biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Model] Always use Transformers backend for PaliGemma and Gemma3-MM (#26715)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-10-17 13:03:35 +08:00
committed by GitHub
parent 9c2c2287a0
commit 8c017b3490
12 changed files with 54 additions and 1219 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/platforms/rocm.py
View File

@@ -59,9 +59,6 @@ _ROCM_PARTIALLY_SUPPORTED_MODELS: dict[str, str] = {
"Qwen2ForCausalLM": _ROCM_SWA_REASON,
"MistralForCausalLM": _ROCM_SWA_REASON,
"MixtralForCausalLM": _ROCM_SWA_REASON,
"PaliGemmaForConditionalGeneration": (
"ROCm flash attention does not yet fully support 32-bit precision on PaliGemma"
),
"Phi3VForCausalLM": (
"ROCm Triton flash attention may run into compilation errors due to "
"excessive use of shared memory. If this happens, disable Triton FA "