sglang

batch_invariant_ops

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

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

batch_overlap

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

2026-03-02 15:03:42 +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

[Feature] Improve weight loading log (#18651 )

2026-03-03 14:16:13 -08:00

connector

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

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

constrained

Fix grammar sync across TP ranks (#17100 )

2026-01-15 18:38:01 -08:00

debug_utils

Enhance error resilience in dump comparator (#19685 )

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

disaggregation

[model gateway][0/N] router EPD support: add encoder grpc server backend support (#16552 )

2026-03-03 19:38:15 +08:00

distributed

[Fix] NPU deepep hccl buffer and fix IPC safe check (#17804 )

2026-03-03 14:56:06 +08:00

dllm

Refactor: observability code cleanup (#17862 )

2026-02-24 18:07:29 -08:00

elastic_ep

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

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

entrypoints

Fix /health regression from early prebound socket listen (#19805 )

2026-03-03 23:00:46 -08:00

eplb

Refactor: observability code cleanup (#17862 )

2026-02-24 18:07:29 -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

Refactor: observability code cleanup (#17862 )

2026-02-24 18:07:29 -08:00

hardware_backend/npu

[NPU]grok2 model support (#17119 )

2026-03-03 10:24:10 +08:00

layers

[AMD] Fix Qwen3-Coder-Next: Add missing k_scale/v_scale args to extend_attention_fwd in aiter_backend (#19736 )

2026-03-03 22:01:08 -08:00

lora

[Fix] Add lora tied lm head support (for Qwen2.5, Gemma, etc model need) (#18634 )

2026-02-19 00:34:51 +08:00

managers

[sgl]add pin_mem to remove cpu->gpu copy sync point (#19795 )

2026-03-03 21:00:51 -08:00

mem_cache

Fix triton alloc extend kernel (#19780 )

2026-03-03 21:01:16 -08:00

model_executor

Fix dp_attention crash when dp_size < tp_size in warmup dummy run (#19760 )

2026-03-03 19:43:13 -08:00

model_loader

[Quantization] Support config.json quantization_config format, fix exclude_modules matching, and fix KV cache scale loading for Nemotron (#18546 )

2026-02-21 16:14:29 +08:00

models

Remove naive rotary forward overriding. (#19263 )

2026-03-03 21:50:40 -08:00

multimodal

[vlm][internVL] Support processor and embedding inputs for InternVL (#19127 )

2026-02-26 22:46:48 -08:00

multiplex

[Bugfix] fix the display error (inconsistent context) (#17699 )

2026-01-31 23:35:11 -08:00

observability

feat: TTL-based prefix pinning with refresh-on-hit for HiRadixCache (#18941 )

2026-03-01 23:27:22 -08:00

parser

Add KimiK2Detector with tool interruption support (#19696 )

2026-03-03 14:04:49 +08:00

sampling

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

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

speculative

[Feature] implement the standard multi-layer MTP for step3p5 (#18564 )

2026-03-03 22:48:53 -08:00

tokenizer

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

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

utils

[sgl]add pin_mem to remove cpu->gpu copy sync point (#19795 )

2026-03-03 21:00:51 -08:00

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

[model gateway][0/N] router EPD support: add encoder grpc server backend support (#16552 )

2026-03-03 19:38:15 +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

[Bugfix] For cp: Fixed hang problem in prefix cache and kvcache support fp8 in-seq-split mode (#19656 )

2026-03-03 19:19:46 -08:00