vllm/model_executor/models/llama4.py · 4f0f844b1675419fd2171bc5e981a82386ec552b · OpenDAS / vllm_cscc · GitLab

Find file Blame History Permalink

Fix cuda illegal mem access with Llama4 TP8 + rms_norm custom op (#22701) · 4f0f844b
Po-Han Huang (NVIDIA) authored Aug 13, 2025
```
Signed-off-by: Po-Han Huang <pohanh@nvidia.com>
```
4f0f844b

llama4.py 30.8 KB