Jinwoo Jeong 4f51dad1da fix: prevent ReqTimeStats from being dropped during IPC serialization (#31339)

2026-07-30 14:42:38 +08:00

..

arg_groups

Enable GPT-OSS FlashInfer MXFP4 on SM120 (#32668)

2026-07-30 00:04:23 +00:00

batch_invariant_ops

[NPU]adaptation to support deterministic inference (#21197)

2026-06-09 09:22:09 +08:00

batch_overlap

[BugFix] Prevent TBO crash when return_logprob is enabled (#32180)

2026-07-26 00:07:54 -07:00

checkpoint_engine

…

compilation

fix: fix vlm cuda graph shape stability (#30868)

2026-07-19 22:35:51 +08:00

configs

Fix attention backends for models with per-layer head counts (num_attention_heads_per_layer) (#32625)

2026-07-29 20:03:00 -07:00

connector

…

constrained

Use native batched llguidance mask generation (#32412)

2026-07-25 16:36:32 -07:00

debug_utils

[Fix] Update PR #25015 revert for fused topk=1 draft postprocess (#31519)

2026-07-16 21:51:16 -07:00

disaggregation

[PD] Handle abort requests in PP mode (#32797)

2026-07-30 14:39:44 +08:00

distributed

Disable extra NCCL CUDA event synchronization with symm mem (#27089)

2026-07-29 22:39:36 +08:00

dllm

[DLLM] vectorized joint/low-confidence decoding and skip redundant attn init (#21094)

2026-07-30 09:13:14 +03:00

elastic_ep

[Elastic EP] Fix recovery lifecycle and add manual coverage (#31744)

2026-07-25 14:32:47 +08:00

entrypoints

embedding: centralize capabilities and complete OpenAI compatibility (#32481)

2026-07-30 10:28:52 +08:00

eplb

[2/N] elastic-ep: Enable EPLB after scale-up (#30553)

2026-07-30 01:06:56 +08:00

function_call

Fix Inkling tool-call parsing recovery, content handling, and streaming (#32861)

2026-07-30 14:11:09 +08:00

grpc

…

hardware_backend

[MLX] Size request capacity by attention DP (#32115)

2026-07-29 18:18:22 -07:00

kv_canary

Radix Cache Split: Spin off TreeCore (#29901)

2026-07-25 14:31:59 -07:00

layers

Fix attention backends for models with per-layer head counts (num_attention_heads_per_layer) (#32625)

2026-07-29 20:03:00 -07:00

lora

[LoRA] Support LoRA under the breakable/full prefill CUDA graph (#30988)

2026-07-26 22:10:03 -07:00

managers

[PD] Handle abort requests in PP mode (#32797)

2026-07-30 14:39:44 +08:00

mem_cache

Follow up on #30157 post-merge review (#32672)

2026-07-29 15:03:59 -07:00

model_executor

Follow up on #30157 post-merge review (#32672)

2026-07-29 15:03:59 -07:00

model_loader

model: support EmbeddingGemma (#32375)

2026-07-27 10:40:47 +08:00

models

fix(qwen3.5): restrict MoE weights to local PP layers (#32022)

2026-07-29 12:01:09 -07:00

multimodal

vlm: reject moss vision metadata mismatches (#31957)

2026-07-29 07:09:54 +08:00

multiplex

Revert RuntimeContext config-namespace reads/roles (#31813–#31817) (#32100)

2026-07-22 11:52:41 -07:00

observability

fix: prevent ReqTimeStats from being dropped during IPC serialization (#31339)

2026-07-30 14:42:38 +08:00

parser

Fix Inkling tool-call parsing recovery, content handling, and streaming (#32861)

2026-07-30 14:11:09 +08:00

platforms

[Intel GPU] Add XPU Platform support (#31949)

2026-07-24 12:49:29 +08:00

plugins

…

ray

[FEAT] Support fast engine recovery through weight cache (#27139)

2026-07-25 14:31:21 +08:00

sampling

[Refactor] Move sampling tokenizer validation helper (#32694)

2026-07-28 16:48:03 -07:00

session

[HiCache] Merge HiCache event checks to reduce decode overhead (#30511)

2026-07-29 23:48:48 +08:00

speculative

[misc] Remove unused multi_layer_draft_forward_cg module (#32881)

2026-07-29 21:17:16 -07:00

state_capturer

Revert RuntimeContext config-namespace reads/roles (#31813–#31817) (#32100)

2026-07-22 11:52:41 -07:00

tokenizer

Add Inkling model support (#31681)

2026-07-19 22:57:37 -07:00

utils

Support DCP for Kimi Linear model (#32612)

2026-07-28 22:59:58 -07:00

weight_cache

[FEAT] Support fast engine recovery through weight cache (#27139)

2026-07-25 14:31:21 +08:00

weight_sync

Introduce WeightUpdater and WeightExporter components (#31148)

2026-07-14 15:53:33 +08:00

constants.py

[AMD][MXFP4] Online MXFP4 quantization 1/N - dense and MOE models w. original BF16 weight (#18005)

2026-06-03 12:55:24 -07:00

environ.py

Skip mamba lock during decoding (#32228)

2026-07-29 21:53:28 +08:00

runtime_context.py

Revert RuntimeContext config-namespace reads/roles (#31813–#31817) (#32100)

2026-07-22 11:52:41 -07:00

server_args_config_parser.py

fix(server): serialize nested dict config values as JSON (#28094)

2026-06-12 19:27:31 -07:00

server_args.py

embedding: centralize capabilities and complete OpenAI compatibility (#32481)

2026-07-30 10:28:52 +08:00