freeliuzc 6715196924 fix attention bug in spec decoding (#5480)

2025-12-10 12:56:13 +08:00

..

attention

fix attention bug in spec decoding (#5480)

2025-12-10 12:56:13 +08:00

backends

[Metax] support ENABLE_V1_KVCACHE_SCHEDULER (#5163)

2025-11-24 19:19:49 +08:00

moe

[Optimization] 1.fix tp+ep moe_forward; 2.set max_prefill_batch=env.MAX_PREFILL_NUM (#5316)

2025-12-02 13:03:55 +08:00

pool

[Feature] support pooling model dummy_run (#4345)

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

quantization

support moe offline quant (#5142)

2025-11-24 18:59:18 +08:00

sample

[Feature] ThreadPoolExecutor async fill_token_bitmask (#5083)

2025-11-19 10:04:16 +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

[RL]Fix missing is_distributed attribute (#5150)

2025-11-21 14:14:25 +08:00

lm_head.py

[RL]Resolve shape mismatch problems in RL-related modules (#5032)

2025-11-19 11:12:48 +08:00

mtp_linear.py

update (#4985)

2025-11-13 15:58:01 +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