GRU runtime error.

[bigbigzxl]

Describe the bug
in GRU case，with the same parameters and inputs, the output of ONNX is inconstant with Tensorflow.

Urgency
none.

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04) ： (Red Hat 4.8.5-16)] on linux

• ONNX Runtime installed from (source or binary): binary

• ONNX Runtime version: onnx.version='1.6.0'

• Python version: python3.65

• Visual Studio version (if applicable):

• GCC/Compiler version (if compiling from source): GCC 4.8.5 20150623

• CUDA/cuDNN version:
  torch.version： - 1.7.1+cu101
  torch.version.cuda：10.2
  torch.backends.cudnn.version()：7603

• GPU model and memory: Tesla V100， 64GB

To Reproduce
just single gru case.

Expected behavior
the output of onnx is equal with the output of tensorflow.

Screenshots

Additional context
GRU params:
Batch = 3
time step: T = 28
output size: H = 5
input size: X = 10

[ytaous]

@skottmckay - any idea?

[skottmckay]

Can you share the model?

May be due to differences in how output values past the end of the sequence are treated. ORT zeros those out as they're irrelevant. TF may repeat the final value. Either way, they're past the end of the sequence and should be ignored.

[stale]

This issue has been automatically marked as stale due to inactivity and will be closed in 7 days if no further activity occurs. If further support is needed, please provide an update and/or more details.

[snnn]

This issue has been automatically closed as 'not planned' because it has been marked as 'stale' for more than 30 days without activity. If you believe this is still an issue, please feel free to reopen it.