Michael Goin b7adf94c4a Tuned H100/H200 triton fp8 block configs for fused_qkv_a_proj (#23939) ...

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

2025-08-29 10:28:35 -07:00

..

deepgemm

…

bench_block_fp8_gemm.py

Tuned H100/H200 triton fp8 block configs for fused_qkv_a_proj (#23939)

2025-08-29 10:28:35 -07:00

bench_fp8_gemm.py

…

bench_int8_gemm.py

…

bench_nvfp4_gemm.py

…

bench_per_token_quant_fp8.py

…

benchmark_bitblas.py

…

benchmark_cutlass_fp4_moe.py

…

benchmark_grouped_gemm_cutlass.py

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

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

benchmark_layernorm.py

…

benchmark_lora.py

…

benchmark_machete.py

[kernel] Support W4A8 on Hopper (#23198)

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

benchmark_marlin.py

…

benchmark_moe_align_block_size.py

…

benchmark_moe_permute_unpermute.py

…

benchmark_moe.py

[Bugfix] Fix benchmark_moe.py for blockwise fp8. (#23823)

2025-08-28 21:44:09 +08:00

benchmark_mrope.py

…

benchmark_paged_attention.py

…

benchmark_per_token_group_quant.py

…

benchmark_quant.py

…

benchmark_reshape_and_cache_flash.py

…

benchmark_rmsnorm.py

…

benchmark_rope.py

…

benchmark_shapes.py

…

benchmark_silu_mul_fp8_quant.py

[Perf] Small optimizations for silu_mul_fp8_quant_deep_gemm (#23265)

2025-08-21 17:56:15 -04:00

benchmark_trtllm_decode_attention.py

[NVIDIA][torch.compile] Support Flashinfer TRTLLM FP8-q/kv NVFP4-out Attention Kernel (#22703)

2025-08-22 22:09:05 +00:00

benchmark_trtllm_prefill_attention.py

[NVIDIA][torch.compile] Support Flashinfer TRTLLM FP8-q/kv NVFP4-out Attention Kernel (#22703)

2025-08-22 22:09:05 +00:00

benchmark_w8a8_block_fp8.py

Tuned H100/H200 triton fp8 block configs for fused_qkv_a_proj (#23939)

2025-08-29 10:28:35 -07:00

graph_machete_bench.py

…

requirements.txt

…

utils.py

…

weight_shapes.py

[kernel] Support W4A8 on Hopper (#23198)

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