Runtime Error:The weights trying to be saved contained shared tensors.

[IEI-mjx]

The model was quantized successfully.

But failed when saving the model:

[mgoin]

Hi @IEI-mjx does the model have tied embeddings?

[IEI-mjx]

Hi @IEI-mjx does the model have tied embeddings?

Yes

[IEI-mjx]

I skipped this error by copying the shared weight both lm_head.weight and model.embed_tokens.weight.
Since I haven't test this FP8-KV version model, I'm not sure whether this modification to save_quantized_model() is correct or not.