[Feature]: Beam Search also requires diversity

[Jack-mi]

🚀 The feature, motivation and pitch

Basically in decoding stage, we use sampling strategy with parameters like top_p/top_k/temp/xx_penalty to control the diversity of the generated content. However beam_search is still very important at the moment, but lacks relevant parameters to control its diversity, which is already supported in HF as follows: HammingDiversityLogitsProcessor：https://huggingface.co/docs/transformers/zh/internal/generation_utils#transformers.HammingDiversityLogitsProcessor

Alternatives

No response

Additional context

No response

[yeshihai]

Same here!

[yeshihai]

The outputs of vLLM and Hugging Face differ significantly in this regard, making it impossible to achieve a seamless migration.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this issue should remain open. Thank you!

[github-actions[bot]]

This issue has been automatically closed due to inactivity. Please feel free to reopen if you feel it is still relevant. Thank you!