fix(CRITICAL): Dockerfile回退到26e6cb40完全一致的结构——5个RUN step
之前新增的Step 5(bridge build)/Step 6(deploy)/Step 7(gdn)结构 在竞赛平台Docker build失败。真机上全部exit 0但竞赛平台环境不同。 回退到26e6cb40已验证能通过的Dockerfile结构: Step 1: build.sh Step 2: precompile_moe_topk Step 3: precompile_moe_kernels Step 4: patch_ops.sh (已包含bridge build + deploy逻辑) Step 5: precompile_gdn 所有增强功能(bridge build/deploy/ix_unified等)通过patch_ops.sh 在Step 4中完成,不需要额外的Dockerfile RUN step。
This commit is contained in:
52
Dockerfile
52
Dockerfile
@@ -8,44 +8,26 @@ COPY ./qwen3_6_scripts /workspace/qwen3_6_scripts
|
||||
COPY ./computility-run.yaml /workspace/computility-run.yaml
|
||||
COPY ./ex_engine /workspace/ex_engine
|
||||
|
||||
# Step 1: Build EX Engine .so libraries (tolerant)
|
||||
# Step 1: Build EX Engine .so libraries
|
||||
RUN chmod +x /workspace/ex_engine/build.sh && \
|
||||
(bash /workspace/ex_engine/build.sh --corex 2>&1 || true) | tee /workspace/ex_build.log ; \
|
||||
echo "[Dockerfile] ex_engine build done"
|
||||
bash /workspace/ex_engine/build.sh --corex 2>&1 | tee /workspace/ex_build.log ; \
|
||||
echo "[Dockerfile] ex_engine build exit code: $?"
|
||||
|
||||
# Step 2: Precompile MoE CUDA kernels (tolerant)
|
||||
RUN (python3 /workspace/ex_engine/precompile_moe_topk.py 2>&1 || true) | tee -a /workspace/ex_build.log ; \
|
||||
echo "[Dockerfile] moe_topk precompile done"
|
||||
# Step 2: Precompile MoE CUDA kernels
|
||||
RUN python3 /workspace/ex_engine/precompile_moe_topk.py 2>&1 | tee -a /workspace/ex_build.log ; \
|
||||
echo "[Dockerfile] moe_topk precompile exit code: $?"
|
||||
|
||||
# Step 3: Precompile vllm v0.5.5 MoE kernels (tolerant)
|
||||
RUN (python3 /workspace/ex_engine/precompile_moe_kernels.py 2>&1 || true) | tee -a /workspace/ex_build.log ; \
|
||||
echo "[Dockerfile] moe_v055 precompile done"
|
||||
# Step 3: Precompile vllm v0.5.5 MoE kernels
|
||||
RUN python3 /workspace/ex_engine/precompile_moe_kernels.py 2>&1 | tee -a /workspace/ex_build.log ; \
|
||||
echo "[Dockerfile] moe_v055 precompile exit code: $?"
|
||||
|
||||
# Step 4: Deploy patches (serving + engine fixes) — tolerant
|
||||
# Step 4: Deploy patches (serving + engine fixes)
|
||||
# patch_ops.sh also builds bridge, deploys .so and Python modules
|
||||
RUN chmod +x /workspace/qwen3_6_scripts/patch_ops.sh && \
|
||||
(cd /workspace/qwen3_6_scripts && bash ./patch_ops.sh 2>&1 || true) | tee /workspace/patch_ops.log ; \
|
||||
echo "[Dockerfile] patch_ops done"
|
||||
bash /workspace/qwen3_6_scripts/patch_ops.sh 2>&1 | tee /workspace/patch_ops.log ; \
|
||||
echo "[Dockerfile] patch_ops exit code: $?"
|
||||
|
||||
# Step 5: Build ix_unified_bridge.so (tolerant — ixformer symbols resolved at runtime)
|
||||
RUN if [ -f /workspace/ex_engine/build_unified_bridge.sh ]; then \
|
||||
chmod +x /workspace/ex_engine/build_unified_bridge.sh && \
|
||||
(bash /workspace/ex_engine/build_unified_bridge.sh 2>&1 || true) | tee -a /workspace/ex_build.log ; \
|
||||
fi ; \
|
||||
echo "[Dockerfile] bridge build done"
|
||||
|
||||
# Step 6: Deploy ix_unified Python modules to vllm path (tolerant)
|
||||
RUN VLLM_ROOT="$(python3 -c 'import vllm; print(vllm.__path__[0])' 2>/dev/null | grep -v '^INFO\|^WARNING\|^DEBUG' | tail -1)" ; \
|
||||
if [ -z "$VLLM_ROOT" ]; then VLLM_ROOT="/usr/local/corex/lib64/python3/dist-packages/vllm"; fi ; \
|
||||
echo "[Dockerfile] VLLM_ROOT=${VLLM_ROOT}" && \
|
||||
for f in ix_unified.py corex_so_loader.py moe_fused_dispatch.py corex_moe.py; do \
|
||||
[ -f "/workspace/ex_engine/python/$f" ] && cp "/workspace/ex_engine/python/$f" "${VLLM_ROOT}/$f" 2>/dev/null || true ; \
|
||||
done ; \
|
||||
for _so in /workspace/ex_engine/build/ix_unified_bridge*.so; do \
|
||||
[ -f "$_so" ] && cp "$_so" "${VLLM_ROOT}/" 2>/dev/null || true ; \
|
||||
done ; \
|
||||
echo "[Dockerfile] bridge deploy done"
|
||||
|
||||
# Step 7: Precompile GDN kernel (tolerant)
|
||||
RUN (python3 /workspace/qwen3_6_scripts/precompile_gdn.py \
|
||||
/workspace/qwen3_6_scripts/flash_qla_sm70 2>&1 || true) | tee -a /workspace/ex_build.log ; \
|
||||
echo "[Dockerfile] gdn precompile done"
|
||||
# Step 5: Precompile GDN kernel (needs vllm in path, so after patch_ops)
|
||||
RUN python3 /workspace/qwen3_6_scripts/precompile_gdn.py \
|
||||
/workspace/qwen3_6_scripts/flash_qla_sm70 2>&1 | tee -a /workspace/ex_build.log ; \
|
||||
echo "[Dockerfile] gdn precompile exit code: $?"
|
||||
|
||||
Reference in New Issue
Block a user