[https://nvbugs/6693989][fix] Apply the K3 prompt-token offset only when a prompt count is actually known… - #18509

[https://nvbugs/6693989][fix] Apply the K3 prompt-token offset only when a prompt count is actually known…#18509

moraxu merged 2 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

tensorrt-cicd:repair-bot-bug6693989tensorrt-cicd/TensorRT-LLM:repair-bot-bug6693989Copy head branch name to clipboard