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

[NPU] Support mamba cache transfer for NPU (#20364 )

2026-03-12 12:49:21 +08:00

distributed

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

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

dllm

Decouple scheduler log printing from metrics collection (#20107 )

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

elastic_ep

[4/N] (Elastic EP) Back up Expert Weights in DRAM (#17374 )

2026-02-27 15:59:13 +08:00

entrypoints

Rename --stream-output to --incremental-streaming-output (#20614 )

2026-03-14 23:22:33 -07:00

eplb

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

2026-03-11 17:37:43 +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

VLM: add Conv2dLayer/Conv3dLayer to fix PyTorch 2.9.1 CuDNN Conv3d (#20282 )

2026-03-15 19:17:44 +08:00

lora

[NPU] make torch_native lora backend a little bit faster (#17228 )

2026-03-07 20:14:46 +03:00

managers

[Fix] Slice input_embeds to extend_input_len in prepare_for_extend (#20376 )

2026-03-15 00:07:05 -07:00

mem_cache

[bugfix] Add prev_prefix_len parameter to HiMambaRadixCache's _insert_helper() (#20539 )

2026-03-14 09:54:14 +08:00

model_executor

Enable Piecewise CUDA Graph for NemotronH Hybrid (Mamba+Attention) Models (#19903 )

2026-03-12 09:16:38 +08:00

model_loader

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

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

models

VLM: add Conv2dLayer/Conv3dLayer to fix PyTorch 2.9.1 CuDNN Conv3d (#20282 )

2026-03-15 19:17:44 +08:00

multimodal

Fix InternVL and vision attention for non-CUDA backends (e.g. XPU) (#19997 )

2026-03-14 23:24:41 -07:00

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

rename nemotron reasoning parser (#19865 )

2026-03-05 11:27:07 -08:00

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

…

utils

Fix socket utilities and reserve_port for IPv6 dual-stack support (#20491 )

2026-03-15 14:29:10 -07:00

weight_sync

…

constants.py

…

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

VLM: add Conv2dLayer/Conv3dLayer to fix PyTorch 2.9.1 CuDNN Conv3d (#20282 )

2026-03-15 19:17:44 +08:00