miike-ai/vscode-python312

1FROM ubuntu:22.042 3# Prevent interactive prompts4ENV DEBIAN_FRONTEND=noninteractive5ENV PYTHONUNBUFFERED=16ENV HF_HUB_ENABLE_HF_TRANSFER=17ENV CUDA_HOME=/usr/local/cuda8ENV PATH=/usr/local/cuda/bin:${PATH}9ENV LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH}10 11# Install base system dependencies12RUN apt-get update && apt-get install -y \13    wget \14    curl \15    git \16    vim \17    tmux \18    htop \19    build-essential \20    software-properties-common \21    ca-certificates \22    gnupg \23    lsb-release \24    sudo \25    openssh-server \26    nginx \27    supervisor \28    python3.11 \29    python3.11-dev \30    python3-pip \31    && rm -rf /var/lib/apt/lists/*32 33# Add NVIDIA package repositories34RUN wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb && \35    dpkg -i cuda-keyring_1.1-1_all.deb && \36    apt-get update37 38# Install CUDA 12.839RUN apt-get install -y cuda-toolkit-12-8 && \40    rm -rf /var/lib/apt/lists/*41 42# Install cuDNN43RUN apt-get update && apt-get install -y \44    libcudnn9-cuda-12 \45    libcudnn9-dev-cuda-12 \46    && rm -rf /var/lib/apt/lists/*47 48# Set Python 3.11 as default49RUN update-alternatives --install /usr/bin/python python /usr/bin/python3.11 1 && \50    update-alternatives --install /usr/bin/pip pip /usr/bin/pip3 151 52# Upgrade pip53RUN python -m pip install --upgrade pip setuptools wheel54 55# Install PyTorch with CUDA 12.8 support56RUN pip install torch==2.5.1+cu128 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu12857 58# Create non-root user for HF Spaces compatibility59RUN useradd -m -u 1000 user && \60    echo "user ALL=(ALL) NOPASSWD:ALL" >> /etc/sudoers61 62# Install code-server (VSCode in browser)63RUN curl -fsSL https://code-server.dev/install.sh | sh64 65# Install Ollama66RUN curl -fsSL https://ollama.com/install.sh | sh67 68# Install hf_transfer first69RUN pip install hf_transfer70 71# Install core ML packages72RUN pip install \73    accelerate \74    transformers \75    datasets \76    peft \77    bitsandbytes \78    safetensors \79    sentencepiece \80    protobuf \81    scipy \82    einops \83    wandb \84    tensorboard \85    gradio \86    streamlit87 88# Install vLLM89RUN pip install vllm90 91# Install Flash Attention 292RUN pip install ninja packaging && \93    pip install flash-attn --no-build-isolation94 95# Install Triton for 5090 support96RUN pip install triton97 98# Clone and install Unsloth with patches99RUN git clone https://github.com/unslothai/unsloth.git /tmp/unsloth && \100    cd /tmp/unsloth && \101    pip install -e . && \102    cd / && \103    rm -rf /tmp/unsloth/.git104 105# Clone and install Axolotl106RUN git clone https://github.com/axolotl-ai-cloud/axolotl /tmp/axolotl && \107    cd /tmp/axolotl && \108    pip install -e . && \109    cd / && \110    rm -rf /tmp/axolotl/.git111 112# Install Node.js for Open-WebUI113RUN curl -fsSL https://deb.nodesource.com/setup_20.x | bash - && \114    apt-get install -y nodejs && \115    rm -rf /var/lib/apt/lists/*116 117# Clone and setup Open-WebUI118RUN git clone https://github.com/open-webui/open-webui.git /opt/open-webui && \119    cd /opt/open-webui && \120    npm install && \121    npm run build122 123# Create directories with proper permissions124RUN mkdir -p /home/user/app /home/user/.cache /home/user/.config && \125    chown -R user:user /home/user126 127# Configure code-server for user128RUN mkdir -p /home/user/.config/code-server && \129    echo "bind-addr: 0.0.0.0:8080\nauth: none\ncert: false" > /home/user/.config/code-server/config.yaml && \130    chown -R user:user /home/user/.config131 132# Setup SSH133RUN mkdir /var/run/sshd && \134    echo 'user:spaces' | chpasswd && \135    sed -i 's/#PermitRootLogin prohibit-password/PermitRootLogin no/' /etc/ssh/sshd_config136 137# Create supervisor config138RUN mkdir -p /etc/supervisor/conf.d139RUN cat > /etc/supervisor/conf.d/services.conf << 'EOF'140[supervisord]141nodaemon=true142user=root143 144[program:code-server]145command=sudo -u user code-server --bind-addr 0.0.0.0:8080 --auth none146autostart=true147autorestart=true148stderr_logfile=/var/log/code-server.err.log149stdout_logfile=/var/log/code-server.out.log150 151[program:ollama]152command=ollama serve153autostart=true154autorestart=true155environment=OLLAMA_HOST="0.0.0.0",HOME="/home/user"156stderr_logfile=/var/log/ollama.err.log157stdout_logfile=/var/log/ollama.out.log158 159[program:open-webui]160command=cd /opt/open-webui && npm start161autostart=true162autorestart=true163environment=PORT="3000",OLLAMA_BASE_URL="http://localhost:11434"164stderr_logfile=/var/log/open-webui.err.log165stdout_logfile=/var/log/open-webui.out.log166user=user167EOF168 169# Create Gradio app for HF Spaces170RUN cat > /home/user/app/app.py << 'EOF'171import gradio as gr172import subprocess173import os174 175def get_services_status():176    services = {177        "VSCode": "http://localhost:8080",178        "Ollama API": "http://localhost:11434", 179        "Open-WebUI": "http://localhost:3000",180        "vLLM": "http://localhost:8000"181    }182    183    status = "# 🚀 ML Stack Services Status\n\n"184    for service, url in services.items():185        status += f"- **{service}**: {url}\n"186    187    # Get GPU info188    try:189        gpu_info = subprocess.check_output(['nvidia-smi', '--query-gpu=name,memory.total', '--format=csv,noheader'], text=True)190        status += f"\n## 🎮 GPU Status\n```\n{gpu_info}```"191    except:192        status += "\n## ⚠️ No GPUs detected"193    194    return status195 196def launch_vllm(model_name):197    try:198        cmd = f"python -m vllm.entrypoints.openai.api_server --model {model_name} --host 0.0.0.0 --port 8000"199        subprocess.Popen(cmd, shell=True)200        return f"✅ Launching vLLM with model: {model_name}"201    except Exception as e:202        return f"❌ Error: {str(e)}"203 204def pull_ollama_model(model_name):205    try:206        result = subprocess.run(['ollama', 'pull', model_name], capture_output=True, text=True)207        return f"✅ {result.stdout}\n{result.stderr}"208    except Exception as e:209        return f"❌ Error: {str(e)}"210 211# Create Gradio interface212with gr.Blocks(title="ML Stack Control Panel") as demo:213    gr.Markdown("# 🎮 RunPod ML Stack Control Panel")214    215    with gr.Tab("Status"):216        status_btn = gr.Button("🔄 Refresh Status")217        status_output = gr.Markdown()218        status_btn.click(get_services_status, outputs=status_output)219    220    with gr.Tab("vLLM"):221        model_input = gr.Textbox(label="Model Name", value="meta-llama/Llama-2-7b-hf")222        vllm_btn = gr.Button("🚀 Launch vLLM")223        vllm_output = gr.Textbox(label="Output")224        vllm_btn.click(launch_vllm, inputs=model_input, outputs=vllm_output)225    226    with gr.Tab("Ollama"):227        ollama_model = gr.Textbox(label="Model Name", value="llama3.2")228        ollama_btn = gr.Button("📥 Pull Model")229        ollama_output = gr.Textbox(label="Output")230        ollama_btn.click(pull_ollama_model, inputs=ollama_model, outputs=ollama_output)231    232    # Load initial status233    demo.load(get_services_status, outputs=status_output)234 235if __name__ == "__main__":236    # Start supervisor in background237    subprocess.Popen(["/usr/bin/supervisord", "-c", "/etc/supervisor/supervisord.conf"])238    239    # Launch Gradio240    demo.launch(server_name="0.0.0.0", server_port=7860, share=False)241EOF242 243# Create startup script244RUN cat > /home/user/app/start.sh << 'EOF'245#!/bin/bash246cd /home/user/app247python app.py248EOF249RUN chmod +x /home/user/app/start.sh250 251# Fix permissions252RUN chown -R user:user /home/user /opt/open-webui253 254# Expose ports255EXPOSE 22      # SSH256EXPOSE 7860    # Gradio (HF Spaces default)257EXPOSE 8080    # Code-server258EXPOSE 11434   # Ollama259EXPOSE 8000    # vLLM260EXPOSE 3000    # Open-WebUI261 262# Switch to user263USER user264WORKDIR /home/user/app265 266# Set the entrypoint for HF Spaces267CMD ["python", "app.py"]