Add the position_imbeddings param to LlamaAttention.forward

[nagic0]

The function LlamaAttention.forward() has a new param.

https://github.com/huggingface/transformers/blob/2e113422b3504fe6de821bb9911b24273b11aa9c/src/transformers/models/llama/modeling_llama.py#L317

The param position_embeddings is calculated here.

https://github.com/huggingface/transformers/blob/2e113422b3504fe6de821bb9911b24273b11aa9c/src/transformers/models/llama/modeling_llama.py#L917-L918

---

If the parameter is not added here, an error will occur when running the llama model:

TypeError: LlamaAttention.forward() got an unexpected keyword argument 'position_embeddings'

[alessandropalla]

What version of the transformers library raises this issue?

[nagic0]

What version of the transformers library raises this issue?

After v4.43.0

https://github.com/huggingface/transformers/blob/v4.43.0/src/transformers/models/llama/modeling_llama.py#L317

[alessandropalla]

PR is good. Can you please reformat your file using pre-commit (guide here: https://intel.github.io/intel-npu-acceleration-library/developer.html#developer-guide) in order to fix CI and also please add the transformers >= v4.43.0 to requirements.txt in order to not break new installs?

[alessandropalla]

Thanks you very much, merging this now