install flash attention

Dockerfile

@@ -13,8 +13,10 @@ COPY --from=uv /uv /uv
 RUN apt-get update && \
     apt-get upgrade -y && \
     apt-get install -y --no-install-recommends \
-    python3.11 python3.11-venv python3-pip ffmpeg && \
-    apt-get clean && \
+    python3.11 python3.11-venv python3-pip ffmpeg \
+    build-essential \
+    git \
+    && apt-get clean && \
     rm -rf /var/lib/apt/lists/*
 
 # Create virtual environment with uv
@@ -47,7 +49,11 @@ COPY --chown=user requirements.txt .
 
 # Install Python packages with uv caching
 RUN --mount=type=cache,target=$UV_CACHE_DIR,uid=1000,gid=1000 \
-    /uv pip install -r requirements.txt
+    /uv pip install --no-cache-dir -r requirements.txt
+
+# Install build dependencies for flash-attn
+RUN --mount=type=cache,target=$UV_CACHE_DIR,uid=1000,gid=1000 \
+    /uv pip install -U --no-cache-dir setuptools wheel ninja packaging
 
 # Install flash-attn
 RUN --mount=type=cache,target=$UV_CACHE_DIR,uid=1000,gid=1000 \