K11OntheBoat 870dbac370 Use triton qk_norm both in Prefill and Decode (#7213) ...

Co-authored-by: “liuruian” <liuruian@baidu.com>

2026-04-10 15:44:01 +08:00

..

attention

[BugFix] Fix batch_size derivation and relax shape checks in SM90 flash_mask_attn (#7210)

2026-04-09 11:05:10 +08:00

backends

Split enable_mm (#7183)

2026-04-08 11:25:41 +08:00

batch_invariant_ops

[Optimization] Use a separate driver when using Triton with Paddle (#6897)

2026-03-24 10:56:00 +08:00

moe

[OP]Unify MoE op with moe_permute path for bf16 GLM (#7164)

2026-04-09 16:17:56 +08:00

pool

…

quantization

[FDConfig] Support CLI args for quantization params and add cudagraph validation (#7281)

2026-04-10 14:13:42 +08:00

sample

[XPU] Refactor pre process (#6993)

2026-04-01 20:29:55 +08:00

__init__.py

…

activation.py

…

embeddings.py

…

linear.py

[Optimization] merge matmul and add (#6986)

2026-04-03 18:02:03 +08:00

lm_head.py

…

mtp_linear.py

…

normalization.py

Use triton qk_norm both in Prefill and Decode (#7213)

2026-04-10 15:44:01 +08:00

pooler.py

…

rotary_embedding.py

…

utils.py

[Feature] Support NVFP4 Flashinfer-cutedsl MoE on SM100 (#6963)

2026-03-30 11:37:04 +08:00