Cheng Wan 5f7aee726a refactor(moe): de-duplicate triton MoE runner path into shared helpers (#23019) ...

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-04-17 17:05:13 -07:00

..

all_reduce

[AMD][No-Merge] Simplify fused allreduce + RMSNorm and remove hidden_dim allowlist (#21986)

2026-04-11 23:47:08 -07:00

decoding_attention_triton

Fix benchmark import for should_use_tensor_core (#17232)

2026-01-16 17:48:36 -05:00

deepep

Add CLI args to conveniently support tuning more models (#12922)

2026-03-12 23:10:55 -07:00

deepseek

Fix Python 3.11 f-string lint error in deepgemm Blackwell benchmark (#22108)

2026-04-04 21:15:22 +08:00

elementwise

[Benchmark] use flashinfer bench_gpu_time instead of triton do_bench (#20305)

2026-03-12 04:04:30 +00:00

flashinfer_allreduce_fusion

[kernel slimming] Clean many useless sgl-kernel deprecated kernels (#20277)

2026-03-14 16:45:54 +08:00

fused_moe_triton

refactor(moe): de-duplicate triton MoE runner path into shared helpers (#23019)

2026-04-17 17:05:13 -07:00

lora_csgmv

Add offline auto-tuning for LoRA CSGMV kernel (#20391)

2026-04-10 13:10:43 -07:00

quantization

[Intel GPU] Enable DeepSeek R1 inference on XPU (#18461)

2026-03-29 22:35:59 -07:00

scheduler_batch

[Benchmark] use flashinfer bench_gpu_time instead of triton do_bench (#20305)

2026-03-12 04:04:30 +00:00

sliding_window_attention_triton

[Benchmark] use flashinfer bench_gpu_time instead of triton do_bench (#20305)

2026-03-12 04:04:30 +00:00