Toby Roseman
Toby Roseman
What are the dtypes of `text_seq`, `pred_semantic`, and `ref_audio`? It looks like at least one of them is `complex64`. The Core ML Framework does not support models with complex inputs.
Here is a more concise way to reproduce the issue: ```python import torch import coremltools as ct class M(torch.nn.Module): def forward(self, x): return torch.tensor_split(x, 3) x = torch.arange(8) traced_model =...
Windows Subsystem for Linux (WSL) is another option for Windows users. Coremltools should work fine with WSL.
> Is there an option for mac users on arm64? Yes, we support that.
I can reproduce this issue with coremltools version `8.0b1`. The output should indeed have shape (10, 3). This look like a bug in coremltools, not the Framework. Specifically, it looks...
Can you share code to reproduce this result?
Please include a unit test in [test_torch_ops.py](https://github.com/apple/coremltools/blob/main/coremltools/converters/mil/frontend/torch/test/test_torch_ops.py) which fails without this fix and passes with this fix.
Does your PyTorch decoder also produce nans for the same input? What version of macOS are you using? What version of coremltools are you using?
Your code runs fine for me using coremltools version `8.0b1`.
I'm on macOS 14 and it works for me. `'dict' object has no attribute 'flatten'` is not a very helpful error. Looks like we're catching the original exception. You could...