text
stringlengths
1
93.6k
class Classifier():
def __init__(self, params):
self.classifier = params.classifier
self.k_shot = params.k_shot
def predict(self, data_result):
# data_result: dict
# classifier_type:
# 1. protoNet
# 2. SVM
# 3. KNN
# 4. logistic regression
# 5. classifier NN
if (self.classifier == 'protonet'):
predicted_y,loss = one_shot_classifier_prototype_lowerdim(data_result)
elif (self.classifier == 'SVM'):
classifier_SVM = SVC(C=10)
classifier_SVM.fit(data_result['support_feature'], data_result['support_y'])
predicted_y = classifier_SVM.predict(data_result['query_feature'])
elif (self.classifier == 'LR'):
classifier_LR = LogisticRegression()
classifier_LR.fit(data_result['support_feature'], data_result['support_y'])
predicted_y = classifier_LR.predict(data_result['query_feature'])
elif (self.classifier == 'KNN'):
classifier_KNN = KNeighborsClassifier(n_neighbors= self.k_shot)
classifier_KNN.fit(data_result['support_feature'],data_result['support_y'])
predicted_y = classifier_KNN.predict(data_result['query_feature'])
elif(self.classifier == 'cosine'):
distance_cosine_tensor = cosine_dist(data_result['query_feature'],data_result['support_feature'])
#print('tensor cosine dist:', distance_cosine_tensor.size())
#distance_cosine = cosine_similarity(data_result['query_feature'].detach().cpu().numpy(),data_result['support_feature'].detach().cpu().numpy())
#print('numpyt cosine dist:', distance_cosine)
probability = torch.nn.functional.softmax(distance_cosine_tensor, dim=-1)
#print('tensor cosine dist:', distance_cosine_tensor)
query_y = data_result['query_y']
import torch.nn as nn
criterion = nn.CrossEntropyLoss()
probability = probability.unsqueeze(0)
label= query_y.long()
#print('pro:', probability, 'label:',label)
loss = criterion(probability, label)
probability = probability.data.cpu().numpy()
predicted_y = np.argmax(probability, axis=1)
#predicted_y = np.argsort(-distance_cosine)
#predicted_y = predicted_y[:,0]
else:
print('classifier type error.')
return predicted_y,loss
if __name__=='__name__':
myClassifier=classifier()
# <FILESEP>
"""
___ __ _
/ __\ ___ _ __ / _|(_) __ _
/ / / _ \ | '_ \ | |_ | | / _` |
/ /___ | (_) || | | || _|| || (_| |
\____/ \___/ |_| |_||_| |_| \__, |
|___/
"""
# Twitter API credentials. If you need any help have a look at README.md
twitter_credentials = {
"consumer_key": '',
"consumer_secret": '',
"access_token": '',
"access_secret": '',
}
# DON'T WRITE ANYTHING IN CAPS, AS THE BOT AUTOMATICALLY FLATTERS ALL INPUT TEXTS. THUS ANY WORD WITH CAPS WON'T BE RECOGNIZED
# Tags that Twitter will use to look up our tweets. Really important as all the script will be based on them
search_tags = ["giveaway", "contest", "sorteo", "to win"]
# Don't start the bot if friends weren't correctly retrieved
wait_retrieve = False
# Enable this if you want the bot to send a DM in case it detects any message_tags
use_msgs = False
#Ignore tweets that contain any of these words
BadList = ["throat","bone","naked","selfie","photo","onlyfans","nude","+18","femdom","fendom","whatsapp","sex","xxx","daddy","mommy","sugar","vid","#imgxnct","pic","tits" ,"booty" ,"boob", "freenude","cum", "dick" ,"gay" ,"onlyfans.com","hot", "ass", "fuck", "suck", "cock", "lick","pussy" ]
# What words will the bot check in order to retweet a tweet. It's important because if the bot doesnt
# recognize any, it will skip the whole tweet and it wont check if it has to like, msg, or follow
retweet_tags = ["retweet", "retweetea", "retwitea", "rt"]
# What words will trigger to send the author a DM with a random message_text
message_tags = ["message", "dm"]
# What words will the bot check in order to follow the author of the tweet plus all the users mentioned in the text
# (we assume that a retweet tag was recognized)
follow_tags = ["follow", "fl", "sigue", "seguir", "siguenos"]
# What words will the bot look for in order to like a tweet (it also needs to contain a retweet tag)
like_tags = ["like", "fav", "favorite"]
# These are supposed to be random msgs the bot would send if DMing is required
message_text = ["I want to enter to the giveaway!", "Hope to win :D"]
# Add to this list all the users whose contests (actually tweets that contain retweet_tags keywords) the script will