sglang

batch_invariant_ops

[Auto Sync] Update batch_invariant_ops.py (20260221) (#19098 )

2026-02-20 18:27:31 -08:00

batch_overlap

[Feature] Integrate Elastic NIXL-EP into SGLang (#19248 )

2026-03-11 17:37:43 +08:00

checkpoint_engine

update pre-commit config (#18860 )

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

compilation

Piecewise Cuda Graph set default (#16331 )

2026-03-02 23:18:07 +08:00

configs

fix: Nemotron chunk size alias (#20458 )

2026-03-14 23:23:39 -07:00

connector

Remove sync points in mamba cache + prefill cudagraph plumbing for DP (#19639 )

2026-03-02 15:03:42 +08:00

constrained

fix: propagate grammar errors and improve llguidance backend (#20467 )

2026-03-15 16:11:18 -07:00

debug_utils

Enhance error resilience in dump comparator (#19685 )

2026-03-02 19:08:35 +08:00

disaggregation

[Network] Use NetworkAddress for dist_init_method and loopback fallbacks (#20657 )

2026-03-16 19:59:49 -07:00

distributed

[Utils] Move network/socket utilities from common.py to network.py (#20646 )

2026-03-15 20:35:24 -07:00

dllm

Decouple scheduler log printing from metrics collection (#20107 )

2026-03-07 22:09:10 -08:00

elastic_ep

[5/N] (Elastic EP) Use GPU P2P to exchange expert weights during EPLB as much as possible (#12068 )

2026-03-16 18:40:58 +08:00

entrypoints

Fix /generate JSON serialization for non-finite top_logprobs (#20714 )

2026-03-16 20:07:12 -07:00

eplb

[5/N] (Elastic EP) Use GPU P2P to exchange expert weights during EPLB as much as possible (#12068 )

2026-03-16 18:40:58 +08:00

function_call

[Tool Call] Stream DeepSeek-V3.2 function call parameters in JSON format. (#16091 )

2026-03-03 01:46:29 -08:00

grpc

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

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

hardware_backend/npu

[NPU] [Quantization] w4a4 MoE layer support (#18924 )

2026-03-11 16:52:35 +03:00

layers

Fix spec v1 token_ids_logprobs (#20718 )

2026-03-16 19:23:28 -07:00

lora

Enable embedding lookup/lora_a logic for chunked backend (#17692 )

2026-03-16 11:37:58 -07:00

managers

[Network] Use NetworkAddress for dist_init_method and loopback fallbacks (#20657 )

2026-03-16 19:59:49 -07:00

mem_cache

[HiCache] fix: release write-through lock_ref during decode (#20049 )

2026-03-16 14:49:31 +08:00

model_executor

[Network] Use NetworkAddress for dist_init_method and loopback fallbacks (#20657 )

2026-03-16 19:59:49 -07:00

model_loader

[Model] Support Nemotron 3 Super NVFP4 (#20407 )

2026-03-14 00:56:26 -07:00

models

[5/N] (Elastic EP) Use GPU P2P to exchange expert weights during EPLB as much as possible (#12068 )

2026-03-16 18:40:58 +08:00

multimodal

[EPD][VLM] support video/audio input (#17824 )

2026-03-16 14:18:21 +08:00

multiplex

Revert "[Scheduler] Decouple maybe_send_health_check_signal from process_batch_result" (#20259 )

2026-03-10 01:58:48 -07:00

observability

Fix hicache log metrics (#20504 )

2026-03-13 16:29:58 +08:00

parser

[Nemotron] Small reasoning parser fix (#20284 )

2026-03-16 13:29:40 -07:00

ray

Add Ray actor support for scheduler process management (DP=1) (#17684 )

2026-03-05 13:21:23 -08:00

sampling

Revert "feat: reduce constrained-decoding overhead in TP" (#16845 )

2026-01-10 11:39:38 +08:00

speculative

[AMD] Support speculative decoding v2 for aiter backend on ROCm/HIP (#17450 )

2026-03-11 17:01:01 -07:00

tokenizer

[Piecewise Cuda Graph] rename, refactor and add more logging (#13675 )

2025-11-21 13:28:38 +08:00

utils

Fix /generate JSON serialization for non-finite top_logprobs (#20714 )

2026-03-16 20:07:12 -07:00

weight_sync

[RL] Allow passing tensors of different dtypes for FlattenedTensorBucket (#13413 )

2025-11-19 17:24:48 +08:00

constants.py

Support releasing CUDA graph memory when paused (#7873 )

2025-10-28 14:40:50 +08:00

environ.py

[quant] Ignore FP8 quantization layers (#20340 )

2026-03-13 13:59:39 +08:00

server_args_config_parser.py

[ConfigArgumentMerger] Improve ConfigArgumentMerger compatibility with external callers (#17051 )

2026-01-16 15:32:42 +08:00

server_args.py

feat: support human-readable suffixes (25.6k, 1M, 1Mi) for token CLI (#20577 )

2026-03-16 20:05:33 -07:00