Back to all discussions

  •   8 months ago

KeyError: 'model.layers.X.mlp.experts.gate_up_proj' when loading gpt-oss-20b with transformers 4.56

Hi all—
Is anyone else getting a KeyError like 'model.layers.1.mlp.experts.gate_up_proj' when running openai/gpt-oss-20b with the latest transformers and torch?
I've tried both remote and local loading, all files/extensions are correct, and all other models like gpt2 work on my system.

Any workaround or confirmation of this issue? Appreciate any pointers!, Thanks.

  • 1 comment

  •   •   8 months ago

    We had the HuggingFace team try to reproduce it but they don't seem to be able to reproduce it. Do you have any additional details to share?

Comments are closed.