tensorflow: Working code broke after deploying to new installation. ValueError: When using `stateful=True` in a RNN, the batch size must be static. Found dynamic batch size: sequence.shape=(None, xx, xx)

Issue type

Bug

Have you reproduced the bug with TensorFlow Nightly?

Yes

Source

source

TensorFlow version

2.16.1

Custom code

No

OS platform and distribution

Ubuntu 22.04

Mobile device

No

Python version

3.10.12

Bazel version

No response

GCC/compiler version

No response

CUDA/cuDNN version

12.4

GPU model and memory

No response

Current behavior?

So the existing code broke, with this error

ValueError: When using stateful=True in a RNN, the batch size must be static. Found dynamic batch size: sequence.shape=(None, XX, XX),

but when changed to (XX, XX, XX) I get this, ValueError: Input 0 of layer “lstm” is incompatible with the layer: expected ndim=3, found ndim=4. Full shape received: (None, 1, XX, XX)

If static shape is passed, not sure why additional dimension is added. Need help in resolving this. I tried both the release version and nightly version. The issue persists

And also the input is 3D tensor produced by timeseries_dataset_from_array function. Is there any example on training stateful LSTM using these 3D tensors? Not sure if this is a bug really. I am stuck.

Standalone code to reproduce the issue

Dont have stand alone.

Relevant log output

No response

About this issue

  • Original URL
  • State: closed
  • Created 3 years ago
  • Comments: 16 (2 by maintainers)

Most upvoted comments

Try using the input layer instead of batch_input_shape.

InputLayer(batch_shape=(1,None,4))