NotImplementedError: The operator 'aten::isin.Tensor_Tensor_out'

[cyrilzakka]

Trying to run this on the CLI on macOS:

insanely-fast-whisper --device mps  --batch-size 2 --file-name chunk0.mp3

NotImplementedError: The operator 'aten::isin.Tensor_Tensor_out' is not currently implemented for the MPS device. If you want this op to be added in priority during the prototype phase of this feature, please comment on https://github.com/pytorch/pytorch/issues/77764. As a temporary fix, you can set the environment variable `PYTORCH_ENABLE_MPS_FALLBACK=1` to use the CPU as a fallback for this op. WARNING: this will be slower than running natively on MPS.

Any advice?

For non-CLI, I now just run the following:

import os
os.environ['PYTORCH_ENABLE_MPS_FALLBACK'] = '1'
import torch
from transformers import pipeline
from transformers.utils import is_flash_attn_2_available

[cyrilzakka]

After updating to torch (nightly) everything seems to be working fine:

pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu

[pmane]

Did you manage to use GPU without os.environ['PYTORCH_ENABLE_MPS_FALLBACK'] = '1' ?
I was able to use this few months back on GPU, now looks like it is broken.

Can you give me your requirements .txt if you are able to work with GPU ? I am getting same error and not able to resolve

[mertbozkir]

After updating to torch (nightly) everything seems to be working fine:

pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu

Unfortunately, doesn't solve my problem:

NotImplementedError: The operator 'aten::isin.Tensor_Tensor_out' is not currently implemented for the MPS device. If you want this op to be added in priority during the prototype phase of this feature, please comment on https://github.com/pytorch/pytorch/issues/77764. As a temporary fix, you can set the environment variable `PYTORCH_ENABLE_MPS_FALLBACK=1` to use the CPU as a fallback for this op. WARNING: this will be slower than running natively on MPS.