dylanyunlong/project_6
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix: GDN NaN clamp (7 sites) + Dockerfile Step 4 tolerance

Browse Source 
qwen3_5.py (2642 lines, 12 prebuilt .so, no fallback):
- decay_mask: g_diff.clamp(-20,20) before exp()
- Neumann row: .clamp(-65504,65504) on iterative update
- k_cumdecay: g.clamp(-20,20).exp()
- state loop attn_inter: g.clamp(-20,20).exp()
- state loop g_exp_term: .clamp(-20,20)
- state loop g_last: .clamp(-20,20)
- state loop last_state: .clamp(-65504,65504) after update

Dockerfile Step 4: each cp gets 2>/dev/null || true
(matches tolerance pattern of Steps 1-3, 5-8)
This commit is contained in:
Claude
2026-08-11 09:20:14 +00:00
parent ed8bdf8714
commit 2f5be7d635
2 changed files with 22 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
Dockerfile
View File

@@ -24,26 +24,27 @@ RUN python3 /workspace/ex_engine/precompile_moe_kernels.py 2>&1 | tee -a /worksp
echo "[Dockerfile] moe_v055 precompile exit code: $?"
# Step 4: Stage vendor_overrides into qwen3_6_scripts/ so patch_ops.sh finds them
RUN mkdir -p /workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block && \
mkdir -p /workspace/qwen3_6_scripts/vendor_overrides/vllm/model_executor/layers && \
RUN mkdir -p /workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/model_executor/layers && \
cp /workspace/vllm_overrides/core/evictor_v2.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/evictor_v2.py && \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/evictor_v2.py 2>/dev/null || true && \
cp /workspace/vllm_overrides/core/block/cpu_kv_content_cache.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block/cpu_kv_content_cache.py && \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block/cpu_kv_content_cache.py 2>/dev/null || true && \
cp /workspace/vllm_overrides/core/block/cpu_gpu_block_allocator.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block/cpu_gpu_block_allocator.py && \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block/cpu_gpu_block_allocator.py 2>/dev/null || true && \
cp /workspace/vllm_overrides/core/block/prefix_caching_block.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block/prefix_caching_block.py && \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block/prefix_caching_block.py 2>/dev/null || true && \
cp /workspace/vllm_overrides/core/block/block_table.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block/block_table.py && \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block/block_table.py 2>/dev/null || true && \
cp /workspace/vllm_overrides/core/block_manager_v2.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block_manager_v2.py && \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/core/block_manager_v2.py 2>/dev/null || true && \
cp /workspace/vllm_overrides/sampling_params.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/sampling_params.py && \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/sampling_params.py 2>/dev/null || true && \
cp /workspace/vllm_overrides/model_executor/sampling_metadata.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/model_executor/sampling_metadata.py && \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/model_executor/sampling_metadata.py 2>/dev/null || true && \
cp /workspace/vllm_overrides/model_executor/layers/sampler.py \
/workspace/qwen3_6_scripts/vendor_overrides/vllm/model_executor/layers/sampler.py
/workspace/qwen3_6_scripts/vendor_overrides/vllm/model_executor/layers/sampler.py 2>/dev/null || true ; \
echo "[Dockerfile] vendor_overrides staged"
# Step 5: Deploy patches (serving + engine fixes + prebuilt .so)
RUN chmod +x /workspace/qwen3_6_scripts/patch_ops.sh && \