vllm/v1/worker/gpu_model_runner.py · 41fb013d299941486cb026a0e607d2f77a7ec1e8 · OpenDAS / vllm_cscc · GitLab

Find file Blame History Permalink

[V1][Spec Decode] Always use argmax for sampling draft tokens (#16899) · 41fb013d
Woosuk Kwon authored Apr 23, 2025
```
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
```
41fb013d

gpu_model_runner.py 79.6 KB