This issue is for the potential merging of tfa.callbacks.tqdm_progress_bar and tqdm.keras to reduce code duplication. I'll follow up with @casperdcl, the code owner and maintainer of tqdm, https://github.com/tqdm/tqdm, to see what's the best way forward for this effort.
tqdm.keras:
https://github.com/tqdm/tqdm/blob/master/tqdm/keras.py
tfa.callbacks.tqdm_progress_bar:
https://github.com/tensorflow/addons/blob/master/tensorflow_addons/callbacks/tqdm_progress_bar.py
tfa.callback.tqdm_progress_bar tutorial:
https://github.com/tensorflow/addons/blob/master/docs/tutorials/tqdm_progress_bar.ipynb
fyi from https://github.com/tensorflow/addons/pull/916#issuecomment-576937849:
Anyone looked at https://github.com/tqdm/tqdm/#keras-integration? [...] looks like it was developed at the same time as the tensorflow addon. Seems silly to have it in two places.
I'd suggest opening a PR on https://github.com/tqdm/tqdm/ to change anything that may be required and then changing https://github.com/tensorflow/addons/blob/master/tensorflow_addons/callbacks/tqdm_progress_bar.py to just import from tqdm...
cc @shun-lin @gabrieldemarmiesse @seanpmorgan
and from https://github.com/tensorflow/addons/pull/916#issuecomment-576948044:
[...] referring to future development/PRs to ease the burden of maintenance. Right now there's
[tfa].callbacks.tqdm_progress_barbeing developed in tandem withtqdm.keraseven though they are broadly identical. I was suggesting a way to avoid code/logic duplication between them.
also found https://github.com/bstriner/keras-tqdm from @bstriner to consider.
Overall there's
tfa.callbacks.tqdm_progress_bar]tqdm.keras]tqdm usage but no idea about tf/keras correctness. Nobody reviewed it.keras_tqdm]I still would prefer if people open PRs on the main repo ([tqdm.keras]) to make external wrapping easier, and then have very thin wrappers in [tfa.callbacks.tqdm_progress_bar], [keras_tqdm], and even [keras.callbacks].
I would be in favor of a small wrapper in Addons which utilizes a central tqdm.keras
I think keras_tqdm is no longer maintained (last commit in Oct 2017), I will look into how we can merge tqdm.keras and the one we have here in tfa to make better sense in terms of maintenance :)
Most helpful comment
I think keras_tqdm is no longer maintained (last commit in Oct 2017), I will look into how we can merge
tqdm.kerasand the one we have here in tfa to make better sense in terms of maintenance :)