From 577de2fa078ee62d61bca24b75fa59590e2ef66b Mon Sep 17 00:00:00 2001
From: hiyouga <467089858@qq.com>
Date: Wed, 12 Jun 2024 16:50:11 +0800
Subject: [PATCH] fix #4242

---
 Dockerfile                           | 2 +-
 src/llamafactory/chat/vllm_engine.py | 5 +++--
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/Dockerfile b/Dockerfile
index 45849601..3932ff30 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -32,7 +32,7 @@ RUN EXTRA_PACKAGES="metrics"; \
         EXTRA_PACKAGES="${EXTRA_PACKAGES},deepspeed"; \
     fi; \
     pip install -e .[$EXTRA_PACKAGES] && \
-    pip uninstall -y transformer-engine
+    pip uninstall -y transformer-engine flash-attn
 
 # Set up volumes
 VOLUME [ "/root/.cache/huggingface/", "/app/data", "/app/output" ]
diff --git a/src/llamafactory/chat/vllm_engine.py b/src/llamafactory/chat/vllm_engine.py
index d096f6eb..d2850a6e 100644
--- a/src/llamafactory/chat/vllm_engine.py
+++ b/src/llamafactory/chat/vllm_engine.py
@@ -13,10 +13,11 @@ from .base_engine import BaseEngine, Response
 if is_vllm_available():
     from vllm import AsyncEngineArgs, AsyncLLMEngine, RequestOutput, SamplingParams
     from vllm.lora.request import LoRARequest
+
     try:
-        from vllm.multimodal import MultiModalData  # vllm==0.5.0
+        from vllm.multimodal import MultiModalData  # type: ignore (for vllm>=0.5.0)
     except ImportError:
-        from vllm.sequence import MultiModalData  # vllm<0.5.0
+        from vllm.sequence import MultiModalData  # for vllm<0.5.0
 
 
 if TYPE_CHECKING: