Update Dockerfile

Dockerfile

@@ -2,6 +2,7 @@
 FROM vllm/vllm-openai:v0.10.0
 # FROM nvidia/cuda:11.7.1-cudnn8-runtime-ubuntu22.04
 
+ENV OMP_NUM_THREADS=2
 # ENV VLLM_LOGGING_LEVEL=DEBUG
 # ENV HF_HOME=/tmp/.cache/huggingface
 # ENV OMP_NUM_THREADS=1
@@ -11,7 +12,17 @@ FROM vllm/vllm-openai:v0.10.0
 
 # # https://github.com/vllm-project/vllm/blob/v0.10.0/docs/getting_started/quickstart.md?plain=1#L213
 # # `FLASH_ATTN` or `FLASHINFER` or `XFORMERS`.
-# ENV VLLM_ATTENTION_BACKEND=FLASH_ATTN
+# # https://github.com/vllm-project/vllm/blob/main/vllm/envs.py#L426-L435
+# # Backend for attention computation
+# # Available options:
+# # - "TORCH_SDPA": use torch.nn.MultiheadAttention
+# # - "FLASH_ATTN": use FlashAttention
+# # - "XFORMERS": use XFormers
+# # - "ROCM_FLASH": use ROCmFlashAttention
+# # - "FLASHINFER": use flashinfer
+# # - "FLASHMLA": use FlashMLA
+# # Choose XFORMERS that most stable for T4
+# ENV VLLM_ATTENTION_BACKEND=XFORMERS
 
 # RUN apt-get update && apt-get install -y python3 python3-pip git
 # RUN pip3 install --upgrade pip