search

sicara / tf-explain

Interpretability Methods for tf.keras models with Tensorflow 2.x
https://tf-explain.readthedocs.io
MIT License
1.02k stars 112 forks source link

saved files are .v2 files #171

Closed lilianmarey closed 3 years ago

lilianmarey commented 3 years ago

Hi !

I was trying to use tf-explain on my pretrained neural nets, it seems it worked, however, after running my code, the saved files are ".v2" files. I really don't know how to open this type of files, or how to convert it into .png for example.

I wanted to know if getting this type of files was normal, and if it is, how to open it.

Thanks anyway for sharing the library

RaphaelMeudec commented 3 years ago

Those are probably tensorboard files. Can you provide which methods/callbacks you're using?

lilianmarey commented 3 years ago

I'm using OcclusionSensitivityCallback and IntegratedGradientsCallback, I had .v2 files for both. Here is also the architecture of my net and the code I ran Capture d’écran 2021-07-07 à 15 56 47 Capture d’écran 2021-07-07 à 15 56 40

RaphaelMeudec commented 3 years ago

All callbacks are made to draw images directly into tensorboard. You have to open a tensorboard with tensorboard --logdir path_to_logs and go to the images tab

lilianmarey commented 3 years ago

Ok, it worked, thanks a lot !