iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[ROCm][CI][Multimodal] Use ROCm-aware FA availability check for Unlimited-OCR (#47004)

Browse Source 
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
This commit is contained in:
Andreas Karatzas
2026-06-30 14:03:13 +08:00
committed by GitHub
parent e45c8a9f4b
commit 4236514098
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/model_executor/models/config.py
+1 -1
View File
@@ -84,8 +84,8 @@ class UnlimitedOCRForCausalLMConfig(VerifyAndUpdateConfig):
vllm serve baidu/Unlimited-OCR \\
--attention-config '{"backend": "FLEX_ATTENTION"}'
"""
from vllm.v1.attention.backends.fa_utils import is_fa_version_supported
from vllm.v1.attention.backends.registry import AttentionBackendEnum
from vllm.vllm_flash_attn import is_fa_version_supported
attn_config = vllm_config.attention_config
fa4_available = is_fa_version_supported(4)