Spaces:
Running on Zero
Running on Zero
Remove incompatible Docker entrypoint
Browse files- Dockerfile +0 -36
- server_entrypoint.sh +0 -28
Dockerfile
DELETED
|
@@ -1,36 +0,0 @@
|
|
| 1 |
-
FROM python:3.10-slim
|
| 2 |
-
|
| 3 |
-
ENV DEBIAN_FRONTEND=noninteractive \
|
| 4 |
-
CUDA_VISIBLE_DEVICES="" \
|
| 5 |
-
GGML_CUDA=0 \
|
| 6 |
-
OMP_NUM_THREADS=16 \
|
| 7 |
-
OMP_THREAD_LIMIT=16 \
|
| 8 |
-
HF_HOME=/data/.cache/huggingface
|
| 9 |
-
|
| 10 |
-
RUN apt-get update && apt-get install -y --no-install-recommends \
|
| 11 |
-
build-essential cmake git ca-certificates \
|
| 12 |
-
&& rm -rf /var/lib/apt/lists/*
|
| 13 |
-
|
| 14 |
-
WORKDIR /opt
|
| 15 |
-
RUN git clone --depth=1 https://github.com/ggml-org/llama.cpp.git \
|
| 16 |
-
&& cmake -S llama.cpp -B llama.cpp/build \
|
| 17 |
-
-DGGML_NATIVE=OFF \
|
| 18 |
-
-DGGML_OPENMP=OFF \
|
| 19 |
-
-DGGML_CUDA=OFF \
|
| 20 |
-
-DGGML_AVX=ON \
|
| 21 |
-
-DGGML_AVX2=ON \
|
| 22 |
-
-DGGML_FMA=ON \
|
| 23 |
-
-DGGML_AVX512=OFF \
|
| 24 |
-
-DBUILD_SHARED_LIBS=OFF \
|
| 25 |
-
&& cmake --build llama.cpp/build --config Release --target llama-server -j16 \
|
| 26 |
-
&& find llama.cpp/build -type f -name llama-server -perm -111 -exec cp {} /usr/local/bin/llama-server \; \
|
| 27 |
-
&& test -x /usr/local/bin/llama-server \
|
| 28 |
-
&& rm -rf /opt/llama.cpp
|
| 29 |
-
|
| 30 |
-
RUN pip install --no-cache-dir huggingface-hub
|
| 31 |
-
|
| 32 |
-
COPY server_entrypoint.sh /opt/server_entrypoint.sh
|
| 33 |
-
RUN chmod +x /opt/server_entrypoint.sh
|
| 34 |
-
|
| 35 |
-
EXPOSE 7860
|
| 36 |
-
CMD ["/opt/server_entrypoint.sh"]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
server_entrypoint.sh
DELETED
|
@@ -1,28 +0,0 @@
|
|
| 1 |
-
#!/bin/sh
|
| 2 |
-
set -eu
|
| 3 |
-
|
| 4 |
-
MODEL_DIR=/data/models/tinyllama
|
| 5 |
-
MODEL_PATH="$MODEL_DIR/tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf"
|
| 6 |
-
mkdir -p "$MODEL_DIR"
|
| 7 |
-
|
| 8 |
-
if [ ! -f "$MODEL_PATH" ]; then
|
| 9 |
-
echo "Downloading TinyLlama GGUF..."
|
| 10 |
-
hf download \
|
| 11 |
-
TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF \
|
| 12 |
-
tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf \
|
| 13 |
-
--repo-type model \
|
| 14 |
-
--local-dir "$MODEL_DIR"
|
| 15 |
-
fi
|
| 16 |
-
|
| 17 |
-
# Keep this process as PID 1: no Gradio/Python subprocess wrapper.
|
| 18 |
-
exec /usr/local/bin/llama-server \
|
| 19 |
-
--model "$MODEL_PATH" \
|
| 20 |
-
--host 0.0.0.0 \
|
| 21 |
-
--port 7860 \
|
| 22 |
-
--n-gpu-layers 0 \
|
| 23 |
-
--threads 16 \
|
| 24 |
-
--threads-batch 16 \
|
| 25 |
-
--ctx-size 2048 \
|
| 26 |
-
--parallel 1 \
|
| 27 |
-
--no-mmap \
|
| 28 |
-
--no-warmup
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|