Benjamin Chislett b7433ca1a4

[Spec Decode] Efficient padded speculation (#24539) ...

Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>

2025-09-18 01:07:24 -04:00

..

attention

[Bugfix] Refactor Flashinfer TRTLLM attention kernel selection logic (#24600)

2025-09-17 15:36:29 -07:00

core

[Bugfix] Update import path for bc_linter_include (#24766)

2025-09-17 20:33:11 +00:00

engine

[DP] Create placement groups by ray_device_key (#25026)

2025-09-17 08:57:25 +00:00

executor

[P/D]kv_output_aggregator support P TP > D TP (#23917)

2025-09-15 11:36:06 +02:00

metrics

[Metrics] Hide deprecated metrics with gpu_ prefix (#24245)

2025-09-15 20:15:57 -06:00

pool

[Performance] V1 Pooling Models E2E Performance Optimization (#23162)

2025-08-21 13:26:09 +00:00

sample

[V1] Logits processor docs (#22919)

2025-09-17 11:53:12 -07:00

spec_decode

[Spec Decode] Efficient padded speculation (#24539)

2025-09-18 01:07:24 -04:00

structured_output

[Core] Remove tokenizer group in vLLM (#24078)

2025-09-17 08:42:59 +00:00

worker

[Spec Decode] Efficient padded speculation (#24539)

2025-09-18 01:07:24 -04:00

__init__.py

[V1] AsyncLLM Implementation (#9826)

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

cudagraph_dispatcher.py

[CI] Fix linting error on main (#23835)

2025-08-28 06:52:01 -07:00

kv_cache_interface.py

[Core] Get num_encoder_tokens from scheduler config (#24989)

2025-09-16 20:59:07 -07:00

outputs.py

[Perf][V1] Fully overlap model execution (#23569)

2025-09-05 18:20:17 -07:00

request.py

[Multimodal] Remove legacy multimodal fields in favor of MultiModalFeatureSpec (#24548)

2025-09-12 21:42:23 +08:00

serial_utils.py

[Refactor] Define MultiModalKwargsItems separate from MultiModalKwargs (#23053)

2025-08-18 09:52:00 +00:00

utils.py

Allow users to specify kv cache memory size (#21489)

2025-09-11 13:41:07 +00:00