fxyfxy777 39ff38aba1 [OP]Unify MoE op with moe_permute path for bf16 GLM (#7164)

2026-04-09 16:17:56 +08:00

..

__init__.py

…

ep.py

[RL] support moe-topk use topk_reduce_func (#7218)

2026-04-09 11:01:03 +08:00

flashinfer_cutedsl_moe.py

[BugFix] fix flashinfer-cutedsl moe nvfp4 (#7120)

2026-04-03 15:43:19 +08:00

fused_moe_backend_base.py

[Feature] support compute shared experts before combine for better overlap (#6697)

2026-03-17 15:18:51 +08:00

fused_moe_blackwell_backend.py

[Feature] support blackwell gemm in ht (#7053)

2026-04-07 19:52:51 +08:00

fused_moe_cutlass_backend.py

[OP]Unify MoE op with moe_permute path for bf16 GLM (#7164)

2026-04-09 16:17:56 +08:00

fused_moe_deepgemm_backend.py

[OP]Unify MoE op with moe_permute path for bf16 GLM (#7164)

2026-04-09 16:17:56 +08:00

fused_moe_marlin_backend.py

[Feature] support compute shared experts before combine for better overlap (#6697)

2026-03-17 15:18:51 +08:00

fused_moe_triton_backend.py

[Feature] support blackwell gemm in ht (#7053)

2026-04-07 19:52:51 +08:00

fused_moe_wint2_backend.py

[Feature] support compute shared experts before combine for better overlap (#6697)

2026-03-17 15:18:51 +08:00

moe.py

[RL] support moe-topk use topk_reduce_func (#7218)

2026-04-09 11:01:03 +08:00

routing_indices_cache.py

[RL] add stream guard (#6814)

2026-03-13 11:22:26 +08:00

triton_moe_kernels.py

[Optimization] Use a separate driver when using Triton with Paddle (#6897)

2026-03-24 10:56:00 +08:00