ming1753 3148dbca06 [BugFix] fix VL fp8 bug when moe token_num is 0 (#4928) ...

* [BugFix] fix VL fp8 bug when moe token_num is 0

* fix bug

* format

* fix bug

2025-11-12 21:19:36 +08:00

..

attention

[KVCache] support unified cache backend (#4903)

2025-11-12 14:54:52 +08:00

backends

[Metax] optimize flash mla (#4915)

2025-11-12 16:43:46 +08:00

moe

[BugFix] fix VL fp8 bug when moe token_num is 0 (#4928)

2025-11-12 21:19:36 +08:00

pool

[Feature] support pooling model dummy_run (#4345)

2025-10-17 13:30:55 +08:00

quantization

[Metax] support default_v1 loader & thinking model (#4956)

2025-11-12 16:32:26 +08:00

sample

[Perf] Support tensor transmission between work and engine with zero-copy to improve efficiency (#4839)

2025-11-11 15:43:11 +08:00

__init__.py

[LLM] First commit the llm deployment code

2025-06-09 19:20:15 +08:00

activation.py

fix the get_act_fn,_load_st_projector (#4824)

2025-11-06 16:13:35 +08:00

embeddings.py

[Metax] support default_v1 loader & thinking model (#4956)

2025-11-12 16:32:26 +08:00

linear.py

refactor pt loading (#4532)

2025-11-11 21:30:39 +08:00

lm_head.py

refactor pt loading (#4532)

2025-11-11 21:30:39 +08:00

mtp_linear.py

[Speculative Decoding][MTP]Support mtp in epdptp mode (#4614)

2025-10-28 16:02:47 +08:00

normalization.py

[TSP] Support qwen3 moe tsp + cudagraph (#4871)

2025-11-10 23:37:51 +08:00

pooler.py

fix the get_act_fn,_load_st_projector (#4824)

2025-11-06 16:13:35 +08:00

rotary_embedding.py

[Feature] Support Paddle-OCR (#4396)

2025-10-24 23:34:30 +08:00

utils.py

[OPs] MoE support wfp8afp8(channelwise) and improve per_token_quant_fp8 (#4238)

2025-09-24 16:39:51 +08:00