i-peixingyu/K100-vllm
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

添加 K100-vLLM-Patched-v2.0/Dockerfile

Browse Source
This commit is contained in:
i-peixingyu
2026-05-19 18:36:27 +08:00
parent 0498c437e5
commit ad2f32a585
1 changed files with 21 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
K100-vLLM-Patched-v2.0/Dockerfile Normal file
View File

@@ -0,0 +1,21 @@
FROM git.modelhub.org.cn:9443/enginex-hygon/vllm:0.9.2
# 修复1 transformers 库级 patch
COPY patch.py /tmp/patch.py
RUN python3 /tmp/patch.py
# 修复2 ROCm 平台 Triton backend head_size 验证绕过
COPY patch_triton.py /tmp/patch_triton.py
RUN python3 /tmp/patch_triton.py
# 修复3 运行时 tokenizer 配置修复脚本
COPY detect_tokenizer.py /opt/detect_tokenizer.py
COPY fix_tokenizer.py /opt/fix_tokenizer.py
# 修复4 head_size 检测，自动切换 attention backend
COPY detect_head_size.py /opt/detect_head_size.py
COPY entrypoint.sh /opt/entrypoint.sh
RUN chmod +x /opt/entrypoint.sh
ENTRYPOINT ["/opt/entrypoint.sh"]