A Tiny Test Suite for pytorch based Machine Learning models, inspired by mltest. Chase Roberts lists out 4 basic tests in his medium post about mltest. torchtest is mostly a pytorch port of mltest(which was written for tensorflow).
# imports for examples
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
from torchtest import assert_vars_change
inputs = Variable(torch.randn(20, 20))
targets = Variable(torch.randint(0, 2, (20,))).long()
batch = [inputs, targets]
model = nn.Linear(20, 2)
# what are the variables?
print('Our list of parameters', [ np[0] for np in model.named_parameters() ])
# do they change after a training step?
# let's run a train step and see
assert_vars_change(
model=model,
loss_fn=F.cross_entropy,
optim=torch.optim.Adam(model.parameters()),
batch=batch)
""" FAILURE """
# let's try to break this, so the test fails
params_to_train = [ np[1] for np in model.named_parameters() if np[0] is not 'bias' ]
# run test now
assert_vars_change(
model=model,
loss_fn=F.cross_entropy,
optim=torch.optim.Adam(params_to_train),
batch=batch)
# YES! bias did not change
from torchtest import assert_vars_same
# What if bias is not supposed to change, by design?
# test to see if bias remains the same after training
assert_vars_same(
model=model,
loss_fn=F.cross_entropy,
optim=torch.optim.Adam(params_to_train),
batch=batch,
params=[('bias', model.bias)]
)
# it does? good. let's move on
from torchtest import test_suite
# NOTE : bias is fixed (not trainable)
optim = torch.optim.Adam(params_to_train)
loss_fn=F.cross_entropy
test_suite(model, loss_fn, optim, batch,
output_range=(-2, 2),
test_output_range=True
)
# seems to work
""" FAILURE """
# let's tweak the model to fail the test
model.bias = nn.Parameter(2 + torch.randn(2, ))
test_suite(
model,
loss_fn, optim, batch,
output_range=(-1, 1),
test_output_range=True
)
# as expected, it fails; yay!
""" FAILURE """
model.bias = nn.Parameter(float('NaN') * torch.randn(2, ))
test_suite(
model,
loss_fn, optim, batch,
test_nan_vals=True
)
torchtest\torchtest.py", line 151, in _var_change_helper
assert not torch.equal(p0, p1)
RuntimeError: Expected object of backend CPU but got backend CUDA for argument #2 'other'
When you are making use of a GPU, you should explicitly specify device=cuda:0
. By default device
is set to cpu
. See issue #1 for more information.
test_suite(
model, # a model moved to GPU
loss_fn, optim, batch,
test_inf_vals=True,
device='cuda:0'
)