Update Dockerfile

Dockerfile

@@ -1,79 +1,108 @@
-FROM ubuntu:22.04
+FROM nvidia/cuda:12.8.0-devel-ubuntu22.04
 
 # Set environment variables
 ENV DEBIAN_FRONTEND=noninteractive
 ENV HOME=/root
-# CUDA paths for when running on GPU (RunPod)
 ENV CUDA_HOME=/usr/local/cuda
 ENV PATH=${CUDA_HOME}/bin:${PATH}
 ENV LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${LD_LIBRARY_PATH}
+ENV TORCH_CUDA_ARCH_LIST="12.0"
 
-# Install dependencies, code-server, and Ollama
+# Install system dependencies and build tools
 RUN apt-get update && \
-    apt-get install -y curl wget gpg apt-transport-https git software-properties-common && \
-    # Add Python 3.12
-    add-apt-repository ppa:deadsnakes/ppa && \
+    apt-get install -y \
+    curl wget gpg apt-transport-https git software-properties-common \
+    build-essential cmake ninja-build \
+    libopenblas-dev libomp-dev \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install Python 3.12 (as recommended)
+RUN add-apt-repository ppa:deadsnakes/ppa && \
     apt-get update && \
-    apt-get install -y python3.11 python3.11-venv python3.11-dev && \
-    # Make Python 3.12 the default
-    update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.11 1 && \
-    update-alternatives --set python3 /usr/bin/python3.11 && \
-    # Install pip for Python 3.12
-    curl -sS https://bootstrap.pypa.io/get-pip.py | python3.11 && \
-    # Upgrade pip
-    python3 -m pip install --upgrade pip setuptools wheel && \
-    # Install Node.js 22.x from NodeSource
-    curl -fsSL https://deb.nodesource.com/setup_22.x | bash - && \
-    apt-get install -y nodejs && \
-    # Install code-server
-    curl -fsSL https://code-server.dev/install.sh | sh && \
-    apt-get clean && \
-    rm -rf /var/lib/apt/lists/*
+    apt-get install -y python3.12 python3.12-venv python3.12-dev && \
+    update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.12 1 && \
+    update-alternatives --set python3 /usr/bin/python3.12 && \
+    curl -sS https://bootstrap.pypa.io/get-pip.py | python3.12
 
-# Install global npm packages
-RUN npm install -g @anthropic-ai/claude-code @anthropic-ai/dxt
+# Install uv for faster dependency management
+RUN curl -LsSf https://astral.sh/uv/install.sh | sh && \
+    echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc
 
-# Create a directory for the workspace
+# Create workspace
 RUN mkdir -p /workspace
+WORKDIR /workspace
+
+# CRITICAL: Install in the EXACT order specified in the README
+# Step 1: Create venv
+RUN python3 -m venv /opt/unsloth-env
+ENV PATH="/opt/unsloth-env/bin:$PATH"
+
+# Step 2: Install vllm with cu128 (MUST be first)
+RUN /opt/unsloth-env/bin/pip install --upgrade pip setuptools wheel && \
+    /opt/unsloth-env/bin/pip install -U vllm --extra-index-url https://wheels.vllm.ai/nightly
+
+# Step 3: Install unsloth dependencies
+RUN /opt/unsloth-env/bin/pip install unsloth unsloth_zoo bitsandbytes
 
-# Copy and install requirements - fail loudly if not found
-COPY requirements.txt /workspace/requirements.txt
-RUN echo "Installing Python requirements..." && \
-    pip3 install --no-cache-dir -r /workspace/requirements.txt && \
-    echo "Requirements installed successfully!"
+# Step 4: Build xformers from source (no Blackwell wheels exist yet)
+RUN git clone --depth=1 https://github.com/facebookresearch/xformers --recursive /tmp/xformers && \
+    cd /tmp/xformers && \
+    /opt/unsloth-env/bin/pip uninstall -y xformers && \
+    /opt/unsloth-env/bin/python setup.py install && \
+    rm -rf /tmp/xformers
 
-# Create configuration directory for code-server and Ollama
-RUN mkdir -p /root/.config/code-server /root/.ollama
+# Step 5: Update triton to >=3.3.1 for Blackwell
+RUN /opt/unsloth-env/bin/pip install -U "triton>=3.3.1"
 
-# Configure code-server to run on port 8443
-RUN echo "bind-addr: 0.0.0.0:8443\nauth: none\ncert: false" > /root/.config/code-server/config.yaml
+# Step 6: Pin transformers to avoid gradient checkpointing bug
+RUN /opt/unsloth-env/bin/pip install -U "transformers==4.52.4"
 
-# Install Ollama after code-server is set up
+# Step 7: Might need to downgrade numpy
+RUN /opt/unsloth-env/bin/pip install "numpy<=2.2"
+
+# Now install your dev tools (code-server, Node.js, etc)
+RUN curl -fsSL https://deb.nodesource.com/setup_22.x | bash - && \
+    apt-get install -y nodejs && \
+    curl -fsSL https://code-server.dev/install.sh | sh && \
+    npm install -g @anthropic-ai/claude-code @anthropic-ai/dxt && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# Install Ollama
 RUN curl -fsSL https://ollama.com/install.sh | sh || true
 
-# Install some useful VS Code extensions
+# Configure code-server
+RUN mkdir -p /root/.config/code-server /root/.ollama && \
+    echo "bind-addr: 0.0.0.0:8443\nauth: none\ncert: false" > /root/.config/code-server/config.yaml
+
+# Install VS Code extensions
 RUN code-server --install-extension ms-python.python && \
     code-server --install-extension ritwickdey.LiveServer && \
     code-server --install-extension ms-toolsai.jupyter
 
-# Create a startup script
+# Create startup script with proper env activation
 RUN echo '#!/bin/bash\n\
+# Activate the unsloth environment\n\
+source /opt/unsloth-env/bin/activate\n\
+\n\
 # Start Ollama in the background\n\
 /usr/local/bin/ollama serve &\n\
 \n\
 # Give Ollama a moment to start\n\
 sleep 2\n\
 \n\
-# Start code-server in the foreground\n\
+# Start code-server with the activated environment\n\
 exec code-server --disable-telemetry --bind-addr 0.0.0.0:8443 /workspace\n\
 ' > /start.sh && \
     chmod +x /start.sh
 
-# Expose ports for both services
+# Expose ports
 EXPOSE 8443 11434
 
-# Set the workspace as working directory
-WORKDIR /workspace
+# Copy any user requirements AFTER base installation
+COPY requirements.txt* /workspace/
+RUN if [ -f /workspace/requirements.txt ]; then \
+    /opt/unsloth-env/bin/pip install -r /workspace/requirements.txt; \
+    fi
 
-# Start both services
 CMD ["/start.sh"]