Michael Goin 06d490282f

[NVFP4][Perf] Tune NVFP4 input quant kernel for small batch size (#30897) ...

Signed-off-by: mgoin <mgoin64@gmail.com>

2025-12-21 09:41:57 -08:00

..

deepgemm

[Doc]: fix typos in various files (#29230)

2025-11-24 11:10:48 +00:00

bench_block_fp8_gemm.py

…

bench_fp8_gemm.py

…

bench_int8_gemm.py

…

bench_mxfp4_qutlass.py

…

bench_nvfp4_gemm.py

…

bench_nvfp4_quant.py

[NVFP4][Perf] Tune NVFP4 input quant kernel for small batch size (#30897)

2025-12-21 09:41:57 -08:00

bench_nvfp4_qutlass.py

…

bench_per_token_quant_fp8.py

…

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

[Perf][Kernels] Vectorize csrc/activations_kernels.cu (#29512)

2025-12-16 14:56:02 -08:00

benchmark_bitblas.py

…

benchmark_cutlass_fp4_moe.py

…

benchmark_cutlass_moe_fp8.py

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

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

benchmark_device_communicators.py

…

benchmark_fused_collective.py

…

benchmark_grouped_gemm_cutlass.py

…

benchmark_layernorm.py

…

benchmark_lora.py

…

benchmark_machete.py

[Kernel][Quantization] add w4a8 support for marlin kernel (#24722)

2025-11-29 07:19:33 -08:00

benchmark_marlin.py

[Kernel][Quantization] add w4a8 support for marlin kernel (#24722)

2025-11-29 07:19:33 -08:00

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

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

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

benchmark_moe.py

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

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

benchmark_mrope.py

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

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

benchmark_paged_attention.py

…

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

…

benchmark_reshape_and_cache_flash.py

…

benchmark_reshape_and_cache.py

…

benchmark_rmsnorm.py

…

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

…

benchmark_silu_mul_fp8_quant.py

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

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

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

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

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

graph_machete_bench.py

…

requirements.txt

…

utils.py

…

weight_shapes.py

…