bhack
bhack
Yes cause `_process_probability_fn` was going to transform `str` back to the function. Can you send a PR to the specific test to help reproduce this case? https://github.com/tensorflow/addons/blob/master/tensorflow_addons/seq2seq/tests/attention_wrapper_test.py @seanpmorgan
Is this plan still active?
https://stackoverflow.com/questions/53841509/how-does-adaptive-pooling-in-pytorch-work
Is this process still active?
/cc @chenmoneygithub
/cc @chenmoneygithub
> For the onnx issue, we will likely get back to you in next week with a solution It would be nice if it could be exported `torch.export` without extra...
> Note also that it is not possible to do torch.export with variable size tensors because the model uses shape-dependent control flow inside of it (e.g. linear scan over the...
@artemZholus I give you an update. Are you on the official pytorch slack?
@icoderaven Are you able to export with the dynamic `Dim` query points min/max in the single frame scenario? For sure a single export is the better.