3 people 68fc395554 [Bugfix] Keep GateLinear batch invariant under VLLM_BATCH_INVARIANT (#59985) ...

Signed-off-by: yifanFengg <evanfengff@gmail.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>

2026-10-13 15:43:54 -04:00

..

configs

[XPU][MoE] Tune Triton fused MoE for Intel XPU (#53065)

2026-10-08 00:02:13 +08:00

experts

[Kernel] Fuse DeepEP-v2 decode globalize+align into one launch (#53911)

2026-10-13 11:10:38 -04:00

oracle

[Bugfix] Fall back to auto for unquantized MoE layers when --moe-backend is marlin (#60219)

2026-10-13 08:48:13 +00:00

prepare_finalize

[Kernel] Fuse DeepEP-v2 decode globalize+align into one launch (#53911)

2026-10-13 11:10:38 -04:00

router

[Bugfix] Keep GateLinear batch invariant under VLLM_BATCH_INVARIANT (#59985)

2026-10-13 15:43:54 -04:00

runner

[Bugfix][ModelLoader] Make enable_weights_track check quantized layers (#57108)

2026-10-10 10:44:06 +08:00

__init__.py

[ROCm][MoE] Split AiterExperts by weight-quantization key (#60322)

2026-10-13 00:13:33 -05:00

activation.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

all2all_utils.py

[feat] FlashInfer CuteDSL MegaMoE integration (#54049)

2026-10-02 15:43:33 -04:00

b12x.py

[Quantization] Add ModelOpt IQ2_XS, IQ2_XXS, and Q8_0 support via b12x (#59604)

2026-10-13 09:06:33 -07:00

config.py

Add MoNe experts pruning support (#50415)

2026-10-12 23:18:43 +00:00

deep_gemm_mega_moe.py

[Model][MoE] Load MegaMoE experts directly into the DeepGEMM layout (#60096)

2026-10-13 11:52:08 -07:00

deep_gemm_utils.py

[Bugfix] Widen ep_gather output offset to int64 (#58373)

2026-10-07 01:45:04 +00:00

eep_reconfigure.py

[Elastic EP] Reuse CUDA graphs across reconfiguration (#54985)

2026-09-12 10:26:39 -04:00

expert_map_manager.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

expert_substitution.py

Add MoNe experts pruning support (#50415)

2026-10-12 23:18:43 +00:00

flashinfer_moe_ep.py

[Bugfix] Drop enable_in_kernel_fc2_reduce from FlashInfer CuteDSL MegaMoE config (#59900)

2026-10-12 09:05:09 -05:00

fused_flydsl_moe.py

[ROCm][CI][AITER Coverage] Harden MoE sorting-backend/dispatch env-var test matrix (#58393)

2026-09-24 21:14:02 -05:00

fused_globalize_align_block_size.py

[Kernel] Fuse DeepEP-v2 decode globalize+align into one launch (#53911)

2026-10-13 11:10:38 -04:00

fused_moe_method_base.py

[MoE] Defer the TRTLLM-Gen top-k finalize on the modular path (#58635)

2026-09-25 15:38:29 +08:00

fused_moe_modular_method.py

[MoE Refactor] Remove MoE legacy code (#51078)

2026-08-05 14:35:36 -07:00

fused_moe.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

hpc_moe.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

layer.py

Add MoNe experts pruning support (#50415)

2026-10-12 23:18:43 +00:00

modular_kernel.py

[Kernel] Fuse DeepEP-v2 decode globalize+align into one launch (#53911)

2026-10-13 11:10:38 -04:00

moe_align_block_size.py

[Feature][Humming] Humming feature integration (#56685)

2026-09-20 08:11:01 -07:00

moe_fused_mul_sum.py

[Perf] Tune MoE weighted-sum kernel launch configuration (#59731)

2026-10-02 17:09:29 +00:00

moe_output.py

[MoE] Defer the TRTLLM-Gen top-k finalize on the modular path (#58635)

2026-09-25 15:38:29 +08:00

moe_permute_unpermute.py

[Bugfix][MoE] Size the permute scratch from its inputs (#60447)

2026-10-08 15:32:07 -04:00

routed_experts_capturer.py

[Bugfix] Bind routed-experts capture to the MoE layer, not the kernel (#59455)

2026-10-02 15:50:38 +00:00

routed_experts.py

[Quantization] Add ModelOpt IQ2_XS, IQ2_XXS, and Q8_0 support via b12x (#59604)

2026-10-13 09:06:33 -07:00

topk_weight_and_reduce.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

unquantized_fused_moe_method.py

[Bugfix] Bind routed-experts capture to the MoE layer, not the kernel (#59455)

2026-10-02 15:50:38 +00:00

utils.py

[Bugfix][MoE] Apply expert_map remapping for CUTLASS FP4 MOE expert parallel (#59042)

2026-10-12 10:02:56 -04:00