File size: 8,232 Bytes
484b847 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 |
"""
Tests for fluid dynamics residual computations.
Compares torch implementations against numpy reference implementations.
Can be run as:
1. Pytest test: pytest tests/test_fluids.py (skips if data not found)
2. Standalone script: python tests/test_fluids.py --filename <path_to_data>
"""
import argparse
import os
import sys
from pathlib import Path
import h5py
import numpy as np
import pytest
import torch
from loguru import logger
from pdeinvbench.losses.fluids import (
compute_advection,
compute_stream_function,
laplacian,
turbulent_flow_residual,
)
def find_turbulent_flow_data():
"""Try to find turbulent flow data in common locations."""
possible_paths = [
Path("../data/2D_turbulent-flow_nu=0.006153085601625313.h5"),
Path("data/2D_turbulent-flow_nu=0.006153085601625313.h5"),
Path("/data/shared/meta-pde/turbulent-flow-2d/train").glob("*.h5"),
]
for path in possible_paths:
if isinstance(path, Path) and path.exists():
return str(path)
# Handle glob results
try:
for file in path:
if file.exists():
return str(file)
except (TypeError, AttributeError):
pass
return None
@pytest.fixture
def turbulent_flow_datafile():
"""Fixture that provides path to test data."""
data_path = find_turbulent_flow_data()
if data_path is None:
pytest.skip(
"Turbulent flow test data not found. "
"This test requires real PDE data and is skipped in CI/CD. "
"Run manually with: python tests/test_fluids.py --filename <path>"
)
return data_path
def wrapper(func):
"""Convert torch tensor outputs to numpy for comparison."""
def _wrapper(*args):
# Convert numpy to torch
new_args = [
torch.from_numpy(a).float() if isinstance(a, np.ndarray) else a
for a in args
]
out = func(*new_args)
return out.cpu().numpy()
return _wrapper
def _maybe_unsqueeze_np(u):
"""Ensure last dimension exists for channel."""
return u if u.shape[-1] == 1 else np.expand_dims(u, axis=-1)
def compare_funcs(f1, f2):
"""Compare outputs of two functions and log the difference."""
def compare(f1args, f2args):
reference = _maybe_unsqueeze_np(f1(*f1args))
newout = wrapper(f2)(*f2args)
diff = np.linalg.norm(reference - newout)
logger.info(f"Diff between {f1.__name__} and {f2.__name__}: {diff:.2e}")
# Assert reasonable accuracy
assert diff < 1e-3, f"Difference too large: {diff:.2e}"
return diff
return compare
def run_fluids_comparison(filename: str):
"""
Run the fluids residual computation comparison.
Args:
filename: Path to turbulent flow HDF5 data file
"""
# Import numpy reference implementations
from fluids_numpy_reference import (
advection as advection_np_base,
compute_stream_function as compute_stream_function_np,
laplacian as laplacian_np,
tf_residual_numpy,
)
logger.info(f"Loading data from: {filename}")
# Load data
try:
dataset = h5py.File(filename, "r")
traj_idx = "0000"
data = dataset[traj_idx]["data"][:]
logger.info(f"Data shape: {data.shape}")
t = dataset[traj_idx]["grid/t"][:]
x = dataset[traj_idx]["grid/x"][:]
y = dataset[traj_idx]["grid/y"][:]
dataset.close()
except (KeyError, FileNotFoundError) as e:
logger.error(f"Failed to load data: {e}")
raise
# Extract parameters
nu = (
os.path.basename(filename)
.split("=")[-1]
.replace(".h5", "")
.replace(".hdf5", "")
)
nu: float = float(nu)
logger.info(f"Viscosity parameter nu: {nu}")
dx = x[1] - x[0]
dy = y[1] - y[0]
# Compute residual norm as sanity check
residual_norm = np.linalg.norm(wrapper(turbulent_flow_residual)(data, t, x, y, nu))
logger.info(f"Computed residual norm: {residual_norm:.2e}")
# Test 1: Stream function computation (Fourier space)
logger.info("\n=== Testing stream function (Fourier=False) ===")
compute_stream_function_np.__name__ = "compute_stream_function_np, fourier=False"
compare_funcs(compute_stream_function_np, compute_stream_function)(
(data, x, y, False), (data, dx, dy, False)
)
# Test 2: Stream function computation (physical space)
logger.info("\n=== Testing stream function (Fourier=True) ===")
compute_stream_function_np.__name__ = "compute_stream_function_np, fourier=True"
compare_funcs(compute_stream_function_np, compute_stream_function)(
(data, x, y, True), (data, dx, dy, True)
)
# Test 3: Advection term
logger.info("\n=== Testing advection ===")
def advection_np(u, x, y):
return advection_np_base(u, x, y, stream_func=compute_stream_function_np)[0]
advection_np.__name__ = "advection_np"
compare_funcs(advection_np, compute_advection)((data, x, y), (data, dx, dy))
# Test 4: Laplacian
logger.info("\n=== Testing laplacian ===")
laplacian_np.__name__ = "laplacian_np"
compare_funcs(laplacian_np, laplacian)((data, x, y), (data, dx, dy))
# Test 5: Velocity component vx
logger.info("\n=== Testing velocity vx ===")
def advection_np_vx(u, x, y):
return advection_np_base(u, x, y, stream_func=compute_stream_function_np)[1]
advection_np_vx.__name__ = "advection_np for vx"
compare_funcs(
advection_np_vx, lambda *args: compute_advection(*args, return_velocity=True)[1]
)((data, x, y), (data, dx, dy))
# Test 6: Velocity component vy
logger.info("\n=== Testing velocity vy ===")
def advection_np_vy(u, x, y):
return advection_np_base(u, x, y, stream_func=compute_stream_function_np)[2]
advection_np_vy.__name__ = "advection_np for vy"
compare_funcs(
advection_np_vy, lambda *args: compute_advection(*args, return_velocity=True)[2]
)((data, x, y), (data, dx, dy))
# Test 7: Full turbulent flow residual
logger.info("\n=== Testing full turbulent flow residual ===")
compare_funcs(tf_residual_numpy, turbulent_flow_residual)(
(data, t, x, y, nu), (data, t, x, y, nu)
)
logger.success("\n✓ All fluids tests passed!")
@pytest.mark.slow
@pytest.mark.data_required
def test_fluids_residual(turbulent_flow_datafile):
"""
Pytest wrapper for fluids residual comparison.
Automatically skipped if test data is not available.
Run with: pytest tests/test_fluids.py -v -m data_required
"""
run_fluids_comparison(turbulent_flow_datafile)
if __name__ == "__main__":
# Allow running as standalone script with argparse
parser = argparse.ArgumentParser(
description="Validate torch turbulent flow residual computations against numpy reference",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog="""
Examples:
python tests/test_fluids.py --filename data/2D_turbulent-flow_nu=0.006153085601625313.h5
python tests/test_fluids.py --auto-find
""",
)
parser.add_argument(
"--filename", type=str, help="Path to turbulent flow HDF5 data file"
)
parser.add_argument(
"--auto-find",
action="store_true",
help="Automatically search for data file in common locations",
)
args = parser.parse_args()
if args.auto_find or args.filename is None:
logger.info("Searching for turbulent flow data...")
filename = find_turbulent_flow_data()
if filename is None:
logger.error(
"Could not find turbulent flow data. "
"Please specify path with --filename"
)
sys.exit(1)
logger.info(f"Found data at: {filename}")
else:
filename = args.filename
if not Path(filename).exists():
logger.error(f"File not found: {filename}")
sys.exit(1)
try:
run_fluids_comparison(filename)
except Exception as e:
logger.error(f"Test failed with error: {e}")
import traceback
traceback.print_exc()
sys.exit(1)
|