search

tensorflow / lucid

A collection of infrastructure and tools for research in neural network interpretability.
Apache License 2.0
4.65k stars 655 forks source link

can't use finetuned InceptionV1 #217

Open virilo opened 4 years ago

virilo commented 4 years ago

I finetuned a InceptionV1 with flowers17 dataset.

I'm not very sure about how to load the finetuned model to be compatible with Activation Atlas

I've read in https://github.com/tensorflow/lucid/wiki/Importing-Models-into-Lucid that is should be loaded with 'load' method

But I've seen that the Model.load method isn't neither in version 0.3.8 nor 0.3.9-alpha. It seems to be something new, and it's present only in current master (2019/12/16)

I shared the Activation Atlas code with my trained InceptionV1 here:

https://colab.research.google.com/drive/1ngs5lmxzbacx48QXWpsVkqd9ULdrEtr7

It's failing, giving this error: 

TypeError: Tensors in list passed to 'values' of 'ConcatV2' Op have types [float32, float32, float32, int32] that don't all match.

Cell [7] contains the model loading code

TIA