vllm
bd2d83ff - [SpecDecode] Reduce TP communication for large-vocab draft models speculative decoding (#39419)

Commit

36 days ago

[SpecDecode] Reduce TP communication for large-vocab draft models speculative decoding (#39419) Signed-off-by: EanWang211123 <wangyiheng@sangfor.com.cn>

References

#39419 - [SpecDecode] Reduce TP communication for large-vocab draft models speculative decoding

Author

EanWang211123

Parents

bb78168b

vllm bd2d83ff - [SpecDecode] Reduce TP communication for large-vocab draft models speculative decoding (#39419)

vllm
bd2d83ff - [SpecDecode] Reduce TP communication for large-vocab draft models speculative decoding (#39419)