Claude 2f5be7d635 fix: GDN NaN clamp (7 sites) + Dockerfile Step 4 tolerance ...

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)

2026-08-11 09:20:14 +00:00

..

flash_qla_sm70

fix(GDN): gate clamp [-5,0] (decay only) + state clamp ±100

2026-08-10 04:44:05 +00:00

prebuilt/corex-3.2.3-ivcore10

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

qwen3_5

[DEPLOY] Complete submission: baseline + all optimizations

2026-07-30 16:06:20 +00:00

qwen3_5_moe

[DEPLOY] Complete submission: baseline + all optimizations

2026-07-30 16:06:20 +00:00

reasoning

fix(critical): 3 fixes from sub508 diagnosis — n>1 crash guard + thinking format + content fallback

2026-08-07 07:55:04 +00:00

wheels

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

_custom_ops.py

fix(PROVEN): _moe_C compiles and runs on real BI-V100 hardware

2026-08-11 01:50:48 +00:00

api_server.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

arg_utils.py

[ARCH] Eliminate AST patch scripts — full file replacements only

2026-08-05 08:24:43 +00:00

bi100_env.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

bi100_profile.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

block_major_kv_cache.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_attn_head_rms_norm.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_fused_paged_prefill_split4.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_gdn_beta_decay.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_gdn_causal_conv.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_gdn_gated_norm.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_gdn_packed_decode.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_gdn_qk_map.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_moe_direct_routed.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_moe_exact_reduce.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_moe_weight_gather.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

build_corex_paged_kv_gather.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

chat_utils.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

cli_args.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_attn_head_rms_norm.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_block_major_kv_transfer.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_fused_paged_prefill_split4.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_gdn_beta_decay.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_gdn_causal_conv.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_gdn_gated_norm.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_gdn_packed_decode.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_gdn_qk_map.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_moe_direct_routed.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_moe_exact_reduce.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_moe_weight_gather.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_paged_kv_gather.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

corex_query_tiled_paged_prefill.cu

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

gdn_prefix.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

install_prebuilt_corex.sh

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

logits_processor.py

[ARCH] Eliminate AST patch scripts — full file replacements only

2026-08-05 08:24:43 +00:00

mamba_cache.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

model_runner.py

fix(engine): CCCL overflow_cast + checked_allocator patterns for NaN/OOM

2026-08-07 08:56:36 +00:00

paged_attention_v2_pytorch.py

[ENGINE] CCCL SmemResource pattern: pre-allocate staging buffers in V2 paged attention

2026-08-07 04:37:44 +00:00

paged_attn.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_block_major_cache_engine.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_block_major_worker_capacity.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_block_manager_cache_trace.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_corex_swap_blocks.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_executor_startup_debug.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_model_runner.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_numerical_stability.py

fix(cccl): robust fallback — inject module-level safety if regex misses

2026-08-07 09:58:35 +00:00

patch_ops.sh

fix(CRITICAL): merge 26e6cb40 build pipeline + HEAD features — fix docker build

2026-08-11 07:58:14 +00:00

patch_transformers_qwen3_5.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_utils.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_vllm_qwen3_5.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_vllm_tool_parser.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_worker_cache_transfer_order.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_worker_profile_override.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_worker_startup_profile_guard.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_xformers_profile.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_xformers_sdpa_batch_kernel.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_xformers_sdpa_batch.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_xformers_sdpa_seq_kernel.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

patch_xformers_sdpa_seq.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

precompile_gdn.py

build(SM70): precompile GDN CUDA kernel to .so during docker build

2026-08-10 01:08:38 +00:00

prefix_prefill.py

fix(critical): v2 kernel softmax normalization was commented out — outputs were wrong

2026-08-07 02:46:46 +00:00

probe_corex_api.py

fix(build): add .dockerignore + safe probe — fix docker build failure

2026-08-08 11:21:43 +00:00

protocol.py

fix(CRITICAL): docker build容错 + max_completion_tokens + extra=ignore + ix_unified bridge

2026-08-11 07:13:05 +00:00

qwen3_5_base_original.py

arch(CRITICAL): replace custom qwen3_5.py with base original (1369 lines)

2026-08-08 08:10:55 +00:00

qwen3_5.py

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

2026-08-11 09:20:14 +00:00

qwen3coder_tool_parser.py

refactor(tool_parser): CCCL agent_radix_sort_downsweep union TempStorage — phased state

2026-08-08 07:02:22 +00:00

registry.py

[ARCH] Eliminate 2 more patch scripts — registry.py + tool_parsers __init__.py

2026-08-05 08:36:52 +00:00

sampler.py

[BASE] qwen3_6_scripts/sampler.py: CCCL topk unsorted output optimization

2026-08-06 02:55:51 +00:00

scheduler.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

sequence.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

serving_chat.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

serving_tokenization.py

feat(CRITICAL): import wudixzy/competition complete corex stack — 12 prebuilt .so + 13 CUDA kernels + 2615-line qwen3_5.py

2026-08-11 03:55:38 +00:00

tool_parsers_init.py

[ARCH] Eliminate 2 more patch scripts — registry.py + tool_parsers __init__.py

2026-08-05 08:36:52 +00:00

verify_functional.py

fix(verify): remove duplicate TC-22~30 test definitions — CCCL test_then.cu audit

2026-08-07 06:36:12 +00:00

xformers.py

feat(xformers): replace Q-only tiling with Q+KV tiling + online softmax

2026-08-07 01:54:52 +00:00