Woosuk Kwon 2ad6985c49 opt ...

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>

2025-08-15 14:24:50 -07:00

..

attention

opt

2025-08-15 14:24:50 -07:00

core

[Core] Use individual MM items in P0/P1 cache and model runner (#22570)

2025-08-13 07:18:07 -07:00

engine

[BugFix] Fix initial DP request load imbalance (#22910)

2025-08-14 15:20:28 -07:00

executor

[executor] feat: add supports_pp attr to executors (#21786)

2025-08-03 18:04:45 +08:00

metrics

[BugFix] Improve internal DP load balancing (#21617)

2025-08-01 19:45:27 -07:00

pool

[Model][1/N] Support multiple poolers at model level (#21227)

2025-07-21 02:22:21 -07:00

sample

[Bugfix] Add proper comparison for package versions (#22314)

2025-08-06 20:31:03 -07:00

spec_decode

[Core] [N-gram SD Optimization][1/n] Propose tokens with a single KMP (#22437)

2025-08-13 14:44:06 -07:00

structured_output

[Perf] Parallelize fill_bitmask to accelerate high-throughput guided decoding (#21862)

2025-08-05 19:57:49 -07:00

worker

[Log] Debug Once for Randomizing dummy data for DP Rank (#22860)

2025-08-15 11:51:50 -07:00

__init__.py

[V1] AsyncLLM Implementation (#9826)

2024-11-11 23:05:38 +00:00

cudagraph_dispatcher.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

kv_cache_interface.py

[V1] [Hybrid] Support using float32 for state in Hybrid Models (Mamba2, Mamba1, Minimax) (#22928)

2025-08-15 12:57:06 +00:00

outputs.py

[V1] [P/D] Refactor KV Connector Path (#21980)

2025-08-03 04:03:40 -07:00

request.py

[Core] Use individual MM items in P0/P1 cache and model runner (#22570)

2025-08-13 07:18:07 -07:00

serial_utils.py

[Core] Use individual MM items in P0/P1 cache and model runner (#22570)

2025-08-13 07:18:07 -07:00

utils.py

[refactor] improve ConstantList exception specificity (#22156)

2025-08-03 22:14:49 -07:00