Update Dockerfile

Dockerfile

@@ -20,17 +20,20 @@ RUN apt-get update && apt-get install -y \
 # Create and set working directory
 WORKDIR /app
 
-# Install specific versions compatible with A100 and CUDA 12.1
+# Install base packages first
+RUN pip install --no-cache-dir packaging setuptools wheel
+
+# Install PyTorch with CUDA 12.1 support
 RUN pip install --no-cache-dir \
     torch==2.1.2+cu121 \
     --extra-index-url https://download.pytorch.org/whl/cu121
 
-# Install optimized vLLM for A100
+# Install vLLM and other packages in optimal order
 RUN pip install --no-cache-dir \
     vllm==0.3.0 \
     xformers==0.0.22.post7 \
     sentence-transformers==2.2.2 \
-    flash-attn==2.3.3
+    flash-attn==2.3.3 --no-build-isolation
 
 # Create necessary directories
 RUN mkdir -p /data/hf_cache/txagent_models \
@@ -49,5 +52,5 @@ COPY . .
 # Expose port
 EXPOSE 7860
 
-# Command to run the application with optimal settings for A100
+# Command to run the application
 CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]