dylanyunlong/project_6_89d52222
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Commit Graph

  • f265cb8ad3 fix(yaml): align launch params with comp 168 proven config project6-dev 2026-08-10 06:40:48 +00:00
  • 905bf4db2c feat(moe): wire silu_and_mul through C++ bridge in corex_moe.py project6-dev 2026-08-10 06:36:40 +00:00
  • 7127d18491 refactor(bridge): rewrite ix_full_bridge.cpp for actual base image symbols project6-dev 2026-08-10 06:34:54 +00:00
  • b9fd2755d9 debug: probe all ixformer symbol locations project6-dev 2026-08-10 06:33:14 +00:00
  • 3e7fc565ff debug: probe script to find silu_and_mul symbol location project6-dev 2026-08-10 06:32:04 +00:00
  • a54dbda3bb fix(bridge): link against libixformer.so for silu_and_mul symbol project6-dev 2026-08-10 06:29:11 +00:00
  • ac3c8e28eb fix(bridge): c10::nullopt → typed std::optional{} for CoreX torch compat project6-dev 2026-08-10 06:25:15 +00:00
  • ff0bf8c1d6 fix: total_mem → total_memory (torch API) project6-dev 2026-08-10 06:21:40 +00:00
  • c579c75039 test: single-GPU verification script for ix_full_bridge compile + MoE 7-step dispatch chain project6-dev 2026-08-10 06:21:04 +00:00
  • d7fa7b0682 docs: update PROJECT_SUMMARY.md — corex rewrite + dispatch chain analysis project6-dev 2026-08-10 06:17:38 +00:00
  • d86b39d1ae refactor(corex): rewrite 3 dlopen modules to use real ixformer::infer dispatch chain project6-dev 2026-08-10 06:16:39 +00:00
  • 33b7327c1d fix: add error logging to all corex module imports + dtype guard EX Engine 2026-08-10 04:51:16 +00:00
  • 7d4edd4ac7 fix(GDN): force fp16 cast before norm+out_proj — ixformer matmul requires kHalf project6-dev 2026-08-10 04:45:26 +00:00
  • d841c44e55 fix(GDN): dtype guard on _ix_matmul/_ix_bmm — ixformer.matmul requires kHalf EX Engine 2026-08-10 04:45:24 +00:00
  • 41aec33955 fix(GDN): gate clamp [-5,0] (decay only) + state clamp ±100 EX Engine 2026-08-10 04:44:05 +00:00
  • 1ae398eeee fix(interface): corex_moe accepts w13 merged format + no silent fallback EX Engine 2026-08-10 04:36:16 +00:00
  • 44d36e6ccc build: add MoE topk kernel precompile to patch_ops.sh + Dockerfile pipeline project6-dev 2026-08-10 04:26:33 +00:00
  • f32ef97013 feat(MoE): verified CUDA topk_softmax kernel — zero fallback project6-dev 2026-08-10 04:21:38 +00:00
  • 2238604bad test: verify_on_device.py — 真机逐函数验证, 不允许fallback EX Engine 2026-08-10 04:06:21 +00:00
  • f955dd127e feat(EX): ix_full_bridge — all 14 ixformer::infer functions bridged EX Engine 2026-08-10 04:01:35 +00:00
  • 5efb0fcc35 feat(EX): corex_fa2.py — third dlopen module from comp 168 AST chain project6-dev 2026-08-10 04:01:17 +00:00
  • f4e2264a83 ref(EX): import upstream ILU kernels + xllm MoE CUDA sources into ex_engine project6-dev 2026-08-10 03:59:37 +00:00
  • dba027fded fix(deploy): wire corex_gdn.py + corex_moe.py into patch_ops.sh EX Engine 2026-08-10 03:39:49 +00:00
  • 8eba1750fa fix(GDN): clamp gate [-5,2] + state [-65504,65504] to prevent inf/NaN EX Engine 2026-08-10 03:38:39 +00:00
  • 388f6b2d1a feat(MoE): wire full ix_fused_moe_forward as Tier 0 dispatch EX Engine 2026-08-10 03:36:39 +00:00
  • 1be9449883 feat(EX): corex_gdn + corex_moe — dlopen dispatch chain from comp 168 log analysis project6-dev 2026-08-10 03:37:15 +00:00
  • 7839982707 feat(EX): wire xllm CUB topk_softmax kernel into MoE routing EngineX 2026-08-10 03:02:54 +00:00
  • d00daa62f6 feat(MoE): full ixformer pipeline — topk → gen_idx → expand → group_gemm → silu → combine EX Engine 2026-08-10 03:06:00 +00:00
  • e04a3bace9 fix: fail-fast on ix_bridge failure + probe script for real machine EX Engine 2026-08-10 03:04:50 +00:00
  • d21b2505bb fix: wire MoE topk via ixformer C++ bridge + disable broken flash_qla GDN EX Engine 2026-08-10 03:00:24 +00:00
  • 8e6adf20e6 refactor(EX): upstream-aligned kernels + FlashQLA GDN backend EX Engine 2026-08-10 02:55:53 +00:00
  • 002f9879b2 ref(upstream): FULL TREE — Deep-Spark xllm (1470) + ds_vllm csrc/models (703) EX Engine 2026-08-10 02:53:54 +00:00
  • 9e4fb3712f remove enginex/ — duplicate of existing ex_engine/ EngineX 2026-08-10 02:49:32 +00:00
  • ea82b00e54 ref(upstream): add Deep-Spark xllm + vllm MoE/GDN reference code EX Engine 2026-08-10 02:48:17 +00:00
  • b4e055e9a9 feat(enginex): CCCL-style algorithm factor replacement engine — 18 operator dispatch system EngineX 2026-08-10 02:40:13 +00:00
  • b75965d4ea fix(EX): corex ivcore10 build flags + deploy pipeline + topk kernel cleanup EX Engine 2026-08-10 02:31:55 +00:00
  • fcfb764560 feat(EX): Algorithm Factor Replacement Engine — dlopen-based CUDA kernel dispatch EX Engine 2026-08-10 02:25:23 +00:00
  • 121432f8e9 doc: system design — architecture, file map, data flow, build pipeline Claude 2026-08-10 02:07:43 +00:00
  • c077736968 feat(SM70): wire up FlashQLA GDN kernel dispatch in prefill path Claude 2026-08-10 01:42:16 +00:00
  • 47958c4ed2 fix(build): add ninja dependency — required for CUDA kernel compilation Claude 2026-08-10 01:21:28 +00:00
  • 20cd2d8904 build(SM70): precompile GDN CUDA kernel to .so during docker build Claude 2026-08-10 01:08:38 +00:00
  • 8cf73ad39c feat(SM70): add 1Cat-vLLM FlashQLA fused GDN CUDA kernel for BI-V100 Claude 2026-08-10 01:06:32 +00:00
  • 3d5f75fefd fix(d05): remove image_url stripping — model IS multimodal Claude 2026-08-10 00:15:06 +00:00
  • 83d633798f fix(overflow): chunk_size 64→16 — CCCL counter overflow prevention Claude 2026-08-10 00:13:49 +00:00
  • 0a697f5871 arch(scan): dispatch_scan.cuh Phase 1/Phase 2 separation in GDN chunk loop Claude 2026-08-09 10:44:43 +00:00
  • a7537ebee0 doc: add functional test FAIL root cause analysis to CODEPATH_MAP Claude 2026-08-09 00:03:45 +00:00
  • e87470733d accel(ixformer): wire BI-V100 hardware primitives into GDN + MoE compute paths Claude 2026-08-08 22:35:21 +00:00
  • 5cd2780320 fix(CRITICAL): CCCL overflow guard — clamp before cumsum + max-num-seqs=2 Claude 2026-08-08 21:49:39 +00:00
  • 68876acd1b doc: BI-V100 hardware probe raw data (SSH Aug 8 2026) Claude 2026-08-08 18:18:48 +00:00
  • 6b8965a667 accel(ixformer): add BI-V100 hardware op wrappers + silence corex warnings Claude 2026-08-08 18:13:59 +00:00
  • 44003fa829 fix(probe): replace Python probe with direct shell — guaranteed build log output Claude 2026-08-08 15:09:31 +00:00
  • ff971686d4 fix(CRITICAL): max_model_len 100000→80000 (KV cache only 88112) + NaN fix Claude 2026-08-08 15:08:00 +00:00
  • c1065aaf2c fix(build): add .dockerignore + safe probe — fix docker build failure Claude 2026-08-08 11:21:43 +00:00
  • dbfe20fd1c arch(probe): add build-time CoreX API discovery — stop guessing interfaces Claude 2026-08-08 11:16:58 +00:00
  • ee09550263 arch(CoreX): CCCL env_dispatch — try native fused kernels, fallback PyTorch Claude 2026-08-08 11:15:04 +00:00
  • fb2ddb843e fix(patch_ops): correct contradictory deploy log messages Claude 2026-08-08 11:07:41 +00:00
  • 80fa1fe781 arch(CRITICAL): match Sub168 proven engine config exactly Claude 2026-08-08 11:01:52 +00:00
  • 1fed1bc051 fix: add --max-seq-len-to-capture 32768, fix patch_ops.sh contradictory comments Claude 2026-08-08 10:57:02 +00:00
  • d221383fc0 doc(prd): complete base engine migration checklist Claude 2026-08-08 10:48:43 +00:00
  • 6bed911e04 fix(patch_ops): add pip install transformers==4.55.3 from base engine Claude 2026-08-08 10:48:24 +00:00
  • e0fe46a46f arch(CRITICAL): deploy ALL base engine patches — paged_attn, xformers, sequence, scheduler Claude 2026-08-08 10:48:01 +00:00
  • abd3d5640a arch(CRITICAL): replace custom qwen3_5.py with base original (1369 lines) Claude 2026-08-08 08:10:55 +00:00
  • 4daa30a267 fix(CRITICAL): CCCL kernel_segmented_scan — kill nan_to_num, add corex_gdn/corex_moe dispatch Claude 2026-08-08 08:08:52 +00:00
  • c077f7fd40 doc(prd): add block_reduce + exception mapping records Claude 2026-08-08 08:02:31 +00:00
  • e832687893 arch(qwen3_5): dispatch_segmented_sort three-way dispatch — try native CoreX before PyTorch fallback Claude 2026-08-08 08:01:57 +00:00
  • d44ec4d8db perf(qwen3_5): CCCL block_reduce_warp_reductions → reduce DeltaNet loop iterations Claude 2026-08-08 08:01:56 +00:00
  • 4698cd5687 doc(prd): CCCL tuning_batched_topk → sampling strategy mapping Claude 2026-08-08 07:54:46 +00:00
  • 53d816b154 doc(prd): CCCL agent_rle + adjacent_difference → streaming mapping Claude 2026-08-08 07:53:35 +00:00
  • c59b529315 fix(CRITICAL): deploy qwen3_5.py — ModuleNotFoundError kills startup Claude 2026-08-08 07:53:14 +00:00
  • 87cc24b819 doc(prd): CCCL tuning_select_if.cuh complete design → serving_chat.py mapping Claude 2026-08-08 07:52:02 +00:00
  • 85f3240c98 doc(prd): create PRD with CCCL→base mapping table and competition strategy Claude 2026-08-08 07:38:15 +00:00
  • ef540b6f9f fix(serving): CCCL completion_mechanism — remove NaN-era max_tokens cap Claude 2026-08-08 07:37:45 +00:00
  • 68be2ff856 fix(dispatch): radix_sort-inspired size-dispatch — disable thinking for small max_tokens, clamp oversized max_tokens Claude 2026-08-08 07:36:25 +00:00
  • e37b4d283b env(yaml): CCCL buddy_allocator pattern — PYTORCH_CUDA_ALLOC_CONF + OMP_NUM_THREADS Claude 2026-08-08 07:36:03 +00:00
  • b271210af4 fix(critical): allow n=2 to match Sub168 — max_num_seqs=2 in yaml Claude 2026-08-08 07:32:55 +00:00
  • e86b7eacdd refactor(tool_parser): CCCL agent_radix_sort_downsweep union TempStorage — phased state Claude 2026-08-08 07:02:22 +00:00
  • 5a3831e977 refactor(api): CCCL tuning_adjacent_difference policy_selector — dynamic error handling Claude 2026-08-08 07:01:37 +00:00
  • 8be95e422d perf(protocol): CCCL agent_for consume_tile<IsFullTile> — fast path for standard messages Claude 2026-08-08 07:01:05 +00:00
  • 2ea7a19f73 refactor(serving): CCCL dispatch_rle streaming_context pattern — unify per-choice state Claude 2026-08-08 07:00:29 +00:00
  • cafd34fe4a fix(CRITICAL): REVERT to serving-only patches — Sub168 proves CoreX native model is correct Claude 2026-08-08 05:56:58 +00:00
  • 5b8d7c6b76 arch(critical): deploy ALL customized files to container — qwen3_5.py was NEVER running Claude 2026-08-08 05:37:46 +00:00
  • 810aef8c39 fix(critical): match Sub168 proven config — max_model_len=100K, max_num_seqs=1, gpu_mem=0.9 Claude 2026-08-08 05:37:40 +00:00
  • 803e888ae9 fix(build): crash-proof patch_ops.sh — remove set -e, all ops non-fatal project6 2026-08-07 10:37:41 +00:00
  • 2680d62ec8 fix(critical): match Sub168 config exactly + disable risky numerical patch project6 2026-08-07 10:27:53 +00:00
  • 1ba0dd3966 arch(cccl): match Sub168 proven config + bench.py timeout pattern project6-dev 2026-08-07 10:02:53 +00:00
  • 3342d18bcc fix(critical): remove pip install transformers — was breaking corex kernel loading project6 2026-08-07 10:02:09 +00:00
  • 4e04674283 fix(cccl): robust fallback — inject module-level safety if regex misses project6 2026-08-07 09:58:35 +00:00
  • 7153029974 perf(cccl): thread_reduce fast-path — cap tool_call max_tokens to 2048 project6 2026-08-07 09:56:32 +00:00
  • 9870d07073 fix(critical): CCCL-inspired graceful degradation — prevent OOM cascade project6-dev 2026-08-07 09:54:57 +00:00
  • b47a5d4b95 arch(cccl): Agent-pattern numerical stability patch + protocol required fix project6 2026-08-07 09:54:55 +00:00
  • 4eea584c9d fix(deltanet): systematic NaN elimination via CCCL overflow_cast pattern Claude 2026-08-07 09:26:03 +00:00
  • a20e8614a4 fix(critical): stop replacing base image compute files — use corex native kernels Claude 2026-08-07 09:21:43 +00:00
  • 2102146c01 refactor(moe): translate thrust mode.cu pipeline — unique_consecutive replaces manual boundary detect project6 2026-08-07 09:17:23 +00:00
  • 4d9165fa30 arch(tuning): translate CCCL tuning_transform_tile.cuh — derive chunk sizes from hardware project6 2026-08-07 09:15:56 +00:00
  • f140825a56 arch(moe): translate CCCL sync_handler.cuh — register-at-init, resolve-on-first-call project6 2026-08-07 09:14:49 +00:00
  • 43ede018a1 perf(moe): translate CCCL smem_resource_raw.cuh — buffer reuse for MoE output project6 2026-08-07 09:13:32 +00:00
  • c1936a55cb arch(moe): translate CCCL block_histogram.cuh — segment size histogram for expert load analysis project6 2026-08-07 09:10:17 +00:00
  • 83192486d3 perf(deltanet): CCCL thrust::all_of early termination for NaN detection project6 2026-08-07 09:10:01 +00:00
  • 2e2a479c08 perf(moe): translate CCCL dispatch_copy_mdspan.cuh — contiguous slice fast path project6 2026-08-07 09:09:20 +00:00
  • be630106b2 perf(deltanet): CCCL block_scan RAKING_MEMOIZE — precompute all exp() outside loop project6 2026-08-07 09:08:53 +00:00