Matthias Gehre 934acddef9 [Perf] fused_moe: add int4_w4a16 benchmark support and tuning config (#34130) ...

Signed-off-by: Matthias Gehre <matthias.gehre@amd.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>

2026-02-13 00:14:27 -08:00

..

cpu

[Misc] Replace deprecated interface seed_everything (#33474)

2026-01-31 05:38:39 -08:00

deepgemm

[Perf] Create TMA-aligned input scale tensor for DeepGemm on Hopper (#32619)

2026-01-22 15:47:04 -05:00

bench_block_fp8_gemm.py

Remove weight_scale.T special case for SM90 Block FP8 CUTLASS kernel (#28431)

2025-11-11 11:46:04 -05:00

bench_fp8_gemm.py

[Misc] Add SPDX-FileCopyrightText (#20428)

2025-07-04 07:40:42 +00:00

bench_int8_gemm.py

[Benchmark] Refactor benchmark script for fp8 & int8 (#19627)

2025-06-15 15:15:37 +08:00

bench_mxfp4_qutlass.py

[Transform] [Quantization] Add QuTLASS support to vLLM (#24440)

2025-10-10 09:43:40 -07:00

bench_nvfp4_gemm.py

Enable Fbgemm NVFP4 on Dense models (#25609)

2025-09-24 21:12:53 -07:00

bench_nvfp4_quant.py

[Perf][Kernel] Optimize FP4 quantization kernels (SM100F) (#32520)

2026-01-24 18:45:27 -07:00

bench_nvfp4_qutlass.py

[Transform] [Quantization] Add QuTLASS support to vLLM (#24440)

2025-10-10 09:43:40 -07:00

bench_per_token_quant_fp8.py

[Chore]:Extract math and argparse utilities to separate modules (#27188)

2025-10-26 04:03:32 -07:00

benchmark_2d_silu_mul_fp8_quant.py

[Performance][DP/EP] Add silu_mul_per_token_group_quant_fp8_colmajor kernel (#29470)

2025-12-03 18:04:59 +00:00

benchmark_activation.py

[PluggableLayer][1/N] Define PluggableLayer (Fix ci) (#32744)

2026-01-21 11:38:04 -05:00

benchmark_cutlass_moe_fp8.py

[Refactor] Replace activation: str with MoEActivation enum (#33843)

2026-02-11 17:29:32 -08:00

benchmark_cutlass_moe_nvfp4.py

[MoE Refactor] Integrate Naive Prepare Finalize into MK (#32567)

2026-01-27 01:28:02 +00:00

benchmark_device_communicators.py

use the same stream for cuda graph catpure and replay for NCCL (#29207)

2025-12-25 19:10:03 +08:00

benchmark_fused_collective.py

[Kernel] FlashInfer: switch allreduce fusion to unified API (#33985)

2026-02-09 15:43:24 +00:00

benchmark_fused_topk.py

[Kernel] Add topk_sigmoid kernel (#31246)

2026-01-21 22:49:51 +00:00

benchmark_grouped_gemm_cutlass.py

[MoE Refactor] Oracle Select FP8+NVFP4 Kernels In Priority (#32414)

2026-01-21 08:22:33 -05:00

benchmark_layernorm.py

[Misc] Use deprecated for seed_everything (#31780)

2026-01-06 11:29:55 +00:00

benchmark_lora.py

Add unpermute-aware fused MoE LoRA path (#32655)

2026-02-02 09:46:09 +08:00

benchmark_machete.py

[Refactor] Rename gptq_marlin to marlin to match MoE (#32952)

2026-01-23 16:48:12 -05:00

benchmark_marlin.py

[Quantization][Deprecation] Remove Marlin 24 (#32688)

2026-01-28 15:54:59 +00:00

benchmark_mla_k_concat.py

[perf] Use direct copy (broadcast) instead of cat for k_nope/k_pe in MLA prefill (#29710)

2025-12-11 08:20:45 +00:00

benchmark_moe_align_block_size.py

[Kernel][MoE] optimize moe_align_block_size (#29642)

2025-12-07 01:58:47 -08:00

benchmark_moe_permute_unpermute.py

[Refactor] Remove align block size logic in moe_permute (#33449)

2026-02-06 10:57:06 -08:00

benchmark_moe.py

[Perf] fused_moe: add int4_w4a16 benchmark support and tuning config (#34130)

2026-02-13 00:14:27 -08:00

benchmark_mrope.py

[Misc] Use deprecated for seed_everything (#31780)

2026-01-06 11:29:55 +00:00

benchmark_paged_attention.py

[Misc] Use deprecated for seed_everything (#31780)

2026-01-06 11:29:55 +00:00

benchmark_per_token_group_quant.py

Replace torch.cuda.Event with torch.Event for better hardware compatibility (#26985)

2025-11-18 11:34:36 -08:00

benchmark_quant.py

[Misc] Use deprecated for seed_everything (#31780)

2026-01-06 11:29:55 +00:00

benchmark_reshape_and_cache_flash.py

[1/N][Attention] Restructure attention: move files (#31916)

2026-01-09 13:10:24 -08:00

benchmark_reshape_and_cache.py

[Misc] Use deprecated for seed_everything (#31780)

2026-01-06 11:29:55 +00:00

benchmark_rmsnorm.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

benchmark_rope.py

Standardise get_rope to use rope_parameters["partial_rotary_factor"], not rotary_dim (#30389)

2025-12-11 20:45:23 +00:00

benchmark_shapes.py

Disable nm-testing models with issues in CI (#28206)

2025-11-06 06:19:07 -08:00

benchmark_silu_mul_fp8_quant.py

[Misc] Use deprecated for seed_everything (#31780)

2026-01-06 11:29:55 +00:00

benchmark_trtllm_decode_attention.py

Replace torch.cuda.Event with torch.Event for better hardware compatibility (#26985)

2025-11-18 11:34:36 -08:00

benchmark_trtllm_prefill_attention.py

Replace torch.cuda.Event with torch.Event for better hardware compatibility (#26985)

2025-11-18 11:34:36 -08:00

benchmark_w8a8_block_fp8.py

[AMD][QWEN3-NEXT] FP8 Tunings (#32042)

2026-01-27 09:34:13 +00:00

graph_machete_bench.py

[Chore]:Extract math and argparse utilities to separate modules (#27188)

2025-10-26 04:03:32 -07:00

requirements.txt

[Kernel] (2/N) Machete - Integrate into CompressedTensorsWNA16 and GPTQMarlin (#7701)

2024-09-23 13:46:26 -04:00

utils.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

weight_shapes.py

[kernel] Support W4A8 on Hopper (#23198)

2025-08-24 06:18:04 +00:00