convert_model_from_saved_files can not work for Inception V3 and other models

[linchundan88]

Deeplift is a great tool and "Learning Important Features Through Propagating Activation Differences" is a great paper.
However I can not use it for inception v3 and other models(I know deeplift do not implement resnet so far)

File "/home/ubuntu/.conda/envs/dlp/lib/python3.6/site-packages/deeplift/conversion/kerasapi_conversion.py", line 122, in conv2d_conversion
bias=config[KerasKeys.weights][1],
IndexError: list index out of range

[linchundan88]

when using convert_model_from_saved_files(), all kinds of NonlinearMxtsMode parameters.

[AvantiShri]

Thanks for pointing this out...looks like Inception V3 doesn't have bias terms for some layers, which throws the conversion off. I can work on a fix.

[AvantiShri]

Hi @linchundan88, I've implemented the changes in this branch #76 and the conversion works, but for some reason the tests on Travis are stalling. Chances are it's something small like running out of memory, but it could take some time for me to find the bandwidth to debug it. In the meantime, you can look into the alternative implementations described in the FAQ under "My model architecture is not supported by this DeepLIFT implementation. What should I do?" - hope this satisfies your needs!

[linchundan88]

Thanks a lot. I knew DeepExplain and SHAP. I am interested in RevealCancel^_^.