dylanyunlong
0 Followers · 0 Following
  • Joined on 2026-08-06
Repositories
6
 Projects Public Activity Starred Repositories
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-15 00:35:40 +08:00
3bee73207e fix: add cuda_runtime.h to hgemm_bind.cpp for cudaStream_t
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-15 00:25:59 +08:00
09e5261ba6 refactor: hgemm_blocktiling.cu — strict 1:1 from siboehm kernel 6
ab42fc1fd7 feat: hgemm_blocktiling.cu — FP16 GEMM kernel for MoE expert dispatch on BI-V100
Compare 2 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 23:15:24 +08:00
9ca33cf4d5 upstream: add GEMM kernel references from 4 repos for BI-V100 porting
29ecc2e602 feat: moe_expert_gemm.cpp — C++ loop over experts via ixformer_linear (replaces Python for-loop)
0ace44e293 test: cat ixinfer.h + functional batched GEMM test on BI-V100
Compare 3 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 20:05:50 +08:00
bfc4de2cf3 test: probe cuinfer + cublasLt + cublas batched GEMM APIs for MoE expert dispatch
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 20:02:30 +08:00
d6958070cb test: probe CUTLASS + __CUDA_ARCH__ on BI-V100 for grouped GEMM feasibility
50a249e0a3 Revert "feat: batched MoE expert GEMM — replaces Python for-loop"
06d7713db6 feat: batched MoE expert GEMM — replaces Python for-loop
Compare 3 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:40:47 +08:00
93353a1414 test: probe ixformer .so symbols for MoE grouped GEMM
865c18f852 feat: integrate xllm_moe into qwen3_5.py MoE hot path
Compare 2 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:34:29 +08:00
e147c283e3 prebuilt: 5 xllm CUDA kernel .so for BI-V100 (ivcore10)
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:32:37 +08:00
31d3ee99bb fix: MoE kernel include paths — device_utils.cuh + arch_condition.h
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:29:58 +08:00
df6a0f5d47 fix: remove cuda/functional from MoE topk kernels (not available on corex)
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:27:19 +08:00
a50adefdfc feat: xllm MoE CUDA kernels — fused_topk + compute_index + combine
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:21:20 +08:00
49cd7def89 fix: slot_ids dtype long→int32 in cache test
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:19:37 +08:00
1f51feee05 fix: cos_sin_cache dtype float→half in rope test
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:17:17 +08:00
7fc0c1defa test: AST call chain verification for all 4 xllm kernel .so
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:11:58 +08:00
3d816cd18d fix: add ceil_div + DEVICE_INLINE to device_utils.cuh
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:08:37 +08:00
302aa9608a fix: block_copy.cu — DEVICE_INLINE, CHECK_EQ→TORCH_CHECK, cstdint
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:04:00 +08:00
900ae0b1ef fix: block_copy.cu remove utils.h (glog), CHECK→TORCH_CHECK
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 19:00:43 +08:00
a206fc1d43 fix: activation.cu torch/extension.h + LOG(FATAL)→TORCH_CHECK, reshape_paged_cache.cu torch header
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 18:50:53 +08:00
093bfb380f feat: pybind11 bindings for xllm CUDA kernels
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 18:47:57 +08:00
415fff85f1 fix: add DISPATCH_FLOATING_TYPES macro to device_utils.cuh
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-14 18:43:14 +08:00
0359103b9b fix: remove glog/cuda_ops_api.h dependency from all xllm CUDA kernels