Feature fixed weights hist handler issue 2328

[sadra-barikbin]

Fixes #2328

Description:

• Added whitelist arg to WeightsHistHandler to be able to log specific weights to TensorboardLogger.

Check list:

• New tests are added (if a new feature is added)
• New doc strings: description and/or example code are in RST format
• Documentation is updated (if required)

[vfdev-5]

@sadra-barikbin thanks for the PR, the issue was to improve current WeightsHistHandler without introducing another class.

[sadra-barikbin]

Oh. So I add a whitelist parameter to WeightsHistHandler. If it was None, I log all of weights, otherwise just specified weights. Is this OK?

[vfdev-5]

Thanks for the PR @sadra-barikbin
I left few comments on to improve the code. Let me know what do you think

[vfdev-5]

@sadra-barikbin thanks for the updates ! Can you now run this example with specifying whitelist=["conv", ] here:

ignite/examples/contrib/mnist/mnist_with_tensorboard_logger.py

Line 136 in 82a929d

tb_logger.attach(trainer, log_handler=WeightsScalarHandler(model), event_name=Events.ITERATION_COMPLETED(every=100))

and report here the result ?

[sdesrozis]

@sadra-barikbin Thanks very much, LGTM ! Although a comment still to be addressed.

[sadra-barikbin]

Hi, I ran the example. Result is depicted in picture below:

By the way, there was mistakes in lines 138 and 142 in event_name parameter.

[vfdev-5]

Thanks @sadra-barikbin !

[vfdev-5]

@sadra-barikbin can you add please a similar arg to other weights and grad logging handlers ?

[sadra-barikbin]

Sure. Shall I open a new issue for them? You mean just Tensorboard ones?

[vfdev-5]

yes, please open an issue and suggest the classes where we could add such new argument.
By the way, I think we forgot about docs tag for new args in this PR: .. versionchanged:: (check the codebase for examples of usage)
Please send a follow-up update. Thanks