Update Dockerfile

Dockerfile

@@ -1,12 +1,15 @@
 # Use NVIDIA CUDA image for GPU acceleration
 FROM nvidia/cuda:12.1.1-cudnn8-devel-ubuntu22.04 
 
-# Set environment variables
+# Set environment variables - note the additional ones to disable Triton
 ENV DEBIAN_FRONTEND=noninteractive \
     HF_HOME=/app/.cache/huggingface \
     TRANSFORMERS_CACHE=/app/.cache/huggingface \
     TORCH_HOME=/app/.cache/torch \
-    TRITON_DISABLE="1"
+    TRITON_DISABLE="1" \
+    BNB_DISABLE_TRITON="1" \
+    USE_TORCH="1" \
+    BITSANDBYTES_NOWELCOME="1"
 
 # Install system dependencies
 RUN apt-get update && apt-get install -y \
@@ -16,8 +19,9 @@ RUN apt-get update && apt-get install -y \
 # Set working directory
 WORKDIR /app 
 
-# Create necessary directories with proper permissions
+# Create necessary directories with proper permissions - try creating /.triton too
 RUN mkdir -p /app/.cache/huggingface /app/.cache/torch && chmod -R 777 /app/.cache 
+RUN mkdir -p /.triton && chmod 777 /.triton
 
 # Create and activate virtual environment
 RUN python3 -m venv venv
@@ -29,14 +33,9 @@ RUN pip install --no-cache-dir --upgrade pip setuptools wheel
 # Install PyTorch separately
 RUN pip install --no-cache-dir torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
 
-# Install other dependencies one by one to isolate issues
-RUN pip install --no-cache-dir transformers
-RUN pip install --no-cache-dir accelerate
-RUN pip install --no-cache-dir fastapi uvicorn
+# Install other dependencies
+RUN pip install --no-cache-dir transformers accelerate fastapi uvicorn huggingface_hub protobuf
 RUN pip install --no-cache-dir bitsandbytes
-RUN pip install --no-cache-dir huggingface_hub
-RUN pip install --no-cache-dir protobuf
-RUN pip install --no-cache-dir optimum
 
 # Expose API port
 EXPOSE 7860