Expose LayerFactory::LayerTypeList in pycaffe

[lukeyeager]

I want to be able to validate NetParameters with pycaffe without waiting for Caffe to crash on SIGABRT.

F0814 16:58:06.833762 31396 layer_factory.hpp:80] Check failed: registry.count(type) == 1 (0 vs. 1) Unknown layer type: not-a-layer-type (known types: AbsVal, Accuracy, ArgMax, BNLL, Concat, ContrastiveLoss, Convolution, Data, Deconvolution, Dropout, DummyData, Eltwise, EuclideanLoss, Exp, Filter, Flatten, HDF5Data, HDF5Output, HingeLoss, Im2col, ImageData, InfogainLoss, InnerProduct, LRN, Log, MVN, MemoryData, MultinomialLogisticLoss, PReLU, Pooling, Power, Python, ReLU, Reduction, Reshape, SPP, Sigmoid, SigmoidCrossEntropyLoss, Silence, Slice, Softmax, SoftmaxWithLoss, Split, TanH, Threshold, WindowData)
*** Check failure stack trace: ***
Aborted (core dumped)

Exposing LayerTypeList in pycaffe lets me check the layer types and handle errors in a reasonable way.

Is there any way to catch all CHECK/DCHECK errors with pycaffe and turn them into Python exceptions?

[ronghanghu]

Look good to me :)

[jeffdonahue]

this should use self.assertTrue(...) instead of assert ... (edit: apparently assertIn is a thing too, should use that)

[jeffdonahue]

LGTM besides above comment.

[lukeyeager]

Thanks for the review. Changed and squashed. Why is unittest.assertIn() preferable to assert?

Also, can you address this question:

Is there any way to catch all CHECK/DCHECK errors with pycaffe and turn them into Python exceptions?

That would be a huge help.

[jeffdonahue]

Thanks for the update! The unittest.assert* functions will make the test pass/fail appropriately according to their result and print a nicely formatted error message when the assertion fails, where assert will just raise an Exception if its argument is false. (Maybe the exception would also make the test fail, but presumably with a less nice error message.)

Edit: more info here -- apparently a plain AssertionError (as raised by a failing assert) was the old way to signal test failure and will work fine for now, but it may not be supported in the future.

Is there any way to catch all CHECK/DCHECK errors with pycaffe and turn them into Python exceptions?

I think #1683 tried to do this