biondizzle/vllm-deepseek-v32-mtp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

build expirimental cache fix

Browse Source
This commit is contained in:
biondizzle
2026-04-15 11:27:07 +00:00
parent 1c4fe39364
commit 1061388aaa
1 changed files with 2 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
Dockerfile
View File

@@ -1,20 +1,5 @@
FROM vllm/vllm-openai:v0.19.0-cu130
#FROM vllm/vllm-openai:v0.19.0-cu130
FROM atl.vultrcr.com/vllm/vllm-with-lmcache:v0.19.0-cu130-mla-multi-group-kv-cache
# Patch tool parser for GLM regex fix
COPY deepseekv32_tool_parser.py /usr/local/lib/python3.12/dist-packages/vllm/tool_parsers/deepseekv32_tool_parser.py
# Install LMCache for KV cache offloading / sharing across nodes
# Build with system CUDA 13.0 for Blackwell (B200)
RUN apt-get update && apt-get install -y git \
libcusolver-dev-13-0 \
libcusparse-dev-13-0 \
libcublas-dev-13-0 \
libcurand-dev-13-0 \
libcufft-dev-13-0 \
libnvjitlink-dev-13-0 && \
git clone --depth 1 https://github.com/LMCache/LMCache.git /tmp/lmcache && \
cd /tmp/lmcache && \
CUDA_HOME=/usr/local/cuda \
TORCH_CUDA_ARCH_LIST="10.0" \
pip install --no-cache-dir --no-build-isolation . && \
rm -rf /tmp/lmcache