masks_2, s_feats_2, s_probs_2 = _s[1] feats_1, probs_1 = message_1 feats_2, probs_2 = message_2 y_nc = y_all[0] y = y_all[1] # We only care about after communication predictions when measuring the peformance score = None new_target = torch.zeros(1).fill_(_t).long() debuglogger.debug(f'Old target: {target[_]}') na, argmax_y1 = torch.max(y[0][-1], 1) na, argmax_y2 = torch.max(y[1][-1], 1) debuglogger.debug(f'y1 logits: {y[0][-1].data}, y2 logits: {y[1][-1].data}') debuglogger.debug(f'y1: {argmax_y1.data[0]}, y2: {argmax_y2.data[0]}, new_target: {new_target[0]}') if FLAGS.cuda: new_target = new_target.cuda() if change_agent == 1: # Calculate score for agent 2 (dist_2_change, na, na, na, na, na) = get_classification_loss_and_stats(y[1][-1], new_target) debuglogger.debug(f'dist: {dist_2_change.data}') na, na, top_1_2_change = calculate_accuracy( dist_2_change, new_target, 1, FLAGS.top_k_dev) score = top_1_2_change else: # Calculate score for agent 1 (dist_1_change, na, na, na, na, na) = get_classification_loss_and_stats(y[0][-1], new_target) debuglogger.debug(f'dist: {dist_1_change.data}') na, na, top_1_1_change = calculate_accuracy( dist_1_change, new_target, 1, FLAGS.top_k_dev) score = top_1_1_change debuglogger.debug(f'i: {_}_{_t}: New caption: {t}, new target: {_t}, change_agent: {change_agent}, correct: {score[0]}, originally correct: {correct_1[_]}/{correct_2[_]}') # Store results test_language_similarity["total"] += 1 test_language_similarity["orig_shape"].append(s) test_language_similarity["orig_color"].append(c) if score[0] == 1: test_language_similarity["correct"].append(1) else: test_language_similarity["correct"].append(0) if change_agent == 2: # The other agent had their message changed test_language_similarity["originally_correct"].append(correct_1[_]) test_language_similarity["agent_w_changed_msg"].append(1) else: test_language_similarity["originally_correct"].append(correct_2[_]) test_language_similarity["agent_w_changed_msg"].append(2) if ct != c: test_language_similarity["shape"].append(None) test_language_similarity["color"].append(ct) else: test_language_similarity["shape"].append(st) test_language_similarity["color"].append(None) # Track detailed results example_stats['total_permutations'] += 1 example_stats['subtract']["total"] += 1 example_stats['add']["total"] += 1 if score[0] == 1: example_stats["subtract"]["correct"] += 1 example_stats["add"]["correct"] += 1 example_stats["correct_permutations"] += 1 if change_agent == 2: example_stats['originally_correct'] = correct_1[_] else: example_stats['originally_correct'] = correct_2[_] if own_codes_flag: example_stats['total_own_codes'] += 1 if score[0] == 1: example_stats['own_correct'] += 1 ''' ===========================================================''' ''' ===========================================================''' debuglogger.debug(f'Playing game with own codes: {own_codes_flag}') # Now play the game with the switched codes exchange_args["agent_subtract_dict"] = agent_codes_2[_g2] exchange_args["agent_add_dict"] = agent_codes_1[_g1] # Play game, corrupting message _s, message_1, message_2, y_all, r = exchange( agent1, agent2, exchange_args) s_masks_1, s_feats_1, s_probs_1 = _s[0] s_masks_2, s_feats_2, s_probs_2 = _s[1] feats_1, probs_1 = message_1 feats_2, probs_2 = message_2 y_nc = y_all[0] y = y_all[1] # We only care about after communication predictions when measuring the peformance score = None new_target = torch.zeros(1).fill_(_t).long() debuglogger.debug(f'Old target: {target[_]}') na, argmax_y1 = torch.max(y[0][-1], 1) na, argmax_y2 = torch.max(y[1][-1], 1) debuglogger.debug(f'y1 logits: {y[0][-1].data}, y2 logits: {y[1][-1].data}') debuglogger.debug(f'y1: {argmax_y1.data[0]}, y2: {argmax_y2.data[0]}, new_target: {new_target[0]}') if FLAGS.cuda: new_target = new_target.cuda() if change_agent == 1: # Calculate score for agent 2 (dist_2_change, na, na, na, na, na) = get_classification_loss_and_stats(y[1][-1], new_target) debuglogger.debug(f'dist: {dist_2_change.data}') na, na, top_1_2_change = calculate_accuracy( dist_2_change, new_target, 1, FLAGS.top_k_dev) score = top_1_2_change else: # Calculate score for agent 1 (dist_1_change, na, na, na, na, na) = get_classification_loss_and_stats(y[0][-1], new_target) debuglogger.debug(f'dist: {dist_1_change.data}') na, na, top_1_1_change = calculate_accuracy( dist_1_change, new_target, 1, FLAGS.top_k_dev) score = top_1_1_change debuglogger.debug(f'i: {_}_{_t}: New caption: {t}, new target: {_t}, change_agent: {change_agent}, correct: {score[0]}, originally correct: {correct_1[_]}/{correct_2[_]}') # Store results test_language_similarity["total"] += 1 test_language_similarity["orig_shape"].append(s) test_language_similarity["orig_color"].append(c) if score[0] == 1: test_language_similarity["correct"].append(1) else: test_language_similarity["correct"].append(0) if change_agent == 2: # The other agent had their message changed test_language_similarity["originally_correct"].append(correct_1[_]) test_language_similarity["agent_w_changed_msg"].append(1) else: test_language_similarity["originally_correct"].append(correct_2[_]) test_language_similarity["agent_w_changed_msg"].append(2) if ct != c: test_language_similarity["shape"].append(None) test_language_similarity["color"].append(ct) else: test_language_similarity["shape"].append(st) test_language_similarity["color"].append(None) # Track detailed results example_stats['total_permutations'] += 1 example_stats['subtract']["total"] += 1 example_stats['add']["total"] += 1 if score[0] == 1: example_stats["subtract"]["correct"] += 1 example_stats["add"]["correct"] += 1 example_stats["correct_permutations"] += 1 if change_agent == 2: example_stats['originally_correct'] = correct_1[_] else: example_stats['originally_correct'] = correct_2[_] if own_codes_flag: example_stats['total_own_codes'] += 1 if score[0] == 1: example_stats['own_correct'] += 1 ''' ===========================================================''' debuglogger.info(f'Detailed stats: {example_stats}') detail_language_similarity.append(example_stats) debuglogger.info(f'Total msg changed: {test_language_similarity["total"]}, Correct: {sum(test_language_similarity["correct"])}') debuglogger.info(f'Eval total size: {total}') debuglogger.info(f'Eval total correct com: {total_correct_com}') aggregate_stats = {} detail_total = 0 detail_orig_correct = 0 detail_total_permutes_not_filtered = 0 detail_own_total_total = 0 detail_own_total = 0 detail_own_total_filt = 0 detail_own_correct = 0 permutes_total = 0 permutes_correct = 0 for elem in detail_language_similarity: detail_total += 1 detail_total_permutes_not_filtered += elem['total_permutations'] detail_own_total_total += elem['total_own_codes'] if elem["originally_correct"]: detail_orig_correct += 1 detail_own_total += elem['total_own_codes'] if elem['own_correct'] > 0: detail_own_total_filt += elem['total_own_codes'] detail_own_correct += elem['own_correct'] permutes_total += elem['total_permutations'] permutes_correct += elem['correct_permutations'] if elem['subtract']['name'] not in aggregate_stats: aggregate_stats[elem['subtract']['name']] = {'total': 0, 'correct': 0, 'own_correct': 0, 'own_total': 0} aggregate_stats[elem['subtract']['name']]['total'] += elem['subtract']['total'] aggregate_stats[elem['subtract']['name']]['correct'] += elem['subtract']['correct'] aggregate_stats[elem['subtract']['name']]['own_total'] += elem['total_own_codes'] aggregate_stats[elem['subtract']['name']]['own_correct'] += elem['own_correct'] if elem['add']['name'] not in aggregate_stats: aggregate_stats[elem['add']['name']] = {'total': 0, 'correct': 0, 'own_correct': 0, 'own_total': 0} aggregate_stats[elem['add']['name']]['total'] += elem['add']['total'] aggregate_stats[elem['add']['name']]['correct'] += elem['add']['correct'] aggregate_stats[elem['add']['name']]['own_total'] += elem['total_own_codes'] aggregate_stats[elem['add']['name']]['own_correct'] += elem['own_correct'] transform = elem['subtract']['name'] + '_' + elem['add']['name'] if transform not in aggregate_stats: aggregate_stats[transform] = {'total': 0, 'correct': 0, 'own_correct': 0, 'own_total': 0} aggregate_stats[transform]['total'] += elem['subtract']['total'] aggregate_stats[transform]['correct'] += elem['subtract']['correct'] aggregate_stats[transform]['own_total'] += elem['total_own_codes'] aggregate_stats[transform]['own_correct'] += elem['own_correct'] # Log detailed stats results flogger.Log(f'Agents {a1_idx + 1},{a2_idx + 1}: Total examples: {detail_total}, orig correct: {detail_orig_correct}, % correct: {detail_orig_correct / detail_total}') flogger.Log(f'Agents {a1_idx + 1},{a2_idx + 1}: Total permutations: {detail_total_permutes_not_filtered}, own_codes: {detail_own_total_total}, % own {detail_own_total_total / detail_total_permutes_not_filtered}') flogger.Log(f'Agents {a1_idx + 1},{a2_idx + 1}: Total own codes: {detail_own_total}, correct own codes: {detail_own_correct}, % correct {detail_own_correct / detail_own_total}') flogger.Log('Filtering for originally correct examples with at least one permutation with own codes correct...') flogger.Log(f'Agents {a1_idx + 1},{a2_idx + 1}: Permutations total: {permutes_total}, permutations correct: {permutes_correct}, % {permutes_correct / permutes_total}') norm_total = permutes_total - detail_own_total_filt norm_correct = permutes_correct - detail_own_correct flogger.Log(f'Agents {a1_idx + 1},{a2_idx + 1}: Normalized total: {norm_total} normalized correct: {norm_correct}, SIMILARITY: {norm_correct / norm_total}') for key in aggregate_stats: _total = aggregate_stats[key]['total'] _correct = aggregate_stats[key]['correct'] _own_total = aggregate_stats[key]['own_total'] _own_correct = aggregate_stats[key]['own_correct'] _normalized_total = _total - _own_total _normalized_correct = _correct - _own_correct if _total > 0: flogger.Log(f'Agents {a1_idx + 1},{a2_idx + 1}: {key}: total: {_total} correct: {_correct} %: {_correct / _total}') if _normalized_total > 0: flogger.Log(f'Agents {a1_idx + 1},{a2_idx + 1}: {key}: normalized total: {_normalized_total} normalized correct: {_normalized_correct} SIMILARITY: {_normalized_correct / _normalized_total}') else: flogger.Log(f'Agents {a1_idx + 1},{a2_idx + 1}: {key}: normalized total: {_normalized_total} normalized correct: {_normalized_correct} SIMILARITY: {0}') return test_language_similarity def eval_dev(dataset_path, top_k, agent1, agent2, logger, flogger, epoch, step, i_batch, in_domain_eval=True, callback=None, store_examples=False, analyze_messages=True, save_messages=False, agent_tag="_", agent_dicts=None, agent_idxs=None, agent_groups=None): """ Function computing development accuracy and other metrics """ extra = dict() correct_to_analyze = {"masked_im_1": [], "masked_im_2": [], "msg_1": [], "msg_1_ent": [], "msg_1_str": [], "msg_2": [], "msg_2_ent": [], "msg_2_str": [], "probs_1": [], "probs_2": [], "p": [], "target": [], "caption": [], "shapes": [], "colors": [], "texts": [], } incorrect_to_analyze = {"masked_im_1": [], "masked_im_2": [], "msg_1": [], "msg_1_ent": [], "msg_1_str": [], "msg_2": [], "msg_2_ent": [], "msg_2_str": [], "probs_1": [], "probs_2": [], "p": [], "target": [], "caption": [], "shapes": [], "colors": [], "texts": [], } # Keep track of shapes and color accuracy shapes_accuracy = {} for s in SHAPES: shapes_accuracy[s] = {"correct": 0, "total": 0} colors_accuracy = {} for c in COLORS: colors_accuracy[c] = {"correct": 0, "total": 0} shapes_colors_accuracy = {} for c in COLORS: for s in SHAPES: sc = s + '_' + c shapes_colors_accuracy[sc] = {"correct": 0, "total": 0} # Keep track of agent specific performance (given other agent gets it both right) agent1_performance = {"11": 0, # both right "01": 0, # wrong before comms, right after "10": 0, # right before comms, wrong after "00": 0, # both wrong "total": 0} agent2_performance = {"11": 0, # both right "01": 0, # wrong before comms, right after "10": 0, # right before comms, wrong after "00": 0, # both wrong "total": 0} # Keep track of conversation lengths conversation_lengths_1 = [] conversation_lengths_2 = [] # Keep track of message diversity hamming_1 = [] hamming_2 = [] # Keep track of labels true_labels = [] pred_labels_1_nc = [] pred_labels_1_com = [] pred_labels_2_nc = [] pred_labels_2_com = [] # Keep track of number of correct observations total = 0 total_correct_nc = 0 total_correct_com = 0 atleast1_correct_nc = 0 atleast1_correct_com = 0 # Load development images if in_domain_eval: eval_mode = "train" debuglogger.info("Evaluating on in domain validation set") else: eval_mode = FLAGS.dataset_eval_mode debuglogger.info("Evaluating on out of domain validation set") dev_loader = load_shapeworld_dataset(dataset_path, FLAGS.glove_path, eval_mode, FLAGS.dataset_size_dev, FLAGS.dataset_type, FLAGS.dataset_name, FLAGS.batch_size_dev, FLAGS.random_seed, FLAGS.shuffle_dev, FLAGS.img_feat, FLAGS.cuda, truncate_final_batch=False) _batch_counter = 0 for batch in dev_loader: _batch_counter += 1 debuglogger.debug(f'Batch {_batch_counter}') target = batch["target"] im_feats_1 = batch["im_feats_1"] im_feats_2 = batch["im_feats_2"] p = batch["p"] desc = Variable(batch["texts_vec"]) _batch_size = target.size(0) true_labels.append(target.cpu().numpy().reshape(-1)) # GPU support if FLAGS.cuda: im_feats_1 = im_feats_1.cuda() im_feats_2 = im_feats_2.cuda() target = target.cuda() desc = desc.cuda() data = {"im_feats_1": im_feats_1, "im_feats_2": im_feats_2, "p": p} exchange_args = dict() exchange_args["data"] = data exchange_args["target"] = target exchange_args["desc"] = desc exchange_args["train"] = False exchange_args["break_early"] = not FLAGS.fixed_exchange s, message_1, message_2, y_all, r = exchange( agent1, agent2, exchange_args) s_masks_1, s_feats_1, s_probs_1 = s[0] s_masks_2, s_feats_2, s_probs_2 = s[1] feats_1, probs_1 = message_1 feats_2, probs_2 = message_2 y_nc = y_all[0] y = y_all[1] # Mask loss if dynamic exchange length if FLAGS.fixed_exchange: binary_s_masks = None binary_agent1_masks = None binary_agent2_masks = None bas_agent1_masks = None bas_agent2_masks = None y1_masks = None y2_masks = None outp_1 = y[0][-1] outp_2 = y[1][-1] else: # TODO pass # Before communication predictions # Obtain predictions, loss and stats agent 1 (dist_1_nc, maxdist_1_nc, argmax_1_nc, ent_1_nc, nll_loss_1_nc, logs_1_nc) = get_classification_loss_and_stats(y_nc[0], target) # Obtain predictions, loss and stats agent 2 (dist_2_nc, maxdist_2_nc, argmax_2_nc, ent_2_nc, nll_loss_2_nc, logs_2_nc) = get_classification_loss_and_stats(y_nc[1], target) # After communication predictions # Obtain predictions, loss and stats agent 1 (dist_1, maxdist_1, argmax_1, ent_1, nll_loss_1_com, logs_1) = get_classification_loss_and_stats(outp_1, target) # Obtain predictions, loss and stats agent 2 (dist_2, maxdist_2, argmax_2, ent_2, nll_loss_2_com, logs_2) = get_classification_loss_and_stats(outp_2, target) # Store top 1 prediction for confusion matrix pred_labels_1_nc.append(argmax_1_nc.cpu().numpy()) pred_labels_1_com.append(argmax_1.cpu().numpy()) pred_labels_2_nc.append(argmax_2_nc.cpu().numpy()) pred_labels_2_com.append(argmax_2.cpu().numpy()) # Calculate number of correct observations for different types accuracy_1_nc, correct_1_nc, top_1_1_nc = calculate_accuracy( dist_1_nc, target, FLAGS.batch_size_dev, FLAGS.top_k_dev) accuracy_1, correct_1, top_1_1 = calculate_accuracy( dist_1, target, FLAGS.batch_size_dev, FLAGS.top_k_dev) accuracy_2_nc, correct_2_nc, top_1_2_nc = calculate_accuracy( dist_2_nc, target, FLAGS.batch_size_dev, FLAGS.top_k_dev) accuracy_2, correct_2, top_1_2 = calculate_accuracy( dist_2, target, FLAGS.batch_size_dev, FLAGS.top_k_dev) batch_correct_nc = correct_1_nc.float() + correct_2_nc.float() batch_correct_com = correct_1.float() + correct_2.float() batch_correct_top_1_nc = top_1_1_nc.float() + top_1_2_nc.float() batch_correct_top_1_com = top_1_1.float() + top_1_2.float() debuglogger.debug(f'A1 correct com: {correct_1}') debuglogger.debug(f'A2 correct nc: {correct_2}') debuglogger.debug(f'eval batch correct com: {batch_correct_com}') debuglogger.debug(f'eval batch correct nc: {batch_correct_nc}') debuglogger.debug( f'eval batch top 1 correct com: {batch_correct_top_1_com}') debuglogger.debug( f'eval batch top 1 correct nc: {batch_correct_top_1_nc}') # Update accuracy counts total += float(_batch_size) total_correct_nc += (batch_correct_nc == 2).sum() total_correct_com += (batch_correct_com == 2).sum() atleast1_correct_nc += (batch_correct_nc > 0).sum() atleast1_correct_com += (batch_correct_com > 0).sum() debuglogger.debug(f'eval total correct com: {total_correct_com}') debuglogger.debug(f'eval total correct nc: {total_correct_nc}') debuglogger.debug(f'eval atleast1 correct com: {atleast1_correct_com}') debuglogger.debug(f'eval atleast1 correct nc: {atleast1_correct_nc}') debuglogger.debug(f'batch agent 1 nc correct: {correct_1_nc}') debuglogger.debug(f'batch agent 1 com correct: {correct_1}') debuglogger.debug(f'batch agent 2 nc correct: {correct_2_nc}') debuglogger.debug(f'batch agent 2 com correct: {correct_2}') # Track agent specific stats # Agent 1 given Agent 2 both correct a2_idx = (correct_2_nc.float() + correct_2.float()) == 2 a1_00 = (a2_idx & ((correct_1_nc.float() + correct_1.float()) == 0)).sum() a1_10 = (a2_idx & ((correct_1_nc.float() + (1 - correct_1.float()) == 2))).sum() a1_01 = (a2_idx & (((1 - correct_1_nc.float()) + correct_1.float()) == 2)).sum() a1_11 = (a2_idx & ((correct_1_nc.float() + correct_1.float()) == 2)).sum() a1_tot = a2_idx.sum() assert a1_tot == (a1_00 + a1_01 + a1_10 + a1_11) agent1_performance["11"] += a1_11 agent1_performance["01"] += a1_01 agent1_performance["10"] += a1_10 agent1_performance["00"] += a1_00 agent1_performance["total"] += a1_tot # Agent 2 given Agent 1 both correct a1_idx = (correct_1_nc.float() + correct_1.float()) == 2 a2_00 = (a1_idx & ((correct_2_nc.float() + correct_2.float()) == 0)).sum() a2_10 = (a1_idx & ((correct_2_nc.float() + (1 - correct_2.float()) == 2))).sum() a2_01 = (a1_idx & (((1 - correct_2_nc.float()) + correct_2.float()) == 2)).sum() a2_11 = (a1_idx & ((correct_2_nc.float() + correct_2.float()) == 2)).sum() a2_tot = a1_idx.sum() assert a2_tot == (a2_00 + a2_01 + a2_10 + a2_11) agent2_performance["11"] += a2_11 agent2_performance["01"] += a2_01 agent2_performance["10"] += a2_10 agent2_performance["00"] += a2_00 agent2_performance["total"] += a2_tot debuglogger.debug('Agent 1: total {}, 11: {}, 01: {} 00: {}, 10: {}'.format( agent1_performance["total"], agent1_performance["11"], agent1_performance["01"], agent1_performance["00"], agent1_performance["10"])) if agent1_performance["total"] > 0: debuglogger.debug('Agent 1: total {}, 11: {}, 01: {} 00: {}, 10: {}'.format( agent1_performance["total"] / agent1_performance["total"], agent1_performance["11"] / agent1_performance["total"], agent1_performance["01"] / agent1_performance["total"], agent1_performance["00"] / agent1_performance["total"], agent1_performance["10"] / agent1_performance["total"])) debuglogger.debug('Agent 2: total {}, 11: {}, 01: {} 00: {}, 10: {}'.format( agent2_performance["total"], agent2_performance["11"], agent2_performance["01"], agent2_performance["00"], agent2_performance["10"])) if agent2_performance["total"] > 0: debuglogger.debug('Agent 2: total {}, 11: {}, 01: {} 00: {}, 10: {}'.format( agent2_performance["total"] / agent2_performance["total"], agent2_performance["11"] / agent2_performance["total"], agent2_performance["01"] / agent2_performance["total"], agent2_performance["00"] / agent2_performance["total"], agent2_performance["10"] / agent2_performance["total"])) # Gather shape and color stats correct_indices_nc = (batch_correct_nc == 2) correct_indices_com = (batch_correct_com == 2) for _i in range(_batch_size): if batch['shapes'][_i] is not None: shape = batch['shapes'][_i] shapes_accuracy[shape]["total"] += 1 if correct_indices_com[_i]: shapes_accuracy[shape]["correct"] += 1 if batch['colors'][_i] is not None: color = batch['colors'][_i] colors_accuracy[color]["total"] += 1 if correct_indices_com[_i]: colors_accuracy[color]["correct"] += 1 if (batch['colors'][_i] is not None) and (batch['shapes'][_i] is not None): color = batch['colors'][_i] shape = batch['shapes'][_i] sc = shape + '_' + color shapes_colors_accuracy[sc]["total"] += 1 if correct_indices_com[_i]: shapes_colors_accuracy[sc]["correct"] += 1 # Time consuming, so only do this if necessary if store_examples or analyze_messages or save_messages or FLAGS.report_on_complexity: if _batch_counter == 1 and _i == 0: debuglogger.info("Storing message data to analyze later...") if correct_indices_com[_i]: correct_to_analyze = add_data_point(batch, _i, correct_to_analyze, feats_1, feats_2, probs_1, probs_2) else: incorrect_to_analyze = add_data_point(batch, _i, incorrect_to_analyze, feats_1, feats_2, probs_1, probs_2) if _i == 5: debuglogger.debug(f'Message 1: {feats_1}, probs 1: {probs_1}') debuglogger.debug(f'Message 2: {feats_2}, probs 2: {probs_2}') debuglogger.debug(f'Correct dict: {correct_to_analyze}') debuglogger.debug(f'Incorrect dict: {incorrect_to_analyze}') debuglogger.debug(f'shapes dict: {shapes_accuracy}') debuglogger.debug(f'colors dict: {colors_accuracy}') # Keep track of conversation lengths # TODO not relevant yet conversation_lengths_1 += torch.cat(s_feats_1, 1).data.float().sum(1).view(-1).tolist() conversation_lengths_2 += torch.cat(s_feats_2, 1).data.float().sum(1).view(-1).tolist() debuglogger.debug(f'Conversation length 1: {conversation_lengths_1}') debuglogger.debug(f'Conversation length 2: {conversation_lengths_2}') # Keep track of message diversity mean_hamming_1 = 0 mean_hamming_2 = 0 prev_1 = torch.FloatTensor(_batch_size, FLAGS.m_dim).fill_(0) prev_2 = torch.FloatTensor(_batch_size, FLAGS.m_dim).fill_(0) for msg in feats_1: mean_hamming_1 += (msg.data.cpu() - prev_1).abs().sum(1).mean() prev_1 = msg.data.cpu() mean_hamming_1 = mean_hamming_1 / float(len(feats_1)) for msg in feats_2: mean_hamming_2 += (msg.data.cpu() - prev_2).abs().sum(1).mean() prev_2 = msg.data.cpu() mean_hamming_2 = mean_hamming_2 / float(len(feats_2)) hamming_1.append(mean_hamming_1) hamming_2.append(mean_hamming_2) if callback is not None: callback_dict = dict( s_masks_1=s_masks_1, s_feats_1=s_feats_1, s_probs_1=s_probs_1, s_masks_2=s_masks_2, s_feats_2=s_feats_2, s_probs_2=s_probs_2, feats_1=feats_1, feats_2=feats_2, probs_1=probs_1, probs_2=probs_2, y_nc=y_nc, y=y) callback(agent1, agent2, batch, callback_dict) test_language_similarity = {} if agent_dicts is not None: test_language_similarity = get_similarity(dataset_path, in_domain_eval, agent1, agent2, agent_groups[0], agent_groups[1], agent_idxs[0], agent_idxs[1], agent_dicts[0], agent_dicts[1], logger, flogger) debuglogger.info(f'Total msg changed: {test_language_similarity["total"]}, Correct: {sum(test_language_similarity["correct"])}') if store_examples: debuglogger.info(f'Finishing iterating through dev set, storing examples...') store_exemplar_batch(correct_to_analyze, "correct", logger, flogger) store_exemplar_batch(incorrect_to_analyze, "incorrect", logger, flogger) if analyze_messages: debuglogger.info(f'Analyzing messages...') run_analyze_messages(correct_to_analyze, "correct", logger, flogger, epoch, step, i_batch) if save_messages: debuglogger.info(f'Saving messages...') save_messages_and_stats(correct_to_analyze, incorrect_to_analyze, agent_tag) complexity_stats = {} if FLAGS.report_on_complexity: debuglogger.info(f'Reporting on message complexity') # Calc overall stats avg_msg = calculate_average_message([correct_to_analyze['msg_1'], incorrect_to_analyze['msg_1'], correct_to_analyze['msg_2'], incorrect_to_analyze['msg_2']]) avg_ent = calculate_average_entropy([correct_to_analyze['msg_1_ent'], incorrect_to_analyze['msg_1_ent'], correct_to_analyze['msg_2_ent'], incorrect_to_analyze['msg_2_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) print(f'Entropy of average message: {ent_avg_msg}') t, d = count_distinct_messages([correct_to_analyze['msg_1_str'], incorrect_to_analyze['msg_1_str'], correct_to_analyze['msg_2_str'], incorrect_to_analyze['msg_2_str']]) print(f'Total messages: {t}, Num distinct messages: {d}') complexity_stats['total'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Correct answers only avg_msg = calculate_average_message([correct_to_analyze['msg_1'], correct_to_analyze['msg_2']]) avg_ent = calculate_average_entropy([correct_to_analyze['msg_1_ent'], correct_to_analyze['msg_2_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) t, d = count_distinct_messages([correct_to_analyze['msg_1_str'], correct_to_analyze['msg_2_str']]) complexity_stats['correct'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Incorrect answers only avg_msg = calculate_average_message([incorrect_to_analyze['msg_1'], incorrect_to_analyze['msg_2']]) avg_ent = calculate_average_entropy([incorrect_to_analyze['msg_1_ent'], incorrect_to_analyze['msg_2_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) t, d = count_distinct_messages([incorrect_to_analyze['msg_1_str'], incorrect_to_analyze['msg_2_str']]) complexity_stats['incorrect'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Agent 1 total avg_msg = calculate_average_message([correct_to_analyze['msg_1'], incorrect_to_analyze['msg_1']]) avg_ent = calculate_average_entropy([correct_to_analyze['msg_1_ent'], incorrect_to_analyze['msg_1_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) t, d = count_distinct_messages([correct_to_analyze['msg_1_str'], incorrect_to_analyze['msg_1_str']]) complexity_stats['a1_total'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Agent 1 correct avg_msg = calculate_average_message([correct_to_analyze['msg_1']]) avg_ent = calculate_average_entropy([correct_to_analyze['msg_1_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) t, d = count_distinct_messages([correct_to_analyze['msg_1_str']]) complexity_stats['a1_correct'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Agent 1 incorrect avg_msg = calculate_average_message([incorrect_to_analyze['msg_1']]) avg_ent = calculate_average_entropy([incorrect_to_analyze['msg_1_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) t, d = count_distinct_messages([incorrect_to_analyze['msg_1_str']]) complexity_stats['a1_incorrect'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Agent 2 total avg_msg = calculate_average_message([correct_to_analyze['msg_2'], incorrect_to_analyze['msg_2']]) avg_ent = calculate_average_entropy([correct_to_analyze['msg_2_ent'], incorrect_to_analyze['msg_2_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) t, d = count_distinct_messages([correct_to_analyze['msg_2_str'], incorrect_to_analyze['msg_2_str']]) complexity_stats['a2_total'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Agent 2 correct avg_msg = calculate_average_message([correct_to_analyze['msg_2']]) avg_ent = calculate_average_entropy([correct_to_analyze['msg_2_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) t, d = count_distinct_messages([correct_to_analyze['msg_2_str']]) complexity_stats['a2_correct'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Agent 2 incorrect avg_msg = calculate_average_message([incorrect_to_analyze['msg_2']]) avg_ent = calculate_average_entropy([incorrect_to_analyze['msg_2_ent']]) ent_avg_msg = None if avg_msg is None else calculate_entropy(avg_msg) t, d = count_distinct_messages([incorrect_to_analyze['msg_2_str']]) complexity_stats['a2_incorrect'] = {'avg_ent': avg_ent, 'ent_avg_msg': ent_avg_msg, 'num_msg': t, 'distinct_msg': d} # Print confusion matrix true_labels = np.concatenate(true_labels).reshape(-1) pred_labels_1_nc = np.concatenate(pred_labels_1_nc).reshape(-1) pred_labels_1_com = np.concatenate(pred_labels_1_com).reshape(-1) pred_labels_2_nc = np.concatenate(pred_labels_2_nc).reshape(-1) pred_labels_2_com = np.concatenate(pred_labels_2_com).reshape(-1) np.savetxt(FLAGS.conf_mat + "_1_nc", confusion_matrix( true_labels, pred_labels_1_nc), delimiter=',', fmt='%d') np.savetxt(FLAGS.conf_mat + "_1_com", confusion_matrix( true_labels, pred_labels_1_com), delimiter=',', fmt='%d') np.savetxt(FLAGS.conf_mat + "_2_nc", confusion_matrix( true_labels, pred_labels_2_nc), delimiter=',', fmt='%d') np.savetxt(FLAGS.conf_mat + "_2_com", confusion_matrix( true_labels, pred_labels_2_com), delimiter=',', fmt='%d') # Compute statistics conversation_lengths_1 = np.array(conversation_lengths_1) conversation_lengths_2 = np.array(conversation_lengths_2) hamming_1 = np.array(hamming_1) hamming_2 = np.array(hamming_2) extra['conversation_lengths_1_mean'] = conversation_lengths_1.mean() extra['conversation_lengths_1_std'] = conversation_lengths_1.std() extra['conversation_lengths_2_mean'] = conversation_lengths_2.mean() extra['conversation_lengths_2_std'] = conversation_lengths_2.std() extra['hamming_1_mean'] = hamming_1.mean() extra['hamming_2_mean'] = hamming_2.mean() extra['shapes_accuracy'] = shapes_accuracy extra['colors_accuracy'] = colors_accuracy extra['shapes_colors_accuracy'] = shapes_colors_accuracy extra['agent1_performance'] = agent1_performance extra['agent2_performance'] = agent2_performance extra['test_language_similarity'] = test_language_similarity extra['complexity_stats'] = complexity_stats debuglogger.info(f'Eval total size: {total}') debuglogger.info(f'Eval total correct com: {total_correct_com}') total_accuracy_nc = total_correct_nc / total total_accuracy_com = total_correct_com / total atleast1_accuracy_nc = atleast1_correct_nc / total atleast1_accuracy_com = atleast1_correct_com / total # Return accuracy return total_accuracy_nc, total_accuracy_com, atleast1_accuracy_nc, atleast1_accuracy_com, extra def get_and_log_dev_performance(agent1, agent2, dataset_path, in_domain_eval, dev_accuracy_log, logger, flogger, domain, epoch, step, i_batch, store_examples, analyze_messages, save_messages, agent_tag, agent_dicts=None, agent_idxs=None, agent_groups=None): '''Logs performance on the dev set''' total_accuracy_nc, total_accuracy_com, atleast1_accuracy_nc, atleast1_accuracy_com, extra = eval_dev( dataset_path, FLAGS.top_k_dev, agent1, agent2, logger, flogger, epoch, step, i_batch, in_domain_eval=in_domain_eval, callback=None, store_examples=store_examples, analyze_messages=analyze_messages, save_messages=save_messages, agent_tag=agent_tag, agent_dicts=agent_dicts, agent_idxs=agent_idxs, agent_groups=agent_groups) dev_accuracy_log['total_acc_both_nc'].append(total_accuracy_nc) dev_accuracy_log['total_acc_both_com'].append(total_accuracy_com) dev_accuracy_log['total_acc_atl1_nc'].append(atleast1_accuracy_nc) dev_accuracy_log['total_acc_atl1_com'].append(atleast1_accuracy_com) logger.log(key=domain + " Development Accuracy, both right, no comms", val=dev_accuracy_log['total_acc_both_nc'][-1], step=step) logger.log(key=domain + "Development Accuracy, both right, after comms", val=dev_accuracy_log['total_acc_both_com'][-1], step=step) logger.log(key=domain + "Development Accuracy, at least 1 right, no comms", val=dev_accuracy_log['total_acc_atl1_nc'][-1], step=step) logger.log(key=domain + "Development Accuracy, at least 1 right, after comms", val=dev_accuracy_log['total_acc_atl1_com'][-1], step=step) logger.log(key=domain + "Conversation Length A1 (avg)", val=extra['conversation_lengths_1_mean'], step=step) logger.log(key=domain + "Conversation Length A1 (std)", val=extra['conversation_lengths_1_std'], step=step) logger.log(key=domain + "Conversation Length A2 (avg)", val=extra['conversation_lengths_2_mean'], step=step) logger.log(key=domain + "Conversation Length A2 (std)", val=extra['conversation_lengths_2_std'], step=step) logger.log(key=domain + "Hamming 1 (avg)", val=extra['hamming_1_mean'], step=step) logger.log(key=domain + "Hamming 2 (avg)", val=extra['hamming_2_mean'], step=step) if extra['agent1_performance']["total"] > 0: logger.log(key=domain + " Development Accuracy: Agent 1 given Agent 2 both right: 01: ", val=extra['agent1_performance']["01"] / extra['agent1_performance']["total"], step=step) logger.log(key=domain + " Development Accuracy: Agent 1 given Agent 2 both right: 11: ", val=extra['agent1_performance']["11"] / extra['agent1_performance']["total"], step=step) logger.log(key=domain + " Development Accuracy: Agent 1 given Agent 2 both right: 00: ", val=extra['agent1_performance']["00"] / extra['agent1_performance']["total"], step=step) logger.log(key=domain + " Development Accuracy: Agent 1 given Agent 2 both right: 10: ", val=extra['agent1_performance']["10"] / extra['agent1_performance']["total"], step=step) else: logger.log(key=domain + " Development Accuracy: Agent 1 given Agent 2 both right: 0 examples", val=None, step=step) if extra['agent2_performance']["total"] > 0: logger.log(key=domain + " Development Accuracy: Agent 2 given Agent 1 both right: 01: ", val=extra['agent2_performance']["01"] / extra['agent2_performance']["total"], step=step) logger.log(key=domain + " Development Accuracy: Agent 2 given Agent 1 both right: 11: ", val=extra['agent2_performance']["11"] / extra['agent2_performance']["total"], step=step) logger.log(key=domain + " Development Accuracy: Agent 2 given Agent 1 both right: 00: ", val=extra['agent2_performance']["00"] / extra['agent2_performance']["total"], step=step) logger.log(key=domain + " Development Accuracy: Agent 2 given Agent 1 both right: 10: ", val=extra['agent2_performance']["10"] / extra['agent2_performance']["total"], step=step) else: logger.log(key=domain + " Development Accuracy: Agent 1 given Agent 2 both right: 0 examples", val=None, step=step) detail_total = 0 detail_correct = 0 for k in extra['shapes_accuracy']: if extra['shapes_accuracy'][k]['total'] > 0: logger.log(key=domain + " Development Accuracy: " + k + " ", val=extra['shapes_accuracy'][k]['correct'] / extra['shapes_accuracy'][k]['total'], step=step) detail_total += extra['shapes_accuracy'][k]['total'] detail_correct += extra['shapes_accuracy'][k]['correct'] logger.log(key=domain + " Development Accuracy: shapes ", val=detail_correct / detail_total, step=step) detail_total = 0 detail_correct = 0 for k in extra['colors_accuracy']: if extra['colors_accuracy'][k]['total'] > 0: logger.log(key=domain + " Development Accuracy: " + k + " ", val=extra['colors_accuracy'][k]['correct'] / extra['colors_accuracy'][k]['total'], step=step) detail_total += extra['colors_accuracy'][k]['total'] detail_correct += extra['colors_accuracy'][k]['correct'] logger.log(key=domain + " Development Accuracy: colors ", val=detail_correct / detail_total, step=step) detail_total = 0 detail_correct = 0 for k in extra['shapes_colors_accuracy']: if extra['shapes_colors_accuracy'][k]['total'] > 0: logger.log(key=domain + " Development Accuracy: " + k + " ", val=extra['shapes_colors_accuracy'][k]['correct'] / extra['shapes_colors_accuracy'][k]['total'], step=step) detail_total += extra['shapes_colors_accuracy'][k]['total'] detail_correct += extra['shapes_colors_accuracy'][k]['correct'] logger.log(key=domain + " Development Accuracy: shapes_colors ", val=detail_correct / detail_total, step=step) flogger.Log("Epoch: {} Step: {} Batch: {} {} Development Accuracy, both right, no comms: {}".format( epoch, step, i_batch, domain, dev_accuracy_log['total_acc_both_nc'][-1])) flogger.Log("Epoch: {} Step: {} Batch: {} {} Development Accuracy, both right, after comms: {}".format( epoch, step, i_batch, domain, dev_accuracy_log['total_acc_both_com'][-1])) flogger.Log("Epoch: {} Step: {} Batch: {} {} Development Accuracy, at least right, no comms: {}".format( epoch, step, i_batch, domain, dev_accuracy_log['total_acc_atl1_nc'][-1])) flogger.Log("Epoch: {} Step: {} Batch: {} {} Development Accuracy, at least 1 right, after comms: {}".format( epoch, step, i_batch, domain, dev_accuracy_log['total_acc_atl1_com'][-1])) flogger.Log("Epoch: {} Step: {} Batch: {} {} Development Accuracy CHECK, both right, no comms: {}".format( epoch, step, i_batch, domain, total_accuracy_nc)) flogger.Log("Epoch: {} Step: {} Batch: {} {} Development Accuracy CHECK, both right, after comms: {}".format( epoch, step, i_batch, domain, total_accuracy_com)) flogger.Log("Epoch: {} Step: {} Batch: {} {} Development Accuracy CHECK, at least right, no comms: {}".format( epoch, step, i_batch, domain, atleast1_accuracy_nc)) flogger.Log("Epoch: {} Step: {} Batch: {} {} Development Accuracy CHECK, at least 1 right, after comms: {}".format( epoch, step, i_batch, domain, atleast1_accuracy_com)) flogger.Log("Epoch: {} Step: {} Batch: {} {} Conversation Length 1 (avg/std): {}/{}".format( epoch, step, i_batch, domain, extra['conversation_lengths_1_mean'], extra['conversation_lengths_1_std'])) flogger.Log("Epoch: {} Step: {} Batch: {} {} Conversation Length 2 (avg/std): {}/{}".format( epoch, step, i_batch, domain, extra['conversation_lengths_2_mean'], extra['conversation_lengths_2_std'])) flogger.Log("Epoch: {} Step: {} Batch: {} {} Mean Hamming Distance (1/2): {}/{}" .format(epoch, step, i_batch, domain, extra['hamming_1_mean'], extra['hamming_2_mean'])) flogger.Log('Agent 1: total {}, 11: {}, 01: {} 00: {}, 10: {}'.format( extra["agent1_performance"]["total"], extra["agent1_performance"]["11"], extra["agent1_performance"]["01"], extra["agent1_performance"]["00"], extra["agent1_performance"]["10"])) if extra["agent1_performance"]["total"] > 0: flogger.Log('Agent 1: total {}, 11: {}, 01: {} 00: {}, 10: {}'.format( extra["agent1_performance"]["total"] / extra["agent1_performance"]["total"], extra["agent1_performance"]["11"] / extra["agent1_performance"]["total"], extra["agent1_performance"]["01"] / extra["agent1_performance"]["total"], extra["agent1_performance"]["00"] / extra["agent1_performance"]["total"], extra["agent1_performance"]["10"] / extra["agent1_performance"]["total"])) flogger.Log('Agent 2: total {}, 11: {}, 01: {} 00: {}, 10: {}'.format( extra["agent2_performance"]["total"], extra["agent2_performance"]["11"], extra["agent2_performance"]["01"], extra["agent2_performance"]["00"], extra["agent2_performance"]["10"])) if extra["agent2_performance"]["total"] > 0: flogger.Log('Agent 2: total {}, 11: {}, 01: {} 00: {}, 10: {}'.format( extra["agent2_performance"]["total"] / extra["agent2_performance"]["total"], extra["agent2_performance"]["11"] / extra["agent2_performance"]["total"], extra["agent2_performance"]["01"] / extra["agent2_performance"]["total"], extra["agent2_performance"]["00"] / extra["agent2_performance"]["total"], extra["agent2_performance"]["10"] / extra["agent2_performance"]["total"])) detail_total = 0 detail_correct = 0 for k in extra['shapes_accuracy']: if extra['shapes_accuracy'][k]['total'] > 0: flogger.Log('{}: total: {}, correct: {}, accuracy: {}'.format( k, extra['shapes_accuracy'][k]['total'], extra['shapes_accuracy'][k]['correct'], extra['shapes_accuracy'][k]['correct'] / extra['shapes_accuracy'][k]['total'])) detail_total += extra['shapes_accuracy'][k]['total'] detail_correct += extra['shapes_accuracy'][k]['correct'] flogger.Log('{}: {}: total: {}, correct: {}, accuracy: {}'.format( domain, 'TOTAL SHAPES', detail_total, detail_correct, detail_correct / detail_total)) detail_total = 0 detail_correct = 0 for k in extra['colors_accuracy']: if extra['colors_accuracy'][k]['total'] > 0: flogger.Log('{}: total: {}, correct: {}, accuracy: {}'.format( k, extra['colors_accuracy'][k]['total'], extra['colors_accuracy'][k]['correct'], extra['colors_accuracy'][k]['correct'] / extra['colors_accuracy'][k]['total'])) detail_total += extra['colors_accuracy'][k]['total'] detail_correct += extra['colors_accuracy'][k]['correct'] flogger.Log('{}: {}: total: {}, correct: {}, accuracy: {}'.format( domain, 'TOTAL COLORS', detail_total, detail_correct, detail_correct / detail_total)) detail_total = 0 detail_correct = 0 ood_total = 0 ood_correct = 0 id_total = 0 id_correct = 0 for k in extra['shapes_colors_accuracy']: if extra['shapes_colors_accuracy'][k]['total'] > 0: flogger.Log('{}: total: {}, correct: {}, accuracy: {}'.format( k, extra['shapes_colors_accuracy'][k]['total'], extra['shapes_colors_accuracy'][k]['correct'], extra['shapes_colors_accuracy'][k]['correct'] / extra['shapes_colors_accuracy'][k]['total'])) detail_total += extra['shapes_colors_accuracy'][k]['total'] detail_correct += extra['shapes_colors_accuracy'][k]['correct'] if k in OOD_EXAMPLES: ood_total += extra['shapes_colors_accuracy'][k]['total'] ood_correct += extra['shapes_colors_accuracy'][k]['correct'] else: id_total += extra['shapes_colors_accuracy'][k]['total'] id_correct += extra['shapes_colors_accuracy'][k]['correct'] flogger.Log('{}: {}: total: {}, correct: {}, accuracy: {}'.format( domain, 'TOTAL SHAPES_COLORS', detail_total, detail_correct, detail_correct / detail_total)) if id_total > 0: flogger.Log('{}: {}: total: {}, correct: {}, accuracy: {}'.format( domain, 'TOTAL SHAPES_COLORS in domain', id_total, id_correct, id_correct / id_total)) if ood_total > 0: flogger.Log('{}: {}: total: {}, correct: {}, accuracy: {}'.format( domain, 'TOTAL SHAPES_COLORS out of domain', ood_total, ood_correct, ood_correct / ood_total)) if FLAGS.report_on_complexity: flogger.Log(f'{domain}: complexity report: total: avg ent: {extra["complexity_stats"]["total"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["total"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["total"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["total"]["distinct_msg"]}') flogger.Log(f'{domain}: complexity report: correct: avg ent: {extra["complexity_stats"]["correct"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["correct"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["correct"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["correct"]["distinct_msg"]}') flogger.Log(f'{domain}: complexity report: incorrect: avg ent: {extra["complexity_stats"]["incorrect"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["incorrect"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["incorrect"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["incorrect"]["distinct_msg"]}') flogger.Log(f'{domain}: complexity report: a1_total: avg ent: {extra["complexity_stats"]["a1_total"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["a1_total"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["a1_total"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["a1_total"]["distinct_msg"]}') flogger.Log(f'{domain}: complexity report: a1_correct: avg ent: {extra["complexity_stats"]["a1_correct"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["a1_correct"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["a1_correct"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["a1_correct"]["distinct_msg"]}') flogger.Log(f'{domain}: complexity report: a1_incorrect: avg ent: {extra["complexity_stats"]["a1_incorrect"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["a1_incorrect"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["a1_incorrect"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["a1_incorrect"]["distinct_msg"]}') flogger.Log(f'{domain}: complexity report: a2_total: avg ent: {extra["complexity_stats"]["a2_total"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["a2_total"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["a2_total"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["a2_total"]["distinct_msg"]}') flogger.Log(f'{domain}: complexity report: a2_correct: avg ent: {extra["complexity_stats"]["a2_correct"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["a2_correct"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["a2_correct"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["a2_correct"]["distinct_msg"]}') flogger.Log(f'{domain}: complexity report: a2_incorrect: avg ent: {extra["complexity_stats"]["a2_incorrect"]["avg_ent"]}, ent_avg_msg: {extra["complexity_stats"]["a2_incorrect"]["ent_avg_msg"]}, num_msgs: {extra["complexity_stats"]["a2_incorrect"]["num_msg"]}, distinct msg: {extra["complexity_stats"]["a2_incorrect"]["distinct_msg"]}') if agent_dicts is not None: flogger.Log('Test language similarity performance') comp_dict = {} comp_data = extra['test_language_similarity'] correct_orig_correct = 0 correct_orig_incorrect = 0 orig_correct = 0 orig_incorrect = 0 correct = 0 for i in range(len(comp_data['orig_shape'])): if comp_data['shape'][i] is not None: transform = comp_data['orig_shape'][i] + '_' + comp_data['shape'][i] else: transform = comp_data['orig_color'][i] + '_' + comp_data['color'][i] if transform not in comp_dict: comp_dict[transform] = {'total': 0, 'correct': 0, 'p_correct': 0, 'orig_correct': {'total': 0, 'correct': 0, 'p_correct': 0}, 'orig_incorrect': {'total': 0, 'correct': 0, 'p_correct': 0}} comp_dict[transform]['total'] += 1 if comp_data['correct'][i] == 1: comp_dict[transform]['correct'] += 1 correct += 1 if comp_data['originally_correct'][i] == 1: comp_dict[transform]['orig_correct']['total'] += 1 orig_correct += 1 if comp_data['correct'][i]: comp_dict[transform]['orig_correct']['correct'] += 1 correct_orig_correct += 1 else: comp_dict[transform]['orig_incorrect']['total'] += 1 orig_incorrect += 1 if comp_data['correct'][i]: comp_dict[transform]['orig_incorrect']['correct'] += 1 correct_orig_incorrect += 1 flogger.Log(f'Test comp: total: {len(comp_data["orig_shape"])} orig correct: {orig_correct} orig incorrect: {orig_incorrect}') flogger.Log(f'Total correct: {correct}/{correct / len(comp_data["orig_shape"])}, orig_correct_correct {correct_orig_correct}/{correct_orig_correct / orig_correct}, orig_incorrect_correct {correct_orig_incorrect}/{correct_orig_incorrect / orig_incorrect}') for key in comp_dict: comp_dict[key]['p_correct'] = comp_dict[key]['correct'] / comp_dict[key]['total'] if comp_dict[key]['orig_correct']['total'] > 0: comp_dict[key]['orig_correct']['p_correct'] = comp_dict[key]['orig_correct']['correct'] / comp_dict[key]['orig_correct']['total'] if comp_dict[key]['orig_incorrect']['total'] > 0: comp_dict[key]['orig_incorrect']['p_correct'] = comp_dict[key]['orig_incorrect']['correct'] / comp_dict[key]['orig_incorrect']['total'] flogger.Log(f'Transform {key}: total: {comp_dict[key]["total"]}:{comp_dict[key]["correct"]}/{comp_dict[key]["p_correct"]}') flogger.Log(f'Transform {key}: orig_correct: {comp_dict[key]["orig_correct"]["total"]}:{comp_dict[key]["orig_correct"]["correct"]}/{comp_dict[key]["orig_correct"]["p_correct"]}') flogger.Log(f'Transform {key}: orig_incorrect: {comp_dict[key]["orig_incorrect"]["total"]}:{comp_dict[key]["orig_incorrect"]["correct"]}/{comp_dict[key]["orig_incorrect"]["p_correct"]}') return dev_accuracy_log, total_accuracy_com def eval_community(eval_list, models_dict, dev_accuracy_log, logger, flogger, epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=""): eval_type = {1: "Self com, agent connected to multiple pools", 2: "Self com, agent connected to just one pool", 3: "Within pool com, different agents, trained together", 4: "Within pool com, different agents, never trained together", 5: "Cross pool com, different agents, trained together", 6: "Cross pool com, different agents, never trained together"} for num, items in enumerate(eval_list): logger.log(key="Dev accuracy: ", val=eval_type[num + 1], step=step) flogger.Log(f'Dev accuracy: {eval_type[num + 1]}, step: {step}') print(items) if type(items[0]) is list: for nested in items: for elem in nested: if elem is None: pass else: (i, j) = elem agent1 = models_dict["agent" + str(i + 1)] agent2 = models_dict["agent" + str(j + 1)] agent_tag = f'A_{i + 1}_{j + 1}' if i == j: # Create a copy of agents playing with themselves to avoid sharing the hidden state agent2 = Agent(im_feature_type=FLAGS.img_feat, im_feat_dim=FLAGS.img_feat_dim, h_dim=FLAGS.h_dim, m_dim=FLAGS.m_dim, desc_dim=FLAGS.desc_dim, num_classes=FLAGS.num_classes, s_dim=FLAGS.s_dim, use_binary=FLAGS.use_binary, use_attn=FLAGS.visual_attn, attn_dim=FLAGS.attn_dim, use_MLP=FLAGS.use_MLP, cuda=FLAGS.cuda, im_from_scratch=FLAGS.improc_from_scratch, dropout=FLAGS.dropout) agent2.load_state_dict(agent1.state_dict()) if FLAGS.cuda: agent2.cuda() domain = f'In Domain Dev: Agent {i + 1} | Agent {j + 1}, ids [{id(agent1)}]/[{id(agent2)}]: ' _, _ = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_log, logger, flogger, domain, epoch, step, i_batch, store_examples, analyze_messages, save_messages, agent_tag) else: for elem in items: if elem is None: pass else: (i, j) = elem agent1 = models_dict["agent" + str(i + 1)] agent2 = models_dict["agent" + str(j + 1)] agent_tag = f'A_{i + 1}_{j + 1}' if i == j: # Create a copy of agents playing with themselves to avoid sharing the hidden state agent2 = Agent(im_feature_type=FLAGS.img_feat, im_feat_dim=FLAGS.img_feat_dim, h_dim=FLAGS.h_dim, m_dim=FLAGS.m_dim, desc_dim=FLAGS.desc_dim, num_classes=FLAGS.num_classes, s_dim=FLAGS.s_dim, use_binary=FLAGS.use_binary, use_attn=FLAGS.visual_attn, attn_dim=FLAGS.attn_dim, use_MLP=FLAGS.use_MLP, cuda=FLAGS.cuda, im_from_scratch=FLAGS.improc_from_scratch, dropout=FLAGS.dropout) agent2.load_state_dict(agent1.state_dict()) if FLAGS.cuda: agent2.cuda() domain = f'In Domain Dev: Agent {i + 1} | Agent {j + 1}, ids [{id(agent1)}]/[{id(agent2)}]: ' _, _ = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_log, logger, flogger, domain, epoch, step, i_batch, store_examples, analyze_messages, save_messages, agent_tag) def corrupt_message(corrupt_region, agent, binary_message): # Obtain mask mask = Variable(build_mask(corrupt_region, agent.m_dim)) mask_broadcast = mask.view(1, agent_1.m_dim).expand_as(binary_message) # Subtract the mask to change values, but need to get absolute value # to set -1 values to 1 to essentially "flip" all the bits. binary_message = (binary_message - mask_broadcast).abs() return binary_message def exchange(a1, a2, exchange_args): """Run a batched conversation between two agents. There are two parts to an exchange: 1. Each agent receives part of an image, and uses this to select the corresponding text from a selection of texts 2. Agents communicate for a number of steps, then each select the corresponding text again from the same selection of texts Exchange Args: data: Image features - dict containing the image features for agent 1 and agent 2, and the percentage of the image each agent received e.g. { "im_feats_1": im_feats_1, "im_feats_2": im_feats_2, "p": p} target: Class labels. desc: List of description vectors. train: Boolean value indicating training mode (True) or evaluation mode (False). break_early: Boolean value. If True, then terminate batched conversation if both agents are satisfied test_language_similarity: Boolean: whether to test the language similarity using communication vector arithmetric Function Args: a1: agent1 a2: agent2 exchange_args: Other useful arguments. Returns: s: Contains all stop bits [a1, a2] a1 = agent1 (Masks, Values, Probabilties) a2 = agent2 (Masks, Values, Probabilties) message_1: All agent_1 messages. (Values, Probabilities) message_2: All agent_2 messages. (Values, Probabilities) y_all = [y_nc, y] y_nc: (y_1_nc, y_2_nc) y_1_nc: Agent 1 predictions before communication y_2_nc: Agent 2 predictions before communication y = (y_1, y_2) y_1: All predictions that were made by agent 1 after communication y_2: All predictions that were made by agent 2 after communication r = (r_1, r_2) r_1: Estimated rewards of agent_1. r_2: Estimated rewards of agent_2. """ # Randomly select which agent goes first who_goes_first = None use_given_who_goes_first = exchange_args.get("use_given_who_goes_first", False) given_who_goes_first = exchange_args.get("given_who_goes_first", -1) if FLAGS.randomize_comms: if use_given_who_goes_first: if given_who_goes_first == -1: debuglogger.warn(f'No given agent to go first') sys.exit() elif given_who_goes_first == 1: agent1 = a1 agent2 = a2 who_goes_first = 1 debuglogger.debug(f'Agent 1 communicates first') else: agent1 = a2 agent2 = a1 who_goes_first = 2 debuglogger.debug(f'Agent 2 communicates first') else: if random.random() < 0.5: agent1 = a1 agent2 = a2 who_goes_first = 1 debuglogger.debug(f'Agent 1 communicates first') else: agent1 = a2 agent2 = a1 who_goes_first = 2 debuglogger.debug(f'Agent 2 communicates first') else: agent1 = a1 agent2 = a2 who_goes_first = 1 debuglogger.debug(f'Agent 1 communicates first') data = exchange_args["data"] # TODO extend implementation to include data context data_context = None target = exchange_args["target"] desc = exchange_args["desc"] train = exchange_args["train"] break_early = exchange_args.get("break_early", False) corrupt = exchange_args.get("corrupt", False) corrupt_region = exchange_args.get("corrupt_region", None) test_language_similarity = exchange_args.get("test_language_similarity", False) batch_size = data["im_feats_1"].size(0) # Pad with one column of ones. stop_mask_1 = [Variable(torch.ones(batch_size, 1).byte())] stop_feat_1 = [] stop_prob_1 = [] stop_mask_2 = [Variable(torch.ones(batch_size, 1).byte())] stop_feat_2 = [] stop_prob_2 = [] feats_1 = [] probs_1 = [] feats_2 = [] probs_2 = [] y_1_nc = None y_2_nc = None y_1 = [] y_2 = [] r_1 = [] r_2 = [] # First message (default is 0) m_binary = Variable(torch.FloatTensor(batch_size, agent1.m_dim).fill_( FLAGS.first_msg), volatile=not train) if FLAGS.cuda: m_binary = m_binary.cuda() if train: agent1.train() agent2.train() else: agent1.eval() agent2.eval() agent1.reset_state() agent2.reset_state() # The message is ignored initially use_message = False # Run data through both agents if data_context is not None: # No data context at the moment - # TODO debuglogger.warning(f'Data context not supported currently') sys.exit() else: # debuglogger.info(f'Inside exchange: Train status: {train}, Message: {m_binary}') s_1e, m_1e, y_1e, r_1e = agent1( data['im_feats_1'], m_binary, 0, desc, use_message, batch_size, train) s_2e, m_2e, y_2e, r_2e = agent2( data['im_feats_2'], m_binary, 0, desc, use_message, batch_size, train) # Add no message selections to results # Need to be consistent about storing the a1 and a2's even though their roles are randomized during each exchange # agent1 and agent2 is a local name that refers to the order of communication. Storage refers to global labels a1 and a2 if who_goes_first == 1: y_1_nc = y_1e y_2_nc = y_2e else: y_1_nc = y_2e y_2_nc = y_1e for i_exchange in range(FLAGS.max_exchange): debuglogger.debug( f' ================== EXCHANGE {i_exchange} ====================') # The messages are now used use_message = True # Agent 1's message m_1e_binary, m_1e_probs = m_1e # Store last but one communication from agent1 to train with # Last communication is "communication into the void" since the other # agent never gets it. m_binary_1, m_probs_1 = m_1e # Optionally corrupt agent 1's message if corrupt: m_1e_binary = corrupt_message(corrupt_region, agent1, m_1e_binary) # Optionally change agent 1's message if test_language_similarity: if (who_goes_first == 1 and exchange_args["change_agent"] == 1) or (who_goes_first == 2 and exchange_args["change_agent"] == 2): debuglogger.debug(f'Inside exchange: Changing agent 1 message...') a_dict_add = exchange_args["agent_add_dict"] a_dict_sub = exchange_args["agent_subtract_dict"] add = a_dict_add[exchange_args["add"]].unsqueeze(0) sub = a_dict_sub[exchange_args["subtract"]].unsqueeze(0) if FLAGS.cuda: add = add.cuda() sub = sub.cuda() new_m_prob = m_1e_probs.data - sub + add new_m_binary = torch.clamp(new_m_prob, 0, 1).round() m_1e_binary = _Variable(new_m_binary) debuglogger.debug(f'Old msg: {m_binary_1.data}, old prob: {m_probs_1.data}') debuglogger.debug(f'Sub: {sub}, add: {add}') debuglogger.debug(f'New msg: {new_m_binary}, new prob: {new_m_prob}') # Optionally mute communication channel if FLAGS.no_comms_channel: m_1e_binary = m_binary # Run data through agent 2 if data_context is not None: # TODO debuglogger.warning(f'Data context not supported currently') sys.exit() else: # debuglogger.info(f'Inside exchange: Train status: {train}, Message to agent 2: {m_1e_binary}') s_2e, m_2e, y_2e, r_2e = agent2( data['im_feats_2'], m_1e_binary, i_exchange, desc, use_message, batch_size, train) # Agent 2's message m_2e_binary, m_2e_probs = m_2e # Optionally corrupt agent 2's message if corrupt: debuglogger.info(f'Corrupting message...') m_2e_binary = corrupt_message(corrupt_region, agent2, m_2e_binary) # Optionally change agent 2's message if test_language_similarity: if (who_goes_first == 1 and exchange_args["change_agent"] == 2) or (who_goes_first == 2 and exchange_args["change_agent"] == 1): debuglogger.debug(f'Inside exchange: Changing agent 2 message...') a_dict_add = exchange_args["agent_add_dict"] a_dict_sub = exchange_args["agent_subtract_dict"] add = a_dict_add[exchange_args["add"]].unsqueeze(0) sub = a_dict_sub[exchange_args["subtract"]].unsqueeze(0) if FLAGS.cuda: add = add.cuda() sub = sub.cuda() new_m_prob = m_2e_probs.data - sub + add new_m_binary = torch.clamp(new_m_prob, 0, 1).round() m_2e_binary = _Variable(new_m_binary) debuglogger.debug(f'Old msg: {m_binary_1.data}, old prob: {m_probs_1.data}') debuglogger.debug(f'Sub: {sub}, add: {add}') debuglogger.debug(f'New msg: {new_m_binary}, new prob: {new_m_prob}') # Optionally mute communication channel if FLAGS.no_comms_channel: m_2e_binary = m_binary # Run data through agent 1 if data_context is not None: pass else: # debuglogger.info(f'Inside exchange: Train status: {train}, Message to agent 1: {m_2e_binary}') s_1e, m_1e, y_1e, r_1e = agent1( data['im_feats_1'], m_2e_binary, i_exchange, desc, use_message, batch_size, train) # Store rest of communication and stop information s_binary_1, s_prob_1 = s_1e s_binary_2, s_prob_2 = s_2e m_binary_2, m_probs_2 = m_2e # Save for later # TODO check stop mask # Need to be consistent about storing the a1 and a2's even though their roles are randomized during each exchange # agent1 and agent2 is a local name that refers to the order of communication. Storage refers to global labels a1 and a2 if who_goes_first == 1: stop_mask_1.append(torch.min(stop_mask_1[-1], s_binary_1.byte())) stop_mask_2.append(torch.min(stop_mask_2[-1], s_binary_2.byte())) stop_feat_1.append(s_binary_1) stop_feat_2.append(s_binary_2) stop_prob_1.append(s_prob_1) stop_prob_2.append(s_prob_2) feats_1.append(m_binary_1) feats_2.append(m_binary_2) probs_1.append(m_probs_1) probs_2.append(m_probs_2) y_1.append(y_1e) y_2.append(y_2e) r_1.append(r_1e) r_2.append(r_2e) else: stop_mask_1.append(torch.min(stop_mask_2[-1], s_binary_2.byte())) stop_mask_2.append(torch.min(stop_mask_1[-1], s_binary_1.byte())) stop_feat_1.append(s_binary_2) stop_feat_2.append(s_binary_1) stop_prob_1.append(s_prob_2) stop_prob_2.append(s_prob_1) feats_1.append(m_binary_2) feats_2.append(m_binary_1) probs_1.append(m_probs_2) probs_2.append(m_probs_1) y_1.append(y_2e) y_2.append(y_1e) r_1.append(r_2e) r_2.append(r_1e) # Terminate exchange if everyone is done conversing if break_early and stop_mask_1[-1].float().sum().data[0] == 0 and stop_mask_2[-1].float().sum().data[0] == 0: break # The final mask must always be zero. stop_mask_1[-1].data.fill_(0) stop_mask_2[-1].data.fill_(0) s = [(stop_mask_1, stop_feat_1, stop_prob_1), (stop_mask_2, stop_feat_2, stop_prob_2)] message_1 = (feats_1, probs_1) message_2 = (feats_2, probs_2) y = (y_1, y_2) y_nc = (y_1_nc, y_2_nc) y_all = [y_nc, y] r = (r_1, r_2) return s, message_1, message_2, y_all, r def get_outp(y, masks): def negent(yy): probs = F.softmax(yy, dim=1) return (torch.log(probs + 1e-8) * probs).sum(1).mean() # TODO: This is wrong for the dynamic exchange, and we might want a "per example" # entropy for either exchange (this version is mean across batch). negentropy = list(map(negent, y)) # TODO check ok for new agents if masks is not None: batch_size = y[0].size(0) exchange_steps = len(masks) inp = torch.cat([yy.view(batch_size, 1, -1) for yy in y], 1) mask = torch.cat(masks, 1).view( batch_size, exchange_steps, 1).expand_as(inp) outp = torch.masked_select(inp, mask.detach()).view(batch_size, -1) if FLAGS.debug: # Each mask index should have exactly 1 true value. assert all([mm.data[0] == 1 for mm in torch.cat(masks, 1).sum(1)]) return outp, negentropy else: return y[-1], negentropy def calculate_loss_binary(binary_features, binary_probs, rewards, baseline_rewards, entropy_penalty): '''Calculates the reinforcement learning loss on the agent communication vectors''' log_p_z = Variable(binary_features.data) * torch.log(binary_probs + 1e-8) + \ (1 - Variable(binary_features.data)) * \ torch.log(1 - binary_probs + 1e-8) log_p_z = log_p_z.sum(1) weight = Variable(rewards) - \ Variable(baseline_rewards.clone().detach().data) if rewards.size(0) > 1: # Ensures weights are not larger than 1 weight = weight / max(1., torch.std(weight.data).item()) loss = torch.mean(-1 * weight * log_p_z) # Must do both sides of negent, otherwise is skewed towards 0. initial_negent = (torch.log(binary_probs + 1e-8) * binary_probs).sum(1).mean() inverse_negent = (torch.log((1. - binary_probs) + 1e-8) * (1. - binary_probs)).sum(1).mean() negentropy = initial_negent + inverse_negent if entropy_penalty is not None: loss = (loss + entropy_penalty * negentropy) return loss, negentropy def multistep_loss_binary(binary_features, binary_probs, rewards, baseline_rewards, masks, entropy_penalty): ''' Same as calculate loss binary but with multiple communications per exchange''' if masks is not None: # TODO - implement for new agents pass else: outp = list(map(lambda feat, prob, scores: calculate_loss_binary(feat, prob, rewards, scores, entropy_penalty), binary_features, binary_probs, baseline_rewards)) losses = [o[0] for o in outp] entropies = [o[1] for o in outp] loss = sum(losses) / len(binary_features) return loss, entropies def calculate_loss_bas(baseline_scores, rewards): loss_bas = nn.MSELoss()(baseline_scores, Variable(rewards)) return loss_bas def multistep_loss_bas(baseline_scores, rewards, masks): if masks is not None: # TODO - check for new agents pass else: losses = list(map(lambda scores: calculate_loss_bas(scores, rewards), baseline_scores)) loss = sum(losses) / len(baseline_scores) return loss def calculate_accuracy(prediction_dist, target, batch_size, top_k): '''Calculates the prediction accuracy using correct@top_k Returns: - accuracy: float - correct: boolean vector of batch_size elements. 1 indicates prediction was correct@top_k - top_1: boolean vector of batch_size elements. 1 indicates prediction was correct (top 1) ''' assert batch_size == target.size(0) target_exp = target.view(-1, 1).expand(batch_size, top_k) top_k_ind = torch.from_numpy( prediction_dist.data.cpu().numpy().argsort()[:, -top_k:]).long() correct = (top_k_ind == target_exp.cpu()).sum(dim=1) top_1_ind = torch.from_numpy( prediction_dist.data.cpu().numpy().argsort()[:, -1:]).long() top_1 = (top_1_ind == target.view(-1, 1).cpu()).sum(dim=1) accuracy = correct.sum() / float(batch_size) return accuracy, correct, top_1 def get_classification_loss_and_stats(predictions, targets): ''' Arguments: - predictions: predicted logits for the classes - targets: correct classes Returns: - dist: logs of the predicted probability distribution over the classes - argmax: predicted class - argmax_prob: predicted class probability - ent: average entropy of the predicted probability distributions (over the batch) - nll_loss: Negative Log Likelihood loss between the predictions and targets - logs: Individual log likelihoods across the batch ''' dist = F.log_softmax(predictions, dim=1) maxdist, argmax = dist.data.max(1) probs = F.softmax(predictions, dim=1) ent = (torch.log(probs + 1e-8) * probs).sum(1).mean() debuglogger.debug(f'Mean entropy: {-ent.data.item()}') nll_loss = nn.NLLLoss()(dist, Variable(targets)) logs = loglikelihood(Variable(dist.data), Variable(targets.view(-1, 1))) return (dist, maxdist, argmax, ent, nll_loss, logs) def run(): flogger = FileLogger(FLAGS.log_file) logger = Logger( env=FLAGS.env, experiment_name=FLAGS.experiment_name, enabled=FLAGS.visdom) flogger.Log("Flag Values:\n" + json.dumps(FLAGS.FlagValuesDict(), indent=4, sort_keys=True)) if not os.path.exists(FLAGS.json_file): with open(FLAGS.json_file, "w") as f: f.write(json.dumps(FLAGS.FlagValuesDict(), indent=4, sort_keys=True)) # Initialize Agents agents = [] optimizers_dict = {} models_dict = {} frozen_agents = {} # Only used for training agent communities num_agents_per_community = None intra_pool_connect_p = None train_vec_prob = None # Probability distribution over agent pairs, used to sample pairs per batch agent_idx_list = None # Mapping from train_vec indices to agent pairs eval_agent_list = None # List of agent pairs to evaluate each dev eval # Check agent setup if FLAGS.num_agents > 2 and not (FLAGS.agent_pools or FLAGS.agent_communities): flogger.Log("{} is too many agents. There can only be two agents if FLAGS.agent_pools and FLAGS.agent_communities are false".format(FLAGS.num_agents)) sys.exit() if FLAGS.agent_pools and FLAGS.agent_communities: flogger.Log("You cannot train an agent pool and community at the same time. Please select one of the other") sys.exit() if not FLAGS.use_binary: flogger.Log("Non binary agent communication not implemented yet. Set FLAGS.use_binary to True") sys.exit() if FLAGS.agent_communities: if FLAGS.num_communities != len(FLAGS.num_agents_per_community): flogger.Log("Number of communities doesn't match length of community size list") sys.exit() if FLAGS.num_communities != len(FLAGS.community_checkpoints): flogger.Log("Number of communities doesn't match length of community checkpoint list. Use None to train from scratch.") sys.exit() if FLAGS.num_communities != len(FLAGS.intra_pool_connect_p): flogger.Log("Number of communities doesn't match length of intra community connectivity list") sys.exit() if FLAGS.community_type == "chain" and FLAGS.num_communities <= 2: flogger.Log("There must be at least 3 communities to train in a chain") sys.exit() num_agents_per_community = [int(x) for x in FLAGS.num_agents_per_community] intra_pool_connect_p = [float(x) for x in FLAGS.intra_pool_connect_p] if FLAGS.num_agents != sum(num_agents_per_community): flogger.Log("Total number of agents does not match sum of agents in each community") sys.exit() flogger.Log(f'Total number of agents: {FLAGS.num_agents}, Agents per community: {num_agents_per_community}') if FLAGS.test_language_similarity and not FLAGS.eval_only: flogger.Log("Can only test language similarity in eval only mode") sys.exit() if FLAGS.test_language_similarity: if len(FLAGS.agent_dicts) != FLAGS.num_agents: flogger.Log("There must be a code dictionary per agent") sys.exit() # Create agents for _ in range(FLAGS.num_agents): agent = Agent(im_feature_type=FLAGS.img_feat, im_feat_dim=FLAGS.img_feat_dim, h_dim=FLAGS.h_dim, m_dim=FLAGS.m_dim, desc_dim=FLAGS.desc_dim, num_classes=FLAGS.num_classes, s_dim=FLAGS.s_dim, use_binary=FLAGS.use_binary, use_attn=FLAGS.visual_attn, attn_dim=FLAGS.attn_dim, use_MLP=FLAGS.use_MLP, cuda=FLAGS.cuda, im_from_scratch=FLAGS.improc_from_scratch, dropout=FLAGS.dropout) flogger.Log("Agent {} id: {} Architecture: {}".format(_ + 1, id(agent), agent)) total_params = sum([functools.reduce(lambda x, y: x * y, p.size(), 1.0) for p in agent.parameters()]) flogger.Log("Total Parameters: {}".format(total_params)) agents.append(agent) # Optimizer if FLAGS.optim_type == "SGD": optimizer_agent = optim.SGD( agent.parameters(), lr=FLAGS.learning_rate) elif FLAGS.optim_type == "Adam": optimizer_agent = optim.Adam( agent.parameters(), lr=FLAGS.learning_rate) elif FLAGS.optim_type == "RMSprop": optimizer_agent = optim.RMSprop( agent.parameters(), lr=FLAGS.learning_rate) else: raise NotImplementedError optim_name = "optimizer_agent" + str(_ + 1) agent_name = "agent" + str(_ + 1) optimizers_dict[optim_name] = optimizer_agent models_dict[agent_name] = agent if FLAGS.agent_communities: flogger.Log(f'Training {FLAGS.num_communities} communities of agents, type: {FLAGS.community_type}') flogger.Log("Number of agents: {}".format(len(agents))) for k in optimizers_dict: flogger.Log("Optimizer {}: {}".format(k, optimizers_dict[k])) # Create additional data structures if training with agent communites # These handle to sampling of agents during training, and selection of agent pairs when evaluating of the dev set if FLAGS.agent_communities: (train_vec_prob, agent_idx_list) = build_train_matrix(num_agents_per_community, FLAGS.community_type, intra_pool_connect_p, FLAGS.inter_pool_connect_p, FLAGS.ratio_adjust, FLAGS.intra_inter_ratio) eval_agent_list = build_eval_list(num_agents_per_community, FLAGS.community_type, train_vec_prob) flogger.Log(f'Train matrix: {torch.from_numpy(np.reshape(train_vec_prob, (FLAGS.num_agents, FLAGS.num_agents)))}') flogger.Log(f'Eval agent list: {eval_agent_list}') # Training metrics epoch = 0 step = 0 best_dev_acc = 0 # Optionally load previously saved models if os.path.exists(FLAGS.checkpoint): flogger.Log("Loading from: " + FLAGS.checkpoint) data = torch_load(FLAGS.checkpoint, models_dict, optimizers_dict) flogger.Log("Loaded at step: {} and best dev acc: {}".format( data['step'], data['best_dev_acc'])) step = data['step'] best_dev_acc = data['best_dev_acc'] if FLAGS.agent_communities: # Load train and eval matrices flogger.Log(f"Train and eval matrices for {FLAGS.checkpoint} exists?: {os.path.isfile(FLAGS.checkpoint + '_train_vec.pkl')}") if os.path.isfile(FLAGS.checkpoint + '_train_vec.pkl'): train_vec_prob = pickle.load(open(FLAGS.checkpoint + '_train_vec.pkl', 'rb')) agent_idx_list = pickle.load(open(FLAGS.checkpoint + '_agent_idx_list.pkl', 'rb')) eval_agent_list = pickle.load(open(FLAGS.checkpoint + '_eval_agent_list.pkl', 'rb')) else: # Try just .pt suffix. Used when evaluating checkpoints saved after xx steps e.g. .pt_50000 # Or for the best checkpoint, .pt_best. # During normal training the train and eval matrices are only associated with .pt suffix stripped_checkpoint = FLAGS.checkpoint.split("pt_")[0] + 'pt' flogger.Log("Loading train and eval matrices from: " + FLAGS.checkpoint + " failed") flogger.Log("Trying " + stripped_checkpoint) train_vec_prob = pickle.load(open(stripped_checkpoint + '_train_vec.pkl', 'rb')) agent_idx_list = pickle.load(open(stripped_checkpoint + '_agent_idx_list.pkl', 'rb')) eval_agent_list = pickle.load(open(stripped_checkpoint + '_eval_agent_list.pkl', 'rb')) flogger.Log(f'Train matrix: {torch.from_numpy(np.reshape(train_vec_prob, (FLAGS.num_agents, FLAGS.num_agents)))}') flogger.Log(f'Eval agent list: {eval_agent_list}') elif FLAGS.agent_communities: torch_load_communities(FLAGS.community_checkpoints, num_agents_per_community, models_dict, optimizers_dict) if FLAGS.agent_communities: # Save additional community data pickle.dump(train_vec_prob, open(FLAGS.checkpoint + '_train_vec.pkl', 'wb')) pickle.dump(agent_idx_list, open(FLAGS.checkpoint + '_agent_idx_list.pkl', 'wb')) pickle.dump(eval_agent_list, open(FLAGS.checkpoint + '_eval_agent_list.pkl', 'wb')) # Copy agents to measure how much the language has changed frozen_agents = copy.deepcopy(models_dict) frozen_optimizers = copy.deepcopy(optimizers_dict) if os.path.exists(FLAGS.checkpoint) and FLAGS.agent_communities: # Load original agents into frozen agents so that self play comparison is with original agents flogger.Log("Loading original agent pools into frozen agents") torch_load_communities(FLAGS.community_checkpoints, num_agents_per_community, frozen_agents, frozen_optimizers) # GPU support if FLAGS.cuda: for m in models_dict.values(): m.cuda() for m in frozen_agents.values(): m.cuda() for o in optimizers_dict.values(): recursively_set_device(o.state_dict(), gpu=0) # If training / evaluating with pools of agents or communities of agents sample with each batch if FLAGS.agent_pools or FLAGS.agent_communities: agent1 = None agent2 = None optimizer_agent1 = None optimizer_agent2 = None agent_idxs = [None, None] # Otherwise keep agents fixed for each batch else: if FLAGS.num_agents == 1: agent1 = agents[0] agent2 = agents[0] optimizer_agent1 = optimizers_dict["optimizer_agent1"] optimizer_agent2 = optimizers_dict["optimizer_agent1"] agent_idxs = [1, 1] else: agent1 = agents[0] agent2 = agents[1] optimizer_agent1 = optimizers_dict["optimizer_agent1"] optimizer_agent2 = optimizers_dict["optimizer_agent2"] agent_idxs = [1, 2] # Alternatives to training. if FLAGS.eval_only: if not os.path.exists(FLAGS.checkpoint): raise Exception("Must provide valid checkpoint.") debuglogger.info("Evaluating on validation set") step = i_batch = epoch = 0 # Storage for results dev_accuracy_id = [] dev_accuracy_ood = [] dev_accuracy_self_com = [] for i in range(FLAGS.num_agents): dev_accuracy_id.append({'total_acc_both_nc': [], # % both agents right before comms 'total_acc_both_com': [], # % both agents right after comms 'total_acc_atl1_nc': [], # % at least 1 agent right before comms 'total_acc_atl1_com': [] # % at least 1 agent right after comms }) dev_accuracy_ood.append({'total_acc_both_nc': [], # % both agents right before comms 'total_acc_both_com': [], # % both agents right after comms 'total_acc_atl1_nc': [], # % at least 1 agent right before comms 'total_acc_atl1_com': [] # % at least 1 agent right after comms }) dev_accuracy_self_com.append({'total_acc_both_nc': [], # % both agents right before comms 'total_acc_both_com': [], # % both agents right after comms 'total_acc_atl1_nc': [], # % at least 1 agent right before comms 'total_acc_atl1_com': [] # % at least 1 agent right after comms }) if FLAGS.eval_xproduct: # Complete evaluation on all possible pairs of agents for i in range(FLAGS.num_agents): for j in range(FLAGS.num_agents): flogger.Log("Agent 1: {}".format(i + 1)) logger.log(key="Agent 1: ", val=i + 1, step=step) agent1 = models_dict["agent" + str(i + 1)] flogger.Log("Agent 2: {}".format(j + 1)) logger.log(key="Agent 2: ", val=j + 1, step=step) agent2 = models_dict["agent" + str(j + 1)] if i == j: # Create a copy of agents playing with themselves to avoid sharing the hidden state agent2 = Agent(im_feature_type=FLAGS.img_feat, im_feat_dim=FLAGS.img_feat_dim, h_dim=FLAGS.h_dim, m_dim=FLAGS.m_dim, desc_dim=FLAGS.desc_dim, num_classes=FLAGS.num_classes, s_dim=FLAGS.s_dim, use_binary=FLAGS.use_binary, use_attn=FLAGS.visual_attn, attn_dim=FLAGS.attn_dim, use_MLP=FLAGS.use_MLP, cuda=FLAGS.cuda, im_from_scratch=FLAGS.improc_from_scratch, dropout=FLAGS.dropout) agent2.load_state_dict(agent1.state_dict()) if FLAGS.cuda: agent2.cuda() if i == 0 and j == 0: # Report in domain development accuracy and store examples # TODO: Store examples currently disabled. To fix store examples - image saving disabled because it clogs the memory and makes everything very slow dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{j + 1}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{j + 1}') # Report out of domain development accuracy and store examples dev_accuracy_ood[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_outdomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'Out of Domain Agents {i + 1},{j + 1}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{j + 1}') else: # Report in domain development accuracy dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{j + 1}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{j + 1}') # Report out of domain development accuracy dev_accuracy_ood[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_outdomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'Out of Domain Agents {i + 1},{j + 1}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{j + 1}') elif FLAGS.gen_community_messages: # Get list of agent pairs agent_pairs = get_msg_pairs(FLAGS.community_structure) flogger.Log(f"Agent pairs to generate messages for: {agent_pairs}") for (i, j) in agent_pairs: flogger.Log("Agent 1: {}".format(i + 1)) logger.log(key="Agent 1: ", val=i + 1, step=step) agent1 = models_dict["agent" + str(i + 1)] flogger.Log("Agent 2: {}".format(j + 1)) logger.log(key="Agent 2: ", val=j + 1, step=step) agent2 = models_dict["agent" + str(j + 1)] # Report in domain development accuracy dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{j + i}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=True, agent_tag=f'eval_only_A_{i + 1}_{j + 1}') # Report out of domain development accuracy dev_accuracy_ood[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_path, False, dev_accuracy_ood[i], logger, flogger, f'Out of Domain Agents {i + 1},{j + 1}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag="") elif FLAGS.test_language_similarity: # Load agent dictionaries code_dicts = [] if FLAGS.self_similarity: for i in range(FLAGS.num_agents): _ = pickle.load(open(FLAGS.agent_dicts[i], 'rb')) code_dicts.append(_) for i in range(FLAGS.num_agents - 1): flogger.Log("Agent 1: {}".format(i + 1)) logger.log(key="Agent 1: ", val=i + 1, step=step) agent1 = models_dict["agent" + str(i + 1)] flogger.Log("Agent 2: {}".format(i + 2)) logger.log(key="Agent 2: ", val=i + 2, step=step) agent2 = models_dict["agent" + str(i + 2)] dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{i + 2}', agent_dicts=[code_dicts, code_dicts], agent_idxs=[i, i + 1], agent_groups=[1, 2]) elif FLAGS.ancestor_similarity: # The main difference vs. the other methods for testing language similarity is that both agents always come from group 1. Only the codes for group 2 are present and used to measure the similarity of the current protocol to an ancestor protocol, as well as to trace the ancestry of individual words. # Load codes for checkpointed agents debuglogger.info(f'Agent codes: {FLAGS.agent_dicts}') debuglogger.info(f'Ancestor codes: {FLAGS.agent_supplementary_dicts}') cd = [] for i in range(FLAGS.num_agents): _ = pickle.load(open(FLAGS.agent_dicts[i], 'rb')) cd.append(_) code_dicts.append(cd) # Load ancestor codes. The ancestor codes are given through FLAGS.agent_supplementary_dicts cd = [] for i in range(FLAGS.num_supplementary_agents): _ = pickle.load(open(FLAGS.agent_supplementary_dicts[i], 'rb')) cd.append(_) code_dicts.append(cd) for i in range(FLAGS.num_agents - 1): flogger.Log("Agent 1: {}".format(i + 1)) logger.log(key="Agent 1: ", val=i + 1, step=step) agent1 = models_dict["agent" + str(i + 1)] flogger.Log("Agent 2: {}".format(i + 2)) logger.log(key="Agent 2: ", val=i + 2, step=step) agent2 = models_dict["agent" + str(i + 2)] dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{i + 2}', agent_dicts=[code_dicts[0], code_dicts[1]], agent_idxs=[i, i + 1], agent_groups=[1, 1]) else: '''Please note that this is an experimental feature. To be tested further.''' # Testing similarity between two language pairs. # The pairs are assumed to be 2 of the pairs in a checkpointed community # The pairs are specified through FLAGS.compare_language_pairs and are the index into FLAGS.num_agents_per_community # The code dicts are given in the command line through FLAGS.agent_dicts and FLAGS.agent_supplementary dicts cd = [] for i in range(FLAGS.num_agents): _ = pickle.load(open(FLAGS.agent_dicts[i], 'rb')) cd.append(_) # Extract codes for group 1 idx = int(FLAGS.compare_language_pairs[0]) start = end = 0 g1_bounds = (None, None) for i, p in enumerate(FLAGS.num_agents_per_community): end += int(p) if idx == i: g1_bounds = (start, end) break start += int(p) code_dicts.append(cd[start:end]) # Extract codes for group 2 idx = int(FLAGS.compare_language_pairs[1]) start = end = 0 g2_bounds = (None, None) for i, p in enumerate(FLAGS.num_agents_per_community): end += int(p) if idx == i: g2_bounds = (start, end) break start += int(p) code_dicts.append(cd[start:end]) flogger.Log(f"Comparing groups: {FLAGS.compare_language_pairs}") flogger.Log(f"G1 bounds: {g1_bounds}, G2 bounds: {g2_bounds}") # Evaluate on group 1 pairs for i in range(g1_bounds[0], g1_bounds[1] - 1): flogger.Log("Agent 1: {}".format(i + 1)) logger.log(key="Agent 1: ", val=i + 1, step=step) agent1 = models_dict["agent" + str(i + 1)] flogger.Log("Agent 2: {}".format(i + 2)) logger.log(key="Agent 2: ", val=i + 2, step=step) agent2 = models_dict["agent" + str(i + 2)] dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{i + 2}', agent_dicts=[code_dicts[0], code_dicts[1]], agent_idxs=[i, i + 1], agent_groups=[1, 1]) # Evaluate on group 2 pairs for i in range(g2_bounds[0], g2_bounds[1] - 1): flogger.Log("Agent 1: {}".format(i + 1)) logger.log(key="Agent 1: ", val=i + 1, step=step) agent1 = models_dict["agent" + str(i + 1)] flogger.Log("Agent 2: {}".format(i + 2)) logger.log(key="Agent 2: ", val=i + 2, step=step) agent2 = models_dict["agent" + str(i + 2)] dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{i + 2}', agent_dicts=[code_dicts[0], code_dicts[1]], agent_idxs=[i - g2_bounds[0], i + 1 - g2_bounds[0]], agent_groups=[2, 2]) elif FLAGS.eval_agent_communities: eval_community(eval_agent_list, models_dict, dev_accuracy_id[0], logger, flogger, epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=True, agent_tag="no_tag") else: # For the pairs of agents calculate results # Applies to both pools of agents and an agent pair for i in range(FLAGS.num_agents - 1): flogger.Log("Agent 1: {}".format(i + 1)) logger.log(key="Agent 1: ", val=i + 1, step=step) agent1 = models_dict["agent" + str(i + 1)] flogger.Log("Agent 2: {}".format(i + 2)) logger.log(key="Agent 2: ", val=i + 2, step=step) agent2 = models_dict["agent" + str(i + 2)] if i == 0: # Report in domain development accuracy and analyze messages and store examples dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=True, agent_tag=f'eval_only_A_{i + 1}_{i + 2}') else: # Report in domain development accuracy dev_accuracy_id[i], total_accuracy_com = get_and_log_dev_performance( agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'In Domain Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=True, agent_tag=f'eval_only_A_{i + 1}_{i + 2}') # Report out of domain development accuracy dev_accuracy_ood[i], total_accuracy_com = get_and_log_dev_performance(agent1, agent2, FLAGS.dataset_outdomain_valid_path, True, dev_accuracy_id[i], logger, flogger, f'Out of Domain Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_A_{i + 1}_{i + 2}') # Report in domain development accuracy when agents communicate with themselves for i in range(FLAGS.num_agents): agent1 = models_dict["agent" + str(i + 1)] # Create a copy of agents playing with themselves to avoid sharing the hidden state agent2 = copy.deepcopy(agent1) flogger.Log("Agent {} self communication: id {}".format(i + 1, id(agent))) dev_accuracy_self_com[i], total_accuracy_com = get_and_log_dev_performance( agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_self_com[i], logger, flogger, "Agent " + str(i + 1) + " self communication: In Domain", epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'eval_only_self_com_A_{i + 1}') sys.exit() # Training loop while epoch < FLAGS.max_epoch: flogger.Log("Starting epoch: {}".format(epoch)) # Read dataset randomly into batches if FLAGS.dataset == "shapeworld": dataloader = load_shapeworld_dataset(FLAGS.dataset_path, FLAGS.glove_path, FLAGS.dataset_mode, FLAGS.dataset_size_train, FLAGS.dataset_type, FLAGS.dataset_name, FLAGS.batch_size, FLAGS.random_seed, FLAGS.shuffle_train, FLAGS.img_feat, FLAGS.cuda, truncate_final_batch=False) else: raise NotImplementedError # Keep track of metrics batch_accuracy = {'total_nc': [], # no communicaton 'total_com': [], # after communication 'rewards_1': [], # agent 1 rewards 'rewards_2': [], # agent 2 rewards 'total_acc_both_nc': [], # % both agents right before comms 'total_acc_both_com': [], # % both agents right after comms 'total_acc_atl1_nc': [], # % at least 1 agent right before comms 'total_acc_atl1_com': [], # % at least 1 agent right after comms 'agent1_nc': [], # no communicaton 'agent2_nc': [], # no communicaton 'agent1_com': [], # after communicaton 'agent2_com': [] # after communicaton } dev_accuracy_id = {'total_acc_both_nc': [], # % both agents right before comms 'total_acc_both_com': [], # % both agents right after comms 'total_acc_atl1_nc': [], # % at least 1 agent right before comms 'total_acc_atl1_com': [] # % at least 1 agent right after comms } dev_accuracy_ood = {'total_acc_both_nc': [], # % both agents right before comms 'total_acc_both_com': [], # % both agents right after comms 'total_acc_atl1_nc': [], # % at least 1 agent right before comms 'total_acc_atl1_com': [] # % at least 1 agent right after comms } dev_accuracy_id_pairs = [] dev_accuracy_self_com = [] for i in range(FLAGS.num_agents): dev_accuracy_id_pairs.append({'total_acc_both_nc': [], # % both agents right before comms 'total_acc_both_com': [], # % both agents right after comms 'total_acc_atl1_nc': [], # % at least 1 agent right before comms 'total_acc_atl1_com': [] # % at least 1 agent right after comms }) dev_accuracy_self_com.append({'total_acc_both_nc': [], # % both agents right before comms 'total_acc_both_com': [], # % both agents right after comms 'total_acc_atl1_nc': [], # % at least 1 agent right before comms 'total_acc_atl1_com': [] # % at least 1 agent right after comms }) # Iterate through batches for i_batch, batch in enumerate(dataloader): debuglogger.debug(f'Batch {i_batch}') # Select agents if training with pools or communities if FLAGS.agent_pools: idx = random.randint(0, len(agents) - 1) agent1 = agents[idx] optimizer_agent1 = optimizers_dict["optimizer_agent" + str(idx + 1)] agent_idxs[0] = idx + 1 old_idx = idx if FLAGS.with_replacement: # Sampling second agent with replacement idx = random.randint(0, len(agents) - 1) else: # Sampling second agent without replacement while idx == old_idx: idx = random.randint(0, len(agents) - 1) agent2 = agents[idx] optimizer_agent2 = optimizers_dict["optimizer_agent" + str(idx + 1)] agent_idxs[1] = idx + 1 elif FLAGS.agent_communities: (idx1, idx2) = sample_agents(train_vec_prob, agent_idx_list) agent1 = agents[idx1] optimizer_agent1 = optimizers_dict["optimizer_agent" + str(idx1 + 1)] agent_idxs[0] = idx1 + 1 agent2 = agents[idx2] optimizer_agent2 = optimizers_dict["optimizer_agent" + str(idx2 + 1)] agent_idxs[1] = idx2 + 1 elif FLAGS.num_agents == 1: # Training with one agents agent1 = agents[0] agent2 = agents[0] optimizer_agent1 = optimizers_dict["optimizer_agent1"] optimizer_agent2 = optimizers_dict["optimizer_agent1"] agent_idxs = [1, 1] else: # Training with just a pair of agents agent1 = agents[0] agent2 = agents[1] optimizer_agent1 = optimizers_dict["optimizer_agent1"] optimizer_agent2 = optimizers_dict["optimizer_agent2"] agent_idxs = [1, 2] # Create a copy of agents playing with themselves to avoid sharing the hidden state if FLAGS.num_agents == 1 or (agent_idxs[0] == agent_idxs[1]): agent2 = Agent(im_feature_type=FLAGS.img_feat, im_feat_dim=FLAGS.img_feat_dim, h_dim=FLAGS.h_dim, m_dim=FLAGS.m_dim, desc_dim=FLAGS.desc_dim, num_classes=FLAGS.num_classes, s_dim=FLAGS.s_dim, use_binary=FLAGS.use_binary, use_attn=FLAGS.visual_attn, attn_dim=FLAGS.attn_dim, use_MLP=FLAGS.use_MLP, cuda=FLAGS.cuda, im_from_scratch=FLAGS.improc_from_scratch, dropout=FLAGS.dropout) agent2.load_state_dict(agent1.state_dict()) if FLAGS.cuda: agent2.cuda() debuglogger.debug(f'Agent 1: {agent_idxs[0]}, Agent 1: {agent_idxs[1]}') # Converted to Variable in get_classification_loss_and_stats target = batch["target"] im_feats_1 = batch["im_feats_1"] # Already Variable im_feats_2 = batch["im_feats_2"] # Already Variable p = batch["p"] desc = Variable(batch["texts_vec"]) # GPU support if FLAGS.cuda: im_feats_1 = im_feats_1.cuda() im_feats_2 = im_feats_2.cuda() target = target.cuda() desc = desc.cuda() data = {"im_feats_1": im_feats_1, "im_feats_2": im_feats_2, "p": p} exchange_args = dict() exchange_args["data"] = data exchange_args["target"] = target exchange_args["desc"] = desc exchange_args["train"] = True exchange_args["break_early"] = not FLAGS.fixed_exchange s, message_1, message_2, y_all, r = exchange( agent1, agent2, exchange_args) s_masks_1, s_feats_1, s_probs_1 = s[0] s_masks_2, s_feats_2, s_probs_2 = s[1] feats_1, probs_1 = message_1 feats_2, probs_2 = message_2 y_nc = y_all[0] y = y_all[1] # Mask loss if dynamic exchange length if FLAGS.fixed_exchange: binary_s_masks = None binary_agent1_masks = None binary_agent2_masks = None bas_agent1_masks = None bas_agent2_masks = None y1_masks = None y2_masks = None outp_1 = y[0][-1] outp_2 = y[1][-1] else: # TODO pass # Before communication predictions # Obtain predictions, loss and stats agent 1 (dist_1_nc, maxdist_1_nc, argmax_1_nc, ent_1_nc, nll_loss_1_nc, logs_1_nc) = get_classification_loss_and_stats(y_nc[0], target) # Obtain predictions, loss and stats agent 2 (dist_2_nc, maxdist_2_nc, argmax_2_nc, ent_2_nc, nll_loss_2_nc, logs_2_nc) = get_classification_loss_and_stats(y_nc[1], target) # After communication predictions # Obtain predictions, loss and stats agent 1 (dist_1, maxdist_1, argmax_1, ent_1, nll_loss_1_com, logs_1) = get_classification_loss_and_stats(outp_1, target) # Obtain predictions, loss and stats agent 2 (dist_2, maxdist_2, argmax_2, ent_2, nll_loss_2_com, logs_2) = get_classification_loss_and_stats(outp_2, target) # Store prediction entropies if FLAGS.fixed_exchange: ent_agent1_y = [ent_1] ent_agent2_y = [ent_2] else: # TODO - not implemented yet ent_agent1_y = [] ent_agent2_y = [] # Calculate accuracy accuracy_1_nc, correct_1_nc, top_1_1_nc = calculate_accuracy( dist_1_nc, target, FLAGS.batch_size, FLAGS.top_k_train) accuracy_1, correct_1, top_1_1 = calculate_accuracy( dist_1, target, FLAGS.batch_size, FLAGS.top_k_train) accuracy_2_nc, correct_2_nc, top_1_2_nc = calculate_accuracy( dist_2_nc, target, FLAGS.batch_size, FLAGS.top_k_train) accuracy_2, correct_2, top_1_2 = calculate_accuracy( dist_2, target, FLAGS.batch_size, FLAGS.top_k_train) # Calculate accuracy total_correct_nc = correct_1_nc.float() + correct_2_nc.float() total_correct_com = correct_1.float() + correct_2.float() total_accuracy_nc = (total_correct_nc == 2).sum() / float(FLAGS.batch_size) total_accuracy_com = (total_correct_com == 2).sum() / float(FLAGS.batch_size) atleast1_accuracy_nc = ( total_correct_nc > 0).sum() / float(FLAGS.batch_size) atleast1_accuracy_com = ( total_correct_com > 0).sum() / float(FLAGS.batch_size) # Calculate rewards # rewards = difference between performance before and after communication # Only use top 1 total_correct_top_1_nc = top_1_1_nc.float() + top_1_2_nc.float() total_correct_top_1_com = top_1_1.float() + top_1_2.float() if FLAGS.reward_type == 'cooperative': rewards_1 = (total_correct_top_1_com.float() - total_correct_top_1_nc.float()) rewards_2 = rewards_1 elif FLAGS.reward_type == 'cooperative_nodiff': rewards_1 = total_correct_top_1_com.float() rewards_2 = rewards_1 elif FLAGS.reward_type == 'selfish_nodiff': rewards_1 = top_1_1.float() rewards_2 = top_1_2.float() elif FLAGS.reward_type == 'selfish': rewards_1 = top_1_1.float() - top_1_1_nc.float() rewards_2 = top_1_2.float() - top_1_2_nc.float() else: debuglogger.warn(f'Reward type {FLAGS.reward_type} not recognized') sys.exit() debuglogger.debug( f'total correct top 1 com: {total_correct_top_1_com}') debuglogger.debug( f'total correct top 1 nc: {total_correct_top_1_nc}') debuglogger.debug(f'total correct com: {total_correct_com}') debuglogger.debug(f'total correct nc: {total_correct_nc}') debuglogger.debug(f'rewards_1: {rewards_1}') debuglogger.debug(f'rewards_2: {rewards_2}') # Store results batch_accuracy['agent1_nc'].append(accuracy_1_nc) batch_accuracy['agent2_nc'].append(accuracy_2_nc) batch_accuracy['agent1_com'].append(accuracy_1) batch_accuracy['agent2_com'].append(accuracy_2) batch_accuracy['total_nc'].append(total_correct_nc) batch_accuracy['total_com'].append(total_correct_com) batch_accuracy['rewards_1'].append(rewards_1) batch_accuracy['rewards_1'].append(rewards_1) batch_accuracy['total_acc_both_nc'].append(total_accuracy_nc) batch_accuracy['total_acc_both_com'].append(total_accuracy_com) batch_accuracy['total_acc_atl1_nc'].append(atleast1_accuracy_nc) batch_accuracy['total_acc_atl1_com'].append(atleast1_accuracy_com) # Cross entropy loss for each agent nll_loss_1 = FLAGS.nll_loss_weight_nc * nll_loss_1_nc + \ FLAGS.nll_loss_weight_com * nll_loss_1_com nll_loss_2 = FLAGS.nll_loss_weight_nc * nll_loss_2_nc + \ FLAGS.nll_loss_weight_com * nll_loss_2_com loss_agent1 = nll_loss_1 loss_agent2 = nll_loss_2 # If training communication channel if FLAGS.use_binary: if not FLAGS.fixed_exchange: # TODO - fix # Stop loss # TODO - check old use of entropy_s # The receiver might have no z-loss if we stop after first message from sender. debuglogger.warning( f'Error: multistep adaptive exchange not implemented yet') sys.exit() elif FLAGS.max_exchange == 1: loss_binary_1, ent_bin_1 = calculate_loss_binary( feats_1[0], probs_1[0], rewards_1, r[0][0], FLAGS.entropy_agent1) loss_binary_2, ent_bin_2 = calculate_loss_binary( feats_2[0], probs_2[0], rewards_2, r[1][0], FLAGS.entropy_agent2) loss_baseline_1 = calculate_loss_bas(r[0][0], rewards_1) loss_baseline_2 = calculate_loss_bas(r[1][0], rewards_2) ent_agent1_bin = [ent_bin_1] ent_agent2_bin = [ent_bin_2] elif FLAGS.max_exchange > 1: loss_binary_1, ent_bin_1 = multistep_loss_binary( feats_1, probs_1, rewards_1, r[0], binary_agent1_masks, FLAGS.entropy_agent1) loss_binary_2, ent_bin_2 = multistep_loss_binary( feats_2, probs_2, rewards_2, r[1], binary_agent2_masks, FLAGS.entropy_agent2) loss_baseline_1 = multistep_loss_bas(r[0], rewards_1, bas_agent1_masks) loss_baseline_2 = multistep_loss_bas(r[1], rewards_2, bas_agent2_masks) ent_agent1_bin = ent_bin_1 ent_agent2_bin = ent_bin_2 debuglogger.debug(f'Loss bin 1: {loss_binary_1} bin 2: {loss_binary_2}') debuglogger.debug(f'Loss baseline 1: {loss_baseline_1} baseline 2: {loss_baseline_2}') debuglogger.debug(f'Entropy bin 1: {ent_agent1_bin} Entropy bin 2: {ent_agent1_bin}') if FLAGS.use_binary: loss_agent1 += FLAGS.rl_loss_weight * loss_binary_1 loss_agent2 += FLAGS.rl_loss_weight * loss_binary_2 if not FLAGS.fixed_exchange: # TODO pass else: loss_baseline_1 = Variable(torch.zeros(1)) loss_baseline_2 = Variable(torch.zeros(1)) loss_agent1 += FLAGS.baseline_loss_weight * loss_baseline_1 loss_agent2 += FLAGS.baseline_loss_weight * loss_baseline_2 if FLAGS.num_agents == 1 or (agent_idxs[0] == agent_idxs[1]): debuglogger.info(f'Agent 1 and 2 are the same') optimizer_agent1.zero_grad() agent2.zero_grad() # Get gradients from both sides of the exchange loss_agent1.backward(retain_graph=True) nn.utils.clip_grad_norm(agent1.parameters(), max_norm=1.) loss_agent2.backward() nn.utils.clip_grad_norm(agent2.parameters(), max_norm=1.) # Add agent2 gradients to agent1 gradients for p1, p2 in zip(agent1.parameters(), agent2.parameters()): if p1.grad is not None: p1.grad += p2.grad # Only need to update agent1 optimizer_agent1.step() else: debuglogger.debug(f'Different agents') # Update agent1 optimizer_agent1.zero_grad() loss_agent1.backward() nn.utils.clip_grad_norm(agent1.parameters(), max_norm=1.) optimizer_agent1.step() # Update agent2 optimizer_agent2.zero_grad() loss_agent2.backward() nn.utils.clip_grad_norm(agent2.parameters(), max_norm=1.) optimizer_agent2.step() # Print logs regularly if step % FLAGS.log_interval == 0: # Average batch accuracy avg_batch_acc_total_nc = np.array( batch_accuracy['total_acc_both_nc'][-FLAGS.log_interval:]).mean() avg_batch_acc_total_com = np.array( batch_accuracy['total_acc_both_com'][-FLAGS.log_interval:]).mean() avg_batch_acc_atl1_nc = np.array( batch_accuracy['total_acc_atl1_nc'][-FLAGS.log_interval:]).mean() avg_batch_acc_atl1_com = np.array( batch_accuracy['total_acc_atl1_com'][-FLAGS.log_interval:]).mean() # Log accuracy log_acc = "Epoch: {} Step: {} Batch: {} Agent 1: {} Agent 2: {} Training Accuracy:\nBefore comms: Both correct: {} At least 1 correct: {}\nAfter comms: Both correct: {} At least 1 correct: {}".format(epoch, step, i_batch, agent_idxs[0], agent_idxs[1], avg_batch_acc_total_nc, avg_batch_acc_atl1_nc, avg_batch_acc_total_com, avg_batch_acc_atl1_com) flogger.Log(log_acc) # Agent1 log_loss_agent1 = "Epoch: {} Step: {} Batch: {} Loss Agent1: {}".format( epoch, step, i_batch, loss_agent1.data.item()) flogger.Log(log_loss_agent1) # Agent 1 breakdown log_loss_agent1_detail = "Epoch: {} Step: {} Batch: {} Loss Agent1: NLL: {} (BC:{} / AC:{}), RL: {}, Baseline: {} ".format( epoch, step, i_batch, nll_loss_1.data.item(), nll_loss_1_nc.data.item(), nll_loss_1_com.data.item(), loss_binary_1.data.item(), loss_baseline_1.data.item()) flogger.Log(log_loss_agent1_detail) # Agent2 log_loss_agent2 = "Epoch: {} Step: {} Batch: {} Loss Agent2: {}".format( epoch, step, i_batch, loss_agent2.data.item()) flogger.Log(log_loss_agent2) # Agent 2 breakdown log_loss_agent2_detail = "Epoch: {} Step: {} Batch: {} Loss Agent2: NLL: {} (BC:{} / AC:{}), RL: {}, Baseline: {} ".format( epoch, step, i_batch, nll_loss_2.data.item(), nll_loss_2_nc.data.item(), nll_loss_2_com.data.item(), loss_binary_2.data.item(), loss_baseline_2.data.item()) flogger.Log(log_loss_agent2_detail) # Log predictions log_pred = "Predictions: Target | Agent1 BC | Agent1 AC | Agent2 BC | Agent2 AC: {}".format( torch.cat([target, argmax_1_nc, argmax_1, argmax_2_nc, argmax_2], 0).view(-1, FLAGS.batch_size)) flogger.Log(log_pred) # Log Entropy for both Agents if FLAGS.use_binary: if len(ent_agent1_bin) > 0: log_ent_agent1_bin = "Entropy Agent1 Binary" for i, ent in enumerate(ent_agent1_bin): log_ent_agent1_bin += "\n{}. {}".format( i, -ent.data.item()) log_ent_agent1_bin += "\n" flogger.Log(log_ent_agent1_bin) if len(ent_agent2_bin) > 0: log_ent_agent2_bin = "Entropy Agent2 Binary" for i, ent in enumerate(ent_agent2_bin): log_ent_agent2_bin += "\n{}. {}".format( i, -ent.data.item()) log_ent_agent2_bin += "\n" flogger.Log(log_ent_agent2_bin) if len(ent_agent1_y) > 0: log_ent_agent1_y = "Entropy Agent1 Predictions\n" log_ent_agent1_y += "No comms entropy {}\n Comms entropy\n".format( -ent_1_nc.data.item()) for i, ent in enumerate(ent_agent1_y): log_ent_agent1_y += "\n{}. {}".format(i, -ent.data.item()) log_ent_agent1_y += "\n" flogger.Log(log_ent_agent1_y) if len(ent_agent2_y) > 0: log_ent_agent2_y = "Entropy Agent2 Predictions\n" log_ent_agent2_y += "No comms entropy {}\n Comms entropy\n".format( -ent_2_nc.data.item()) for i, ent in enumerate(ent_agent2_y): log_ent_agent2_y += "\n{}. {}".format(i, -ent.data.item()) log_ent_agent2_y += "\n" flogger.Log(log_ent_agent2_y) # Agent 1 logger.log(key="Loss Agent 1 (Total)", val=loss_agent1.data.item(), step=step) logger.log(key="Loss Agent 1 (NLL)", val=nll_loss_1.data.item(), step=step) logger.log(key="Loss Agent 1 (NLL NC)", val=nll_loss_1_nc.data.item(), step=step) logger.log(key="Loss Agent 1 (NLL COM)", val=nll_loss_1_com.data.item(), step=step) if FLAGS.use_binary: logger.log(key="Loss Agent 1 (RL)", val=loss_binary_1.data.item(), step=step) logger.log(key="Loss Agent 1 (BAS)", val=loss_baseline_1.data.item(), step=step) if not FLAGS.fixed_exchange: # TODO pass # Agent 2 logger.log(key="Loss Agent 2 (Total)", val=loss_agent2.data.item(), step=step) logger.log(key="Loss Agent 2 (NLL)", val=nll_loss_2.data.item(), step=step) logger.log(key="Loss Agent 2 (NLL NC)", val=nll_loss_2_nc.data.item(), step=step) logger.log(key="Loss Agent 2 (NLL COM)", val=nll_loss_2_com.data.item(), step=step) if FLAGS.use_binary: logger.log(key="Loss Agent 2 (RL)", val=loss_binary_2.data.item(), step=step) logger.log(key="Loss Agent 2 (BAS)", val=loss_baseline_2.data.item(), step=step) if not FLAGS.fixed_exchange: # TODO pass # Accuracy metrics logger.log(key="Training Accuracy (Total, BC)", val=avg_batch_acc_total_nc, step=step) logger.log(key="Training Accuracy (At least 1, BC)", val=avg_batch_acc_atl1_nc, step=step) logger.log(key="Training Accuracy (Total, COM)", val=avg_batch_acc_total_com, step=step) logger.log(key="Training Accuracy (At least 1, COM)", val=avg_batch_acc_atl1_com, step=step) # Report development accuracy if step % FLAGS.log_dev == 0: # Report in domain development accuracy and checkpoint if best result log_agents = "Epoch: {} Step: {} Batch: {} Agent 1: {} Agent 2: {}".format( epoch, step, i_batch, agent_idxs[0], agent_idxs[1]) flogger.Log(log_agents) dev_accuracy_id, total_accuracy_com = get_and_log_dev_performance( agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id, logger, flogger, "In Domain", epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag="no_tag") if step >= FLAGS.save_after and total_accuracy_com > best_dev_acc: best_dev_acc = total_accuracy_com flogger.Log( "Checkpointing with best In Domain Development Accuracy (both right after comms): {}".format(best_dev_acc)) # Optionally store additional information data = dict(step=step, best_dev_acc=best_dev_acc) torch_save(FLAGS.checkpoint + "_best", data, models_dict, optimizers_dict, gpu=0 if FLAGS.cuda else -1) # Re-run in domain dev performance and log examples and analyze messages for a number of pairs of agents # Time consuming to run, only run if dev_acc high enough if best_dev_acc > 0.75 and FLAGS.agent_pools: for i in range(FLAGS.num_agents - 1): flogger.Log("Agent 1: {}".format(i + 1)) logger.log(key="Agent 1: ", val=i + 1, step=step) _agent1 = models_dict["agent" + str(i + 1)] flogger.Log("Agent 2: {}".format(i + 2)) logger.log(key="Agent 2: ", val=i + 2, step=step) _agent2 = models_dict["agent" + str(i + 2)] if i == 0: # Report in domain development accuracy and store examples dev_accuracy_id_pairs[i], total_accuracy_com = get_and_log_dev_performance( _agent1, _agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id_pairs[i], logger, flogger, f'In Domain: Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'A_{i + 1}_{i + 2}') else: # Report in domain development accuracy dev_accuracy_id_pairs[i], total_accuracy_com = get_and_log_dev_performance( _agent1, _agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id_pairs[i], logger, flogger, f'In Domain: Agents {i + 1},{i + 2}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'A_{i + 1}_{i + 2}') # Report out of domain development accuracy dev_accuracy_ood, total_accuracy_com = get_and_log_dev_performance( agent1, agent2, FLAGS.dataset_outdomain_valid_path, True, dev_accuracy_ood, logger, flogger, "Out of Domain", epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag="no_tag") # Report in domain development accuracy when training agent communities if FLAGS.agent_communities and step % FLAGS.log_community_eval == 0: eval_community(eval_agent_list, models_dict, dev_accuracy_id_pairs[0], logger, flogger, epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag="no_tag") # Log how much the language has changed by measuring the performance of one agent from each pool playing a frozen version of itself. offset = 0 for _, p in enumerate(num_agents_per_community): idx = offset + p - 1 # Select last agent from each community to play with itself key = "agent" + str(idx + 1) dev_accuracy_id, total_accuracy_com = get_and_log_dev_performance( models_dict[key], frozen_agents[key], FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id, logger, flogger, f'In Domain, Pool {_ + 1}, agent {idx + 1} playing with frozen version of itself, ids: {id(models_dict[key])}/{id(frozen_agents[key])}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag="no_tag") offset += p # Report in domain development accuracy when agents communicate with themselves if (not FLAGS.agent_communities) and step % FLAGS.log_self_com == 0: for i in range(FLAGS.num_agents): agent1 = models_dict["agent" + str(i + 1)] # Create a copy of agents playing with themselves to avoid sharing the hidden state agent2 = Agent(im_feature_type=FLAGS.img_feat, im_feat_dim=FLAGS.img_feat_dim, h_dim=FLAGS.h_dim, m_dim=FLAGS.m_dim, desc_dim=FLAGS.desc_dim, num_classes=FLAGS.num_classes, s_dim=FLAGS.s_dim, use_binary=FLAGS.use_binary, use_attn=FLAGS.visual_attn, attn_dim=FLAGS.attn_dim, use_MLP=FLAGS.use_MLP, cuda=FLAGS.cuda, im_from_scratch=FLAGS.improc_from_scratch, dropout=FLAGS.dropout) agent2.load_state_dict(agent1.state_dict()) if FLAGS.cuda: agent2.cuda() flogger.Log("Agent {} self communication: id {}".format(i + 1, id(agent))) dev_accuracy_self_com[i], total_accuracy_com = get_and_log_dev_performance( agent1, agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_self_com[i], logger, flogger, "Agent " + str(i + 1) + " self communication: In Domain", epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'self_com_A_{i + 1}') # Every FLAGS.check_accuracy_interval steps check if the agents have reached an average accuracy of 75% # If yes, then saved a version if step > 0 and (step % FLAGS.check_accuracy_interval == 0): # Only check the first 8 x 8 agents max - otherwise too time consuming _agent_accuracy = [] for i in range(min(FLAGS.num_agents, 8)): for j in range(min(FLAGS.num_agents, 8)): debuglogger.info(f"Accuracy check i:{i} j:{j}") _agent1 = models_dict["agent" + str(i + 1)] if i == j: # Create a copy of agents playing with themselves to avoid sharing the hidden state _agent2 = Agent(im_feature_type=FLAGS.img_feat, im_feat_dim=FLAGS.img_feat_dim, h_dim=FLAGS.h_dim, m_dim=FLAGS.m_dim, desc_dim=FLAGS.desc_dim, num_classes=FLAGS.num_classes, s_dim=FLAGS.s_dim, use_binary=FLAGS.use_binary, use_attn=FLAGS.visual_attn, attn_dim=FLAGS.attn_dim, use_MLP=FLAGS.use_MLP, cuda=FLAGS.cuda, im_from_scratch=FLAGS.improc_from_scratch, dropout=FLAGS.dropout) _agent2.load_state_dict(agent1.state_dict()) if FLAGS.cuda: _agent2.cuda() else: _agent2 = models_dict["agent" + str(j + 1)] dev_accuracy_id_pairs[i], total_accuracy_com = get_and_log_dev_performance( _agent1, _agent2, FLAGS.dataset_indomain_valid_path, True, dev_accuracy_id_pairs[i], logger, flogger, f'Average Check: In Domain: Agents {i + 1},{j + 1}', epoch, step, i_batch, store_examples=False, analyze_messages=False, save_messages=False, agent_tag=f'A_{i + 1}_{j + 1}') _agent_accuracy.append(total_accuracy_com) if FLAGS.num_agents == 2: _avg_accuracy = np.mean([_agent_accuracy[1], _agent_accuracy[2]]) else: _avg_accuracy = np.mean(_agent_accuracy) flogger.Log(f"Step {step}: Average accuracy: {_avg_accuracy}, Individual accuracies: {_agent_accuracy}") if _avg_accuracy > 0.80: flogger.Log(f"Checkpointing a model with {_avg_accuracy} average accuracy at {step} steps") # Optionally store additional information data = dict(step=step, best_dev_acc=best_dev_acc) torch_save(FLAGS.checkpoint + "_{0:.4f}".format(_avg_accuracy), data, models_dict, optimizers_dict, gpu=0 if FLAGS.cuda else -1) flogger.Log(f"Accuracy reached at least 75% on average, stopping training at step {step}...") sys.exit() # Save model periodically (overwrites most recent) if step >= FLAGS.save_after and step % FLAGS.save_interval == 0: flogger.Log("Checkpointing.") # Optionally store additional information data = dict(step=step, best_dev_acc=best_dev_acc) torch_save(FLAGS.checkpoint, data, models_dict, optimizers_dict, gpu=0 if FLAGS.cuda else -1) # Save separate copy of model every FLAGS.save_distinct_interval steps if step >= FLAGS.save_after and step % FLAGS.save_distinct_interval == 0: flogger.Log(f"Checkpointing a distinct model at {step} steps") # Optionally store additional information data = dict(step=step, best_dev_acc=best_dev_acc) torch_save(FLAGS.checkpoint + "_" + str(step), data, models_dict, optimizers_dict, gpu=0 if FLAGS.cuda else -1) # Increment batch step step += 1 # Increment epoch epoch += 1 flogger.Log("Finished training.") """ Preset Model Configurations 1. Fixed - Fixed conversation length. 2. Adaptive - Adaptive conversation length using STOP bit. 3. FixedAttention - Fixed with Visual Attention. 4. AdaptiveAttention - Adaptive with Visual Attention. """ def Fixed(): FLAGS.img_feat = "avgpool_512" FLAGS.img_feat_dim = 512 FLAGS.fixed_exchange = True FLAGS.visual_attn = False def Adaptive(): FLAGS.img_feat = "avgpool_512" FLAGS.img_feat_dim = 512 FLAGS.fixed_exchange = False FLAGS.visual_attn = False def FixedAttention(): FLAGS.img_feat = "layer4_2" FLAGS.img_feat_dim = 512 FLAGS.fixed_exchange = True FLAGS.visual_attn = True FLAGS.attn_dim = 256 FLAGS.attn_extra_context = False FLAGS.attn_context_dim = 1000 def AdaptiveAttention(): FLAGS.img_feat = "layer4_2" FLAGS.img_feat_dim = 512 FLAGS.fixed_exchange = False FLAGS.visual_attn = True FLAGS.attn_dim = 256 FLAGS.attn_extra_context = True FLAGS.attn_context_dim = 1000 def flags(): # Debug settings gflags.DEFINE_string("branch", None, "") gflags.DEFINE_string("sha", None, "") gflags.DEFINE_boolean("debug", False, "") gflags.DEFINE_string("debug_log_level", 'INFO', "") # Convenience settings gflags.DEFINE_integer("save_after", 1000, "Min step (num batches) after which to save") gflags.DEFINE_integer( "save_interval", 1000, "How often to save after min batches have been reached") gflags.DEFINE_integer( "save_distinct_interval", 50000, "How often to save a distinct model after min batches have been reached") gflags.DEFINE_integer( "check_accuracy_interval", 25000, "How often to check if the accuracy has reached 75%") gflags.DEFINE_string("checkpoint", None, "Path to save data") gflags.DEFINE_string("conf_mat", None, "Path to save confusion matrix") gflags.DEFINE_string("log_path", "./logs", "Path to save logs") gflags.DEFINE_string("log_file", None, "") gflags.DEFINE_string("id_eval_csv_file", None, "Path to in domain eval log file") gflags.DEFINE_string("ood_eval_csv_file", None, "Path to out of domain eval log file") gflags.DEFINE_string( "json_file", None, "Where to store all flags for an experiment") gflags.DEFINE_string("log_load", None, "") gflags.DEFINE_boolean("eval_only", False, "") gflags.DEFINE_boolean("eval_xproduct", False, "Whether to evaluate the full cross product of agent pairs") gflags.DEFINE_boolean("eval_agent_communities", False, "Whether to evaluate on the community evaluation pairs") gflags.DEFINE_boolean("test_language_similarity", False, "Whether to test language similarity by changing trained agent messages, eval only option") gflags.DEFINE_boolean("self_similarity", False, "Whether the language similarity being evaluated is with itself") gflags.DEFINE_boolean("ancestor_similarity", False, "Whether the language similarity being evaluated is with an ancestor language") # Extract Settings gflags.DEFINE_boolean("binary_only", False, "Only extract binary data (no training)") gflags.DEFINE_string("binary_output", None, "Where to store binary data") # Performance settings gflags.DEFINE_boolean("cuda", False, "") # Display settings gflags.DEFINE_string("env", "main", "") gflags.DEFINE_boolean("visdom", False, "") gflags.DEFINE_string("experiment_name", None, "") gflags.DEFINE_integer("log_interval", 50, "") gflags.DEFINE_integer("log_dev", 1000, "") gflags.DEFINE_integer("log_self_com", 10000, "") gflags.DEFINE_integer("log_community_eval", 10000, "") gflags.DEFINE_boolean("report_on_complexity", False, "") # Data settings gflags.DEFINE_integer("wv_dim", 100, "Dimension of the word vectors") gflags.DEFINE_string("dataset", "shapeworld", "What type of dataset to use") gflags.DEFINE_string( "dataset_path", "./Shapeworld/data/oneshape_simple_textselect", "Root directory of the dataset") gflags.DEFINE_string( "dataset_indomain_valid_path", "./Shapeworld/data/oneshape_valid/oneshape_simple_textselect", "Root directory of the in domain validation dataset") gflags.DEFINE_string( "dataset_outdomain_valid_path", "./Shapeworld/data/oneshape_valid_all_combos/oneshape_simple_textselect", "Root directory of the in domain validation dataset") gflags.DEFINE_string("dataset_mode", "train", "") gflags.DEFINE_enum("dataset_eval_mode", "validation", ["validation", "test"], "") gflags.DEFINE_string("dataset_type", "agreement", "Task type") gflags.DEFINE_string("dataset_name", "oneshape_simple_textselect", "Name of dataset (should correspond to the root directory name automatically generated using ShapeWorld generate.py)") gflags.DEFINE_integer("dataset_size_train", 100, "How many examples to use") gflags.DEFINE_integer("dataset_size_dev", 100, "How many examples to use") gflags.DEFINE_string( "glove_path", "./glove.6B/glove.6B.100d.txt", "") gflags.DEFINE_boolean("shuffle_train", True, "") gflags.DEFINE_boolean("shuffle_dev", False, "") gflags.DEFINE_integer("random_seed", 7, "") gflags.DEFINE_enum( "resnet", "34", ["18", "34", "50", "101", "152"], "Specify Resnet variant.") gflags.DEFINE_boolean("improc_from_scratch", False, "Whether to train the image processor from scratch") gflags.DEFINE_integer("image_size", 128, "Width and height in pixels of the images to give to the agents") gflags.DEFINE_boolean("vertical_mask", False, "Whether to just use a vertical mask on images. Otherwise the mask is random") # Model settings gflags.DEFINE_enum("model_type", None, [ "Fixed", "Adaptive", "FixedAttention", "AdaptiveAttention"], "Preset model configurations.") gflags.DEFINE_enum("img_feat", "avgpool_512", [ "layer4_2", "avgpool_512", "fc"], "Specify which layer output to use as image") gflags.DEFINE_enum("data_context", "fc", [ "fc"], "Specify which layer output to use as context for attention") gflags.DEFINE_integer("img_feat_dim", 512, "Dimension of the image features") gflags.DEFINE_integer( "h_dim", 100, "Hidden dimension for all hidden representations in the network") gflags.DEFINE_integer("m_dim", 64, "Dimension of the messages") gflags.DEFINE_integer( "desc_dim", 100, "Dimension of the input description vectors") gflags.DEFINE_integer( "num_classes", 10, "How many texts the agents have to choose from") gflags.DEFINE_integer("s_dim", 1, "Stop probability output dim") gflags.DEFINE_boolean("use_binary", True, "Encoding whether agents uses binary features") gflags.DEFINE_boolean("no_comms_channel", False, "Whether to mute the communications channel") gflags.DEFINE_list("agent_dicts", ['None', 'None'], "list of paths to average message code dictionaries") gflags.DEFINE_list("agent_supplementary_dicts", ['None', 'None'], "list of paths to extra average message code dictionaries. Used for testing language similarity between an existing and ancestor community") gflags.DEFINE_boolean("randomize_comms", False, "Whether to randomize the order in which agents communicate") gflags.DEFINE_string("reward_type", "cooperative", "String describing the reward type to train the message channel") gflags.DEFINE_boolean("agent_pools", False, "Whether to have a pool of agents to train instead of two fixed agents") gflags.DEFINE_integer("num_agents", 2, "How many agents total (single pool or community)") gflags.DEFINE_integer("num_supplementary_agents", 2, "How many agents total in an additional community. Used for testing language similarity between an existing and ancestor community.") gflags.DEFINE_list("compare_language_pairs", [0, 1], "Index of the two groups in a community to compare") gflags.DEFINE_boolean("with_replacement", False, "Whether to sample agents from pool with replacement") gflags.DEFINE_boolean("agent_communities", False, "Whether to have a community of agents to train instead of two fixed agents") gflags.DEFINE_integer("num_communities", 2, "How many communities of agents") gflags.DEFINE_string("community_type", "dense", "Type of agent community: dense or chain") gflags.DEFINE_list("community_checkpoints", ['None', 'None'], "list of checkpoints per community") gflags.DEFINE_boolean("gen_community_messages", False, "") gflags.DEFINE_string("community_structure", "55555", "String listing the community structure") gflags.DEFINE_list("num_agents_per_community", [5, 5], "Number of agents per community. Specify one per community, can be different") gflags.DEFINE_list("intra_pool_connect_p", [1.0, 1.0], "Percentage of intra pool connectivity. Specify one value per community, can be different") gflags.DEFINE_float("inter_pool_connect_p", 0.1, "Percentage of inter pool connectvity. Specify one value") gflags.DEFINE_integer("ratio_adjust", 1, "Whether to adjust the inter:intra training ratio") gflags.DEFINE_float("intra_inter_ratio", 1.0, "Ratio of intra to inter pool training. Default is 1.0. Agents across pools and within pools will be trained equally.") gflags.DEFINE_float("first_msg", 0, "Value to fill the first message with") gflags.DEFINE_boolean("visual_attn", False, "agents attends over image") gflags.DEFINE_boolean( "use_MLP", False, "use MLP to generate prediction scores") gflags.DEFINE_integer("attn_dim", 256, "") gflags.DEFINE_boolean("attn_extra_context", False, "") gflags.DEFINE_integer("attn_context_dim", 4096, "") gflags.DEFINE_float("dropout", 0.3, "How much dropout to apply when training using the original image") gflags.DEFINE_integer("top_k_dev", 3, "Top-k error in development") gflags.DEFINE_integer("top_k_train", 3, "Top-k error in training") # Optimization settings gflags.DEFINE_enum("optim_type", "RMSprop", ["Adam", "SGD", "RMSprop"], "") gflags.DEFINE_integer("batch_size", 32, "Minibatch size for train set.") gflags.DEFINE_integer("batch_size_dev", 50, "Minibatch size for dev set.") gflags.DEFINE_float("learning_rate", 1e-4, "Used in optimizer.") gflags.DEFINE_integer("max_epoch", 500, "") gflags.DEFINE_float("entropy_s", None, "") gflags.DEFINE_float("entropy_agent1", None, "") gflags.DEFINE_float("entropy_agent2", None, "") gflags.DEFINE_float("nll_loss_weight_nc", 1.0, "") gflags.DEFINE_float("nll_loss_weight_com", 1.0, "") gflags.DEFINE_float("rl_loss_weight", 1.0, "") gflags.DEFINE_float("baseline_loss_weight", 1.0, "") # Conversation settings gflags.DEFINE_integer("max_exchange", 1, "") gflags.DEFINE_boolean("fixed_exchange", True, "") gflags.DEFINE_boolean( "bit_flip", False, "Whether sender's messages are corrupted.") gflags.DEFINE_string("corrupt_region", None, "Comma-separated ranges of bit indexes (e.g. ``0:3,5'').") def default_flags(): if FLAGS.log_load: log_flags = json.loads(open(FLAGS.log_load).read()) for k in log_flags.keys(): if k in FLAGS.FlagValuesDict().keys(): setattr(FLAGS, k, log_flags[k]) FLAGS(sys.argv) # Optionally override predefined flags. if FLAGS.model_type: eval(FLAGS.model_type)() FLAGS(sys.argv) # Optionally override predefined flags. if not FLAGS.use_binary: FLAGS.exchange_samples = 0 if not FLAGS.experiment_name: timestamp = str(int(time.time())) FLAGS.experiment_name = "{}-so_{}-wv_{}-bs_{}-{}".format( FLAGS.dataset, FLAGS.m_dim, FLAGS.wv_dim, FLAGS.batch_size, timestamp, ) if not FLAGS.conf_mat: FLAGS.conf_mat = os.path.join( FLAGS.log_path, FLAGS.experiment_name + ".conf_mat.txt") if not FLAGS.log_file: FLAGS.log_file = os.path.join( FLAGS.log_path, FLAGS.experiment_name + ".log") if not FLAGS.id_eval_csv_file: FLAGS.id_eval_csv_file = os.path.join( FLAGS.log_path, FLAGS.experiment_name + ".id_eval.csv") if not FLAGS.ood_eval_csv_file: FLAGS.ood_eval_csv_file = os.path.join( FLAGS.log_path, FLAGS.experiment_name + ".ood_eval.csv") if not FLAGS.json_file: FLAGS.json_file = os.path.join( FLAGS.log_path, FLAGS.experiment_name + ".json") if not FLAGS.checkpoint: FLAGS.checkpoint = os.path.join( FLAGS.log_path, FLAGS.experiment_name + ".pt") if not FLAGS.binary_output: FLAGS.binary_output = os.path.join( FLAGS.log_path, FLAGS.experiment_name + ".bv.hdf5") if not FLAGS.branch: FLAGS.branch = os.popen( 'git rev-parse --abbrev-ref HEAD').read().strip() if not FLAGS.sha: FLAGS.sha = os.popen('git rev-parse HEAD').read().strip() if not torch.cuda.is_available(): FLAGS.cuda = False if FLAGS.debug: np.seterr(all='raise') # silly expanduser FLAGS.glove_path = os.path.expanduser(FLAGS.glove_path) if __name__ == '__main__': flags() FLAGS(sys.argv) default_flags() print(sys.argv) FORMAT = '[%(asctime)s %(levelname)s] %(message)s' logging.basicConfig(format=FORMAT) debuglogger = logging.getLogger('main_logger') debuglogger.setLevel(FLAGS.debug_log_level) if FLAGS.random_seed != -1: random.seed(FLAGS.random_seed) np.random.seed(FLAGS.random_seed) torch.manual_seed(FLAGS.random_seed) else: random.seed() np.random.seed() run()
dls-controls/scanpointgenerator
scanpointgenerator/rois/__init__.py
<filename>scanpointgenerator/rois/__init__.py ### # Copyright (c) 2016, 2017 Diamond Light Source Ltd. # # Contributors: # <NAME> - initial API and implementation and/or initial documentation # <NAME> - initial API and implementation and/or initial documentation # <NAME> - initial API and implementation and/or initial documentation # ### from scanpointgenerator.rois.circular_roi import CircularROI from scanpointgenerator.rois.elliptical_roi import EllipticalROI from scanpointgenerator.rois.linear_roi import LinearROI from scanpointgenerator.rois.point_roi import PointROI from scanpointgenerator.rois.polygonal_roi import PolygonalROI from scanpointgenerator.rois.rectangular_roi import RectangularROI from scanpointgenerator.rois.sector_roi import SectorROI