Isotr0py 74a1ac38b0

[v1] Add PrefixLM support to TritonAttention backend (#30386)

2025-12-17 16:05:24 -08:00

..

attention

[v1] Add PrefixLM support to TritonAttention backend (#30386)

2025-12-17 16:05:24 -08:00

core

[Core][MM] Optimize encoder cache manager by operating with embeddings only (#30475)

2025-12-16 14:18:17 -08:00

engine

[Bugfix] Fix RequestOutput miss lora_request (#30636)

2025-12-16 01:36:35 -08:00

executor

[Logs] Optimize startup logs 4 (#29903)

2025-12-13 16:12:53 -05:00

kv_offload

CPU KV Offloading: Use more CUDA streams (#29013)

2025-12-14 23:50:45 +00:00

metrics

feat(metrics): Add prefill KV compute metric excluding cached tokens (#30189)

2025-12-09 00:08:48 +00:00

pool

[Model][6/N] Improve all pooling task | Support chunked prefill with ALL pooling (#27145)

2025-12-04 13:44:15 +00:00

sample

[bugfix] fix bug when top_logprobs=0 with spec decoding (#30059)

2025-12-12 09:03:35 -08:00

spec_decode

[ROCm][Bugfix] Add MLACommonMetadata to allowed attention types for speculative decoding (#30430)

2025-12-11 19:25:01 +00:00

structured_output

Fix lazy import (#30858)

2025-12-17 03:33:50 -08:00

worker

[Chore] Factor out logic for requesting initial memory (#30868)

2025-12-17 07:32:17 -08:00

__init__.py

…

cudagraph_dispatcher.py

[BugFix] Fix assert batch_descriptor.num_tokens == num_tokens_padded (#30173)

2025-12-09 10:36:12 -05:00

kv_cache_interface.py

[Feature] Prefill Context Parallel (PCP) basic support (#28718)

2025-11-19 15:52:44 -05:00

outputs.py

[KVConnector] Add KV events to KV Connectors (#28309)

2025-12-11 15:30:29 +01:00

request.py

[Core][MM] Optimize encoder cache manager by operating with embeddings only (#30475)

2025-12-16 14:18:17 -08:00

serial_utils.py

[Model] Move multimodal_cpu_fields definition to field config (#30181)

2025-12-06 13:40:02 +00:00

utils.py

…