Passing initial_epoch parameter to callbacks' self.params in tf.keras.model.fit #39839
Labels
comp:keras
Keras related issues
stat:awaiting tensorflower
Status - Awaiting response from tensorflower
type:feature
Feature requests
Please make sure that this is a feature request. As per our GitHub Policy, we only address code/doc bugs, performance issues, feature requests and build/installation issues on GitHub. tag:feature_template
System information
Describe the feature and the current behavior/state.
TensorFlow Addons are developing a progress bar (TQDM Progress bar) and we receive and issue where when user specify
initialEpoch > 0
duringmodel.fit
and when usingtfa.callbacks.TQDMProgressBar
the progress bar will never reach 100% because inTQDMProgressBar
code we never consider the case when user setinitialEpoch
model.fit(x=X, y=y, class_weight=None, batch_size=batchSize, verbose=0, callbacks=tfa.callbacks.TQDMProgressBar(), validation_split=0.2, shuffle=True, epochs=epochCount, initial_epoch=initialEpoch)
tensorflow/addons#1748
While searching for a solution, I noticed that
initial_epoch
is never passed to Callback'sself.params
dictionary and thus making it hard to implement the feature where users set an initial epoch, thus I am asking if it would make sense for TensorFlow to passinitial_epoch
to Callback'sself.params
for us to implement that feature. The other way around would be to ask user to specifyinitial_epoch
again in the progress bar but that would not be ideal. Thank you so much for your time!Will this change the current api? How?
No, but this will add another key into Callbacks'
self.params
dictionary.Who will benefit with this feature?
Users of TensorFlow Addons TQDM Progress Bar and thus that may need
initial_epoch
in their custom callbacks.Any Other info.
TensorFlow Addons TQDM Progress Bar Source code:
https://github.com/tensorflow/addons/blob/master/tensorflow_addons/callbacks/tqdm_progress_bar.py
The text was updated successfully, but these errors were encountered: