Liangsheng Yin 85d077f44d Introduce global alloc_len_per_decode & clean check decode memory (#15115)

2026-01-26 10:26:20 -08:00

..

ascend

[NPU] upgrade npu mf_apater plugin (#15853)

2026-01-13 09:02:10 +08:00

base

Support PD disaggregation with different TP/DP size for Qwen3-Next (#16056)

2026-01-25 15:34:02 +08:00

common

[PD] Optimize MHA models pp util calculation logic (#17306)

2026-01-19 11:23:15 +08:00

fake

[PD] Add PD support for hybrid model (Qwen3-Next, DeepSeek V3.2 Exp) (#10912)

2025-10-15 18:59:14 -07:00

mooncake

Support PD disaggregation with different TP/DP size for Qwen3-Next (#16056)

2026-01-25 15:34:02 +08:00

nixl

[NIXL] Add custom NIXL backend selection for KVManager (#17146)

2026-01-26 14:35:38 +08:00

decode_kvcache_offload_manager.py

[PD / HiCache]fix deocde kvcache offload manager memory leak (#12774)

2025-11-14 03:31:53 +08:00

decode_schedule_batch_mixin.py

Adjust wrong mtp meaning introduce by mimo (#15632)

2025-12-23 02:06:46 +08:00

decode.py

Introduce global alloc_len_per_decode & clean check decode memory (#15115)

2026-01-26 10:26:20 -08:00

encode_receiver.py

Skip mm feature pool init to avoid EPD OOM (#16388)

2026-01-23 10:53:45 +08:00

encode_server.py

Support EPD error handling (#16670)

2026-01-21 18:47:00 +08:00

kv_events.py

fix(tcp-port): replace bind_server_socket to get_zmq_socket(Port conflict) (#11961)

2025-11-12 18:05:16 +08:00

prefill.py

Support PD disaggregation with different TP/DP size for Qwen3-Next (#16056)

2026-01-25 15:34:02 +08:00

utils.py

Fix metrics and request tracing (TimeStats) (#11123)

2025-10-01 13:03:07 -07:00