fix facebook/hf-seamless-m4t-medium crash

[sywangyi]

What does this PR do?

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[jiminha]

@sywangyi Could you explain why other two models work but not facebook/hf-seamless-m4t-medium ?

[sywangyi]

facebook/hf-seamless-m4t-medium hit the condition that decode input id length is 2. and decoder_start_token_id is not in the decode input id, which cause issue in https://github.com/huggingface/optimum-habana/blob/main/optimum/habana/transformers/generation/utils.py#L249

[jiminha]

facebook/hf-seamless-m4t-medium hit the condition that decode input id length is 2. and decoder_start_token_id is not in the decode input id, which cause issue in https://github.com/huggingface/optimum-habana/blob/main/optimum/habana/transformers/generation/utils.py#L249

Is this only model that you are aware of which has decode input id length of 2?
When does below statement return true? Is this model specific, and only happening on this ht-seamless-m4t model?
elif (decoder_input_ids[:, 0] != decoder_start_token_id[:, 0]).all().item():

What other tests do you recommend us to run to make sure this new code in utils.py doesn't break any other models run?