Trevor MorrisandClaude Opus 4.7 9c53853ea3 Use pack topk ids triton kernel for flashinfer_trtllm_routed (#25702) ...

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-06-16 19:27:14 -07:00

..

arg_groups

Custom spec algorithm can handle server args (#28162)

2026-06-16 17:13:52 -07:00

batch_invariant_ops

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

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

batch_overlap

Unify NVTX annotation helpers and split the enable gate per subsystem (#28165)

2026-06-14 00:04:59 -07:00

checkpoint_engine

update pre-commit config (#18860)

2026-02-16 00:18:31 +08:00

compilation

fix(fa3): no NaN embeddings with fa_skip_kv_cache under piecewise CUDA graph (#27343)

2026-06-15 13:46:22 -07:00

configs

Revert "[AMD][Quantization] Online MXFP4 quantization 2/N - FP8 to MXFP4 requantization on AMD GPUs" (#28213)

2026-06-14 13:34:06 -07:00

connector

[connector] Add Azure Blob Storage connector (az:// and *.blob.core.windows.net URLs) (#23995)

2026-05-11 14:20:22 -07:00

constrained

fix: prevent stale bitmask leakage in LLGuidance grammar backend (#23653)

2026-06-13 13:27:08 -07:00

debug_utils

[lint] Enable Ruff UP037 to drop redundant quoted annotations (#27984)

2026-06-11 17:38:05 -07:00

disaggregation

[PD] Optimize SWA allocation (#28085)

2026-06-15 11:01:55 -07:00

distributed

[CAR] Let custom allreduce support VMM based allocation (#27593)

2026-06-16 04:49:05 -07:00

dllm

Fix Req fill_len (fill_ids) having dual semantics by restricting to truncated/committed semantics (#26659)

2026-06-08 14:52:48 +08:00

elastic_ep

chore: bump mooncake version to 0.3.11.post1 (#25989)

2026-05-25 16:51:03 +08:00

entrypoints

fix(openai): validate assistant tool call arguments before chat template (#28035)

2026-06-16 12:21:23 +00:00

eplb

LPLB: linear-programming load balancer for MoE expert parallelism (#24515)

2026-06-16 10:19:42 -07:00

function_call

Support GLM-4.7 function calling via structural tags (#28149)

2026-06-14 14:57:07 +08:00

grpc

[gRPC] Extract gRPC servicer into standalone package (#20478)

2026-03-13 09:13:29 -07:00

hardware_backend

✨ [llm][npu][quant] Add W8A8 MXFP8 quantization support for Qwen3 Dense on Ascend NPU (#22352)

2026-06-16 09:45:18 +03:00

kv_canary

[Spec] Remove deprecated EAGLE v1 DRAFT_EXTEND forward mode (#28129)

2026-06-13 13:00:21 -07:00

layers

Use pack topk ids triton kernel for flashinfer_trtllm_routed (#25702)

2026-06-16 19:27:14 -07:00

lora

[LoRA] Fix chunked SGMV (csgmv) CUDA graph segment replay (#28371)

2026-06-16 19:19:07 -07:00

managers

[Tokenizer] Fix abort racing server crash when large amount of aborts (#28341)

2026-06-16 14:49:35 -07:00

mem_cache

[HiCache]Asymmetric pool support direct backend (#28446)

2026-06-16 13:17:57 -07:00

model_executor

Fix circular import when sglang.srt.model_executor.runner_backend is imported first (#28002)

2026-06-16 12:04:34 -07:00

model_loader

fix: Fix DSR1 perf regression due to unnecessarily falling back to triton gemm (#28073)

2026-06-15 09:45:09 -04:00

models

[AMD] Feat/dp moe reduce scatter (#28216)

2026-06-16 18:12:39 -07:00

multimodal

[perf] Reuse a pooled HTTP session for multimodal URL downloads (#28056)

2026-06-15 17:47:47 +08:00

multiplex

Move request-ingress methods to SchedulerRequestReceiver (#25610)

2026-05-18 18:31:17 +08:00

observability

[Spec] Clear dead DRAFT_EXTEND objects left after EAGLE v1 removal (#28133)

2026-06-13 13:04:51 -07:00

parser

Fix Responses API request handling (#25881)

2026-06-12 14:47:55 -07:00

platforms

[plugin][distributed] use active platform's backend in get_default_distributed_backend (#23969)

2026-06-11 20:05:17 -07:00

plugins

Multi platform Plugin (#21388)

2026-04-19 17:23:51 -07:00

ray

feat: add SGLANG_RAY_BUNDLE_INDICES for fine-grained Ray bundle index control (#24667)

2026-05-30 02:19:50 -07:00

sampling

[Perf] Avoid per-decode-step host sync in min_new_tokens penalty (#28397)

2026-06-15 23:32:28 -07:00

session

[perf] reduce overhead of fill_ids list reconstruction and decref (#27965)

2026-06-14 00:41:11 -07:00

speculative

Custom spec algorithm can handle server args (#28162)

2026-06-16 17:13:52 -07:00

state_capturer

Fix routed-experts device buffer overflow under DP attention (#26123)

2026-05-30 19:11:06 -07:00

tokenizer

…

utils

fix: get_processor fails when --tokenizer-path lacks model config.json (#25643)

2026-06-16 20:25:05 +03:00

weight_sync

…

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

[2/n] [CP] Add context parallel strategy abstractions (#27313)

2026-06-16 00:20:04 -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

[HiCache]Asymmetric pool support direct backend (#28446)

2026-06-16 13:17:57 -07:00