[https://nvbugs/6427240][fix] Reserve MTP draft tokens in scheduler for one-model speculative decoding - #16101

[https://nvbugs/6427240][fix] Reserve MTP draft tokens in scheduler for one-model speculative decoding#16101

jhaotingc merged 1 commit into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

jhaotingc:fix/mtp-one-model-draft-token-schedulingjhaotingc/TensorRT-LLM:fix/mtp-one-model-draft-token-schedulingCopy head branch name to clipboard