training.py _check_loss_and_target_compatibility() fix crash if y is None

[ahundt]

Don't access y param of model.fit() if it is None. This prevents a crash when model.fit(None,None, ...) is called and loss is provided to model.compile().

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 30 days if no further activity occurs, but feel free to re-open a closed issue if needed.

[ahundt]

This one is still valid

[ragulpr]

+1 for this pull, it seems necessary for various hacks/workarounds for target/predicted compatibility checks and in order to reuse inputs in loss (ex autoencoders w. embedding layers)

Great work! Hope it's merged

[ahundt]

@fchollet I know we passed on most of my other PRs once the target tensor parameter was added, but this one is trivially simple and worth merging.

[fchollet]

Sorry for not checking back on this earlier.

[ahundt]

thanks!