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

..

graph_optimization

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

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

guided_decoding

…

layers

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

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

logits_processor

…

model_loader

[Loader] add multi-thread model loading (#6877)

2026-04-09 23:40:15 -07:00

models

[RL] support moe-topk use topk_reduce_func (#7218)

2026-04-09 11:01:03 +08:00

ops

…

__init__.py

…

entropy_utils.py

…

forward_meta.py

[Models]support GLM4.7 Flash && Ernie_MLA (#7139)

2026-04-03 17:41:33 +08:00

load_weight_utils.py

[Loader] add multi-thread model loading (#6877)

2026-04-09 23:40:15 -07:00

pre_and_post_process.py

…

utils.py

…

xpu_pre_and_post_process.py

…