49 lines
3.2 KiB
Docker
49 lines
3.2 KiB
Docker
# enginex-bi100-compat
|
||
#
|
||
# 天数智芯 天垓100（Iluvatar_bi-100）· 文本生成 · vLLM「兼容增强」引擎。
|
||
#
|
||
# 不动模型、不动卡，只把「引擎镜像层」的结构性不兼容在容器启动前修掉。
|
||
# 做法照搬社区已上线引擎 zhengzhongwei/vLLM-Iluvatar_bi-100-opt：
|
||
# **把镜像内的 vllm 二进制换成本仓的 wrapper**，平台原样下发
|
||
# `vllm serve /model --port 80 ...` 就会流经 wrapper，无需平台改任何命令。
|
||
#
|
||
# 修什么（全部有日志实证，见 README.md）：
|
||
# R3 extra_special_tokens 是 list -> transformers 崩（社区已上线修复，已合并）
|
||
# R3b tokenizer_class 是坏类名 -> 加载异常（社区已上线修复，已合并）
|
||
# R2 模型没有 chat_template -> /v1/chat/completions 400/空输出
|
||
# R4 architectures 是镜像未注册类名 -> KeyError / MODEL_NOT_SUPPORTED
|
||
# R7 镜像缺 ixformer.contrib.vllm.layers -> ModuleNotFoundError（MoE）
|
||
#
|
||
# 基线镜像与 zhengzhongwei/vLLM-Iluvatar_bi-100-opt 一致（平台 bi-100 vllm 框架在用的那个），
|
||
# 保证替换 vllm 二进制的路径与行为可预期。
|
||
|
||
FROM harbor-contest.4pd.io/luopingyi/bi100/vllm:0.6.3
|
||
|
||
LABEL org.opencontainers.image.title="enginex-bi100-compat" \
|
||
org.opencontainers.image.description="Iluvatar bi-100 vLLM engine with preflight compatibility patches (R2/R3/R3b/R4/R7)" \
|
||
com.modelhubxc.engine.target-card="Iluvatar_bi-100" \
|
||
com.modelhubxc.engine.framework="vllm" \
|
||
com.modelhubxc.engine.task-type="text-generation" \
|
||
com.modelhubxc.engine.pattern="zhengzhongwei/vLLM-Iluvatar_bi-100-opt"
|
||
|
||
# 修补脚本 + R7 纯 PyTorch shim（都是小文本文件，无权重）
|
||
COPY preflight.py /opt/
|
||
COPY detect_tokenizer.py /opt/
|
||
COPY vllm_wrapper.sh /opt/
|
||
COPY shims_ixformer_layers.py /opt/shims_src/
|
||
|
||
# 组装 ixformer.contrib.vllm.layers 包（R7 兜底，只有真缺模块时才会被 import）
|
||
RUN set -eux && mkdir -p /opt/shims/ixformer/contrib/vllm/layers && for d in /opt/shims/ixformer /opt/shims/ixformer/contrib /opt/shims/ixformer/contrib/vllm /opt/shims/ixformer/contrib/vllm/layers; do printf '# ModelHub XC compat shim package\n' > "$d/__init__.py"; done && cp /opt/shims_src/shims_ixformer_layers.py /opt/shims/ixformer/contrib/vllm/layers/__init__.py && rm -rf /opt/shims_src && chmod +x /opt/vllm_wrapper.sh /opt/preflight.py && python3 -c "import ast,io;[ast.parse(io.open(f,encoding='utf-8').read()) for f in ['/opt/preflight.py','/opt/detect_tokenizer.py','/opt/shims/ixformer/contrib/vllm/layers/__init__.py']]" && bash -n /opt/vllm_wrapper.sh && echo "[enginex-bi100-compat] preflight syntax OK"
|
||
|
||
# R7：让 shims 全局可 import
|
||
ENV PYTHONPATH=/opt/shims:${PYTHONPATH}
|
||
ENV MODEL_DIR=/model
|
||
|
||
# 关键一步：把 vllm 二进制换成 wrapper（真身改名 vllm_real），
|
||
# 平台下发的 `vllm serve ...` 因此流经本引擎的启动前修补。
|
||
RUN set -eux && VLLM_BIN="$(command -v vllm || echo /usr/local/corex/lib64/python3/dist-packages/bin/vllm)" && \
|
||
if [ ! -f "$VLLM_BIN" ]; then echo "vllm binary not found at $VLLM_BIN"; exit 1; fi && \
|
||
mv "$VLLM_BIN" "${VLLM_BIN}_real" && \
|
||
ln -s /opt/vllm_wrapper.sh "$VLLM_BIN" && \
|
||
echo "[enginex-bi100-compat] vllm -> wrapper installed"
|