Skip to content

Commit f7a6f1d

Browse files
authored
Update Dockerfile
1 parent dafccae commit f7a6f1d

File tree

1 file changed

+13
-57
lines changed

1 file changed

+13
-57
lines changed

Dockerfile

Lines changed: 13 additions & 57 deletions
Original file line numberDiff line numberDiff line change
@@ -1,65 +1,21 @@
1-
# FROM pytorch/pytorch:2.5.1-cuda12.1-cudnn9-devel as compile_server
2-
FROM pytorch/pytorch:2.6.0-cuda12.6-cudnn9-devel as compile_server
1+
FROM nvcr.io/nvidia/pytorch:25.03-py3
32

4-
ARG CPU_INSTRUCT=NATIVE
3+
EXPOSE 28000
54

6-
# 设置工作目录和 CUDA 路径
7-
WORKDIR /workspace
8-
ENV CUDA_HOME=/usr/local/cuda
5+
ENV TZ=Asia/Shanghai
6+
RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone && apt update && apt install python3-tk -y
97

10-
# 安装依赖
11-
RUN apt update -y && apt install -y apt-utils && apt upgrade -y
12-
RUN apt install -y --no-install-recommends \
13-
libtbb-dev \
14-
libssl-dev \
15-
libcurl4-openssl-dev \
16-
libaio1 \
17-
libaio-dev \
18-
libfmt-dev \
19-
libgflags-dev \
20-
zlib1g-dev \
21-
patchelf \
22-
git \
23-
wget \
24-
vim \
25-
gcc \
26-
g++
8+
RUN mkdir /app
279

28-
# 安装CMake
29-
RUN wget https://github.com/Kitware/CMake/releases/download/v4.0.1/cmake-4.0.1-linux-x86_64.sh && echo y | bash cmake-4.0.1-linux-x86_64.sh && rm -rf cmake-4.0.1-linux-x86_64.sh
10+
WORKDIR /app
11+
RUN git clone --recurse-submodules https://github.com/Akegarasu/lora-scripts
3012

31-
# 克隆代码
32-
RUN git clone https://github.com/kvcache-ai/ktransformers.git
33-
# 清理 apt 缓存
34-
RUN rm -rf /var/lib/apt/lists/*
13+
WORKDIR /app/lora-scripts
14+
RUN pip install xformers==0.0.27.post2 --no-deps && pip install -r requirements.txt
3515

36-
# 进入项目目录
37-
WORKDIR /workspace/ktransformers
38-
# 初始化子模块
39-
RUN git submodule update --init --recursive
16+
WORKDIR /app/lora-scripts/scripts
17+
RUN pip install -r requirements.txt
4018

41-
# 升级 pip
42-
RUN pip install --upgrade pip && pip install -U wheel setuptools
19+
WORKDIR /app/lora-scripts
4320

44-
# 安装构建依赖
45-
RUN pip install ninja pyproject numpy cpufeature aiohttp zmq openai
46-
47-
# 安装 flash-attn(提前装可以避免后续某些编译依赖出错)
48-
RUN pip install flash-attn
49-
50-
# 安装 ktransformers 本体(含编译)
51-
RUN CPU_INSTRUCT=${CPU_INSTRUCT} \
52-
USE_BALANCE_SERVE=1 \
53-
KTRANSFORMERS_FORCE_BUILD=TRUE \
54-
TORCH_CUDA_ARCH_LIST="8.0;8.6;8.7;8.9;9.0+PTX" \
55-
pip install . --no-build-isolation --verbose
56-
57-
RUN pip install third_party/custom_flashinfer/
58-
# 清理 pip 缓存
59-
RUN pip cache purge
60-
61-
# 拷贝 C++ 运行时库
62-
RUN cp /usr/lib/x86_64-linux-gnu/libstdc++.so.6 /opt/conda/lib/
63-
64-
# 保持容器运行(调试用)
65-
ENTRYPOINT ["tail", "-f", "/dev/null"]
21+
CMD ["python", "gui.py", "--listen"]

0 commit comments

Comments
 (0)