model.fit() should tell you if batch is not big enough to train with

[isaacgerg]

Suppose you train a model using model.fit() and your batch size is set to N. If the input from a generator is < N, tf keras gives the following error which is difficult to interpret: ValueError: Expect x to be a non-empty array or dataset.

Reference second answer in : https://stackoverflow.com/questions/63231811/valueerror-expect-x-to-be-a-non-empty-array-or-dataset-tensor-flow-lite-model

Propose to have tf keras tell user that batch size is larger than data size in place of current error. The current error implies that the x is empty which isnt this case for all x with size less than N.

[tilakrayal]

@isaacgerg ,

In order to expedite the trouble-shooting process, could you please provide the Tensorflow version,complete code and dataset to reproduce the issue reported here.

Thanks!

[isaacgerg]

Tensorflow 2.4.1

[tilakrayal]

@isaacgerg ,

Could you please provide the complete code and dataset or colab link to reproduce the issue reported here.

[isaacgerg]

@tilakrayal See the stack overflow link in the original post.

[Saduf2019]

@isaacgerg
You set a batch size and then enter input less than batch size it would be treated as empty, you cannot have that.
We cannot have issues to fix user code as this repo is for performance issues and feature request or bugs only.

[isaacgerg]

@Saduf2019 Please read the original issue carefully and note my proposal.

[google-ml-butler]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you.