Francesc Garcia

We had the same issue in our codebase and at the end we found this is a reproducible case from coremltools unit tests. This is why I'm really puzzled about...

About the sizes, there is specific code to reshape the results from bidirectional LSTM in Torch to make it compatible with CoreML. It should be around there btw because the...

Thanks for the insights ... I think all the issues come from an indexing error. Not sure if it's in this code or in the converter. I'm preparing a simpler...

This is the simplest test case I could think of: ``` import torch import torch.nn as nn import numpy as np from coremltools import TensorType from coremltools import convert def...

Thanks @leovinus2001. I'll let this opened until someone from CoreML can have a look.

Did you have time to look into this @aseemw ? Thanks