init
Browse files- Dockerfile +29 -0
Dockerfile
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
FROM python:3.11-slim
|
| 2 |
+
|
| 3 |
+
WORKDIR /app
|
| 4 |
+
|
| 5 |
+
# System build deps for llama-cpp-python (CPU + OpenBLAS for speed)
|
| 6 |
+
RUN apt-get update && apt-get install -y \
|
| 7 |
+
build-essential \
|
| 8 |
+
cmake \
|
| 9 |
+
libopenblas-dev \
|
| 10 |
+
curl \
|
| 11 |
+
&& rm -rf /var/lib/apt/lists/*
|
| 12 |
+
|
| 13 |
+
# Install Python deps — compile llama-cpp-python with OpenBLAS
|
| 14 |
+
COPY requirements.txt .
|
| 15 |
+
RUN CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS" \
|
| 16 |
+
FORCE_CMAKE=1 \
|
| 17 |
+
pip install --no-cache-dir -r requirements.txt
|
| 18 |
+
|
| 19 |
+
COPY app.py .
|
| 20 |
+
|
| 21 |
+
# HuggingFace Spaces expects port 7860
|
| 22 |
+
EXPOSE 7860
|
| 23 |
+
|
| 24 |
+
# Set SPACE_URL at runtime via HF Space secrets/env vars
|
| 25 |
+
ENV SPACE_URL=""
|
| 26 |
+
ENV N_CTX="8192"
|
| 27 |
+
ENV N_THREADS="2"
|
| 28 |
+
|
| 29 |
+
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860", "--workers", "1"]
|