fzyzcjy 87068b5cc7 Support gathering expert distribution details (#6665)

2025-05-27 15:32:59 -07:00

..

multimodal_processors

follow-up: move Idefics2 to a shared location to eliminate unexpected dependency. (#6603)

2025-05-26 19:23:59 -07:00

cache_controller.py

fix bug that gpu0 occupies more memory when hicache is turned on (#5778)

2025-05-09 15:36:08 -07:00

configure_logging.py

Support penalty in overlap mode; return logprob with chunked prefill; improve benchmark scripts (#3988)

2025-03-03 00:12:04 -08:00

data_parallel_controller.py

Add typo checker in pre-commit (#6179)

2025-05-11 12:55:00 +08:00

deepseek_eplb.py

Fix EPLB algorithm fail to run when using 3 nodes for prefill (#6629)

2025-05-26 08:43:24 -07:00

detokenizer_manager.py

Support incremental streaming of logprob/token_ids between scheduler and detokenizer (#6225)

2025-05-12 14:33:38 -07:00

eplb_manager.py

Support dynamically rebalancing experts using EPLB (#6469)

2025-05-21 23:13:21 -07:00

expert_distribution.py

Support gathering expert distribution details (#6665)

2025-05-27 15:32:59 -07:00

expert_location_dispatch.py

Support fake perfectly balanced EP dispatch algorithm (#6571)

2025-05-25 22:35:51 -07:00

expert_location.py

Fix EPLB algorithm fail to run when using 3 nodes for prefill (#6629)

2025-05-26 08:43:24 -07:00

io_struct.py

[PD] Add doc and simplify sender.send (#6019)

2025-05-21 21:22:21 -07:00

mm_utils.py

Refactor vlm embedding routine to use precomputed feature (#6543)

2025-05-24 18:39:21 -07:00

multimodal_processor.py

Clean up imports (#5467)

2025-04-16 15:26:49 -07:00

schedule_batch.py

Support overlapping two batches (#4068)

2025-05-24 17:39:07 -07:00

schedule_policy.py

[Minor] cleanup unused imports (#6358)

2025-05-16 14:52:38 -07:00

scheduler_output_processor_mixin.py

Support incremental streaming of logprob/token_ids between scheduler and detokenizer (#6225)

2025-05-12 14:33:38 -07:00

scheduler.py

Fix profiling will crash the server when using num_steps (#6586)

2025-05-25 22:36:02 -07:00

session_controller.py

Support precomputed multimodal features for Qwen-VL and Gemma3 models. (#6136)

2025-05-16 12:26:15 -07:00

tokenizer_manager.py

[RL] allow weight updation with dp attention enabled (#6311)

2025-05-21 01:58:55 -07:00

tp_worker_overlap_thread.py

Support incremental streaming of logprob/token_ids between scheduler and detokenizer (#6225)

2025-05-12 14:33:38 -07:00

tp_worker.py

Log if cuda graph is used & extend cuda graph capture to cuda-graph-max-bs (#6201)

2025-05-12 00:17:33 -07:00

utils.py

refactor: multimodal data (#4754)

2025-03-31 09:57:51 -07:00