[MM][CG] Support ViT CG for Qwen2.5-VL (#40830)

Signed-off-by: John Calderon <jcalderon@nvidia.com>
This commit is contained in:
John Calderon
2026-05-02 11:10:14 +08:00
committed by GitHub
parent c408fdd663
commit 964a4bc2a5
5 changed files with 539 additions and 22 deletions
+2
View File
@@ -86,9 +86,11 @@ Models opt-in to encoder CUDA Graphs by implementing the [SupportsEncoderCudaGra
| Architecture | Models | CG for Image | CG for Video |
| ------------ | ------ | ------------ | ------------ |
| `Qwen3VLForConditionalGeneration` | `Qwen3-VL` | ✅︎ | ✅︎ |
| `Qwen2_5_VLForConditionalGeneration` | `Qwen2.5-VL` | ✅︎ | ✅︎ |
!!! note
Encoder CUDA Graphs have currently been tested with `--mm-encoder-attn-backend=FLASH_ATTN` and `--mm-encoder-attn-backend=FLASHINFER` on Blackwell GPUs.
For Qwen2.5-VL only FA2 and FA3 has been tested.
## Configuration