Elvir Crnčević 9f04d9d55f

[Qwen3-Next] MoE configs for H100 TP=1,2 and TP2/EP (#24739) ...

Signed-off-by: elvircrn <elvircrn@gmail.com>

2025-09-12 07:54:04 -07:00

..

configs

[Qwen3-Next] MoE configs for H100 TP=1,2 and TP2/EP (#24739)

2025-09-12 07:54:04 -07:00

__init__.py

[Kernels] Clean up FusedMoeMethodBase and modular kernel setup. Remove extra arguments from modular kernel methods. (#22035)

2025-08-15 14:46:00 -04:00

batched_deep_gemm_moe.py

[Feature] Add Hopper DeepGEMM E8M0 for DeepSeekV3.1 scale_fmt (#23666)

2025-08-27 14:09:08 +00:00

batched_triton_or_deep_gemm_moe.py

[Kernels] Clean up FusedMoeMethodBase and modular kernel setup. Remove extra arguments from modular kernel methods. (#22035)

2025-08-15 14:46:00 -04:00

config.py

[Model][gpt-oss] Support DP+EP for GPT-OSS with FlashInfer trtllm-gen MoE (#23819)

2025-08-28 06:56:20 -07:00

cpu_fused_moe.py

Add routed_scaling_factor to MoE grouped topk (#23123)

2025-08-29 21:36:48 -07:00

cutlass_moe.py

[Kernel] CUTLASS MoE FP8: Integrate cuda moe permute/unpermute (#23045)

2025-08-20 10:35:26 -04:00

deep_gemm_moe.py

[Misc] Slight improve deepgemm print (#24085)

2025-09-04 16:06:51 +00:00

deep_gemm_utils.py

[Bugfix] DeepGemm utils : Fix hardcoded type-cast (#21517)

2025-07-24 20:17:29 -07:00

deepep_ht_prepare_finalize.py

[Bugfix] Fix DeepEP config for DP4TP4 (#23619)

2025-09-10 10:37:56 -07:00

deepep_ll_prepare_finalize.py

[Kernels] Overlap shared experts with send/recv (#23273)

2025-09-03 12:35:18 -04:00

flashinfer_cutlass_moe.py

[NVIDIA] Add SM100 Flashinfer Cutlass MoE fp8 backend (#22357)

2025-08-19 18:01:53 -04:00

flashinfer_cutlass_prepare_finalize.py

[Kernels] Overlap shared experts with send/recv (#23273)

2025-09-03 12:35:18 -04:00

fused_batched_moe.py

[Kernels] Overlap shared experts with send/recv (#23273)

2025-09-03 12:35:18 -04:00

fused_marlin_moe.py

[Kernel] Add cuda kernel for gpt_oss activation (#22951)

2025-08-17 05:03:24 +00:00

fused_moe.py

[Doc]: fix typos in Python comments (#24294)

2025-09-05 19:41:12 -07:00

gpt_oss_triton_kernels_moe.py

[Kernels] Clean up FusedMoeMethodBase and modular kernel setup. Remove extra arguments from modular kernel methods. (#22035)

2025-08-15 14:46:00 -04:00

layer.py

[Kernel][B200] mxfp4 fused cutlass moe (#23696)

2025-09-11 17:04:56 -04:00

modular_kernel.py

[Doc]: fix typos in Python comments (#24417)

2025-09-08 00:22:16 -07:00

moe_align_block_size.py

[Refactor] Remove moe_align_block_size_triton (#21335)

2025-07-26 07:08:29 -07:00

moe_pallas.py

[Doc]: fix typos in Python comments (#24026)

2025-09-01 09:38:20 +00:00

moe_permute_unpermute.py

[Kernel] CUTLASS MoE FP8: Integrate cuda moe permute/unpermute (#23045)

2025-08-20 10:35:26 -04:00

moe_torch_iterative.py

[Misc] Add SPDX-FileCopyrightText (#19100)

2025-06-03 11:20:17 -07:00

pplx_prepare_finalize.py

[Kernels] Overlap shared experts with send/recv (#23273)

2025-09-03 12:35:18 -04:00

prepare_finalize.py

[Kernels] Overlap shared experts with send/recv (#23273)

2025-09-03 12:35:18 -04:00

rocm_aiter_fused_moe.py

[Docs] Fix warnings in mkdocs build (continued) (#24092)

2025-09-10 06:23:28 -07:00

routing_simulator.py

[Docs] Fix warnings in mkdocs build (continued) (#24092)

2025-09-10 06:23:28 -07:00

topk_weight_and_reduce.py

[Misc] ModularKernel : Perform WeightAndReduce inside TritonExperts & DeepGemmExperts (#20725)

2025-07-14 19:47:16 +00:00

triton_deep_gemm_moe.py

[Feature] Add Hopper DeepGEMM E8M0 for DeepSeekV3.1 scale_fmt (#23666)

2025-08-27 14:09:08 +00:00

trtllm_moe.py

DP/EP Support for gpt-oss with deepep-ht comm kernel on SM100 (#23608)

2025-08-27 17:33:21 -04:00

utils.py

DP/EP Support for gpt-oss with deepep-ht comm kernel on SM100 (#23608)

2025-08-27 17:33:21 -04:00