New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
SqeezeNet meta data for create_cnn function #1152
Conversation
Thanks! |
Please amend your PR to be able to merge with the latest big refactor. If it's easiest, close this one and reopen another. |
Ok, sure, will do. |
@sgugger Removed all training from the tests except one place with |
Seems there is a bug in tests, going to fix soon. |
@sgugger Seems like there is a timeout issue, however, I don't train any model in the tests I've added. Do you think it is worth to change something in the tests to decrease CI execution time? |
I've tried to address the issues. I think now the @pytest.mark.slow
@pytest.mark.parametrize('arch', [models.resnet18, models.squeezenet1_1])
def test_models_meta(mnist_tiny, arch, zero_image):
learn = create_cnn(mnist_tiny, arch, metrics=[accuracy, error_rate])
pred = learn.predict(zero_image)
assert pred is not None Note the last line. I am not really sure what to assert here so I am just making sure that the output is not |
Seems good, thanks! |
Based on forum's discussion, the PR makes an attempt to bring the support of SqeezeNet architecture into the library. Now you're able to use the architecture like ResNets:
However, I've chosen the layers groups split quite arbitrary (splitting model on maxpool layers) so would be glad to see your comments if there is a better configuration.