Jevin Jiang 621ca2c0ab

[TPU] Increase block size and reset block shapes (#16458)

2025-05-06 13:55:04 -04:00

..

attention

[TPU] Increase block size and reset block shapes (#16458)

2025-05-06 13:55:04 -04:00

core

[v1] Introduce KVCacheBlocks as interface between Scheduler and KVCacheManager (#17479)

2025-05-06 08:50:34 -07:00

engine

Add full API docs and improve the UX of navigating them (#17485)

2025-05-03 19:42:43 -07:00

executor

[V1][PP] Support PP for MultiprocExecutor (#14219)

2025-05-06 07:58:05 -07:00

metrics

[V1][Bugfix]: vllm v1 verison metric num_gpu_blocks is None (#15755)

2025-04-30 18:20:39 +08:00

sample

[Bugfix] Fix triton import with local TritonPlaceholder (#17446)

2025-05-06 17:53:09 +08:00

spec_decode

[v1] AttentionMetadata for each layer (#17394)

2025-05-06 07:58:37 -07:00

stats

Update deprecated Python 3.8 typing (#13971)

2025-03-02 17:34:51 -08:00

structured_output

[V1][Feature] Enable Speculative Decoding with Structured Outputs (#14702)

2025-04-30 00:02:10 +00:00

worker

[v1] AttentionMetadata for each layer (#17394)

2025-05-06 07:58:37 -07:00

__init__.py

[V1] AsyncLLM Implementation (#9826)

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

kv_cache_interface.py

[V1] Implement sliding window attention in kv_cache_manager (#14097)

2025-04-01 00:33:17 -07:00

outputs.py

[V1] Aggregate chunked prompt logprobs in model runner (#14875)

2025-03-24 12:27:57 -04:00

request.py

[Core] Prevent side-channel attacks via cache salting (#17045)

2025-04-30 20:27:21 +08:00

serial_utils.py

[V1] Allow turning off pickle fallback in vllm.v1.serial_utils (#17427)

2025-04-30 16:10:54 +00:00

utils.py

[V1] Move usage stats to worker and start logging TPU hardware (#16211)

2025-04-25 14:06:01 -06:00