sglang

batch_invariant_ops

Support fast gemm when in batch invariant DeepGEMM fallback (#13259 )

2025-11-15 16:34:15 +08:00

batch_overlap

[Feature] Xiaomi MiMo-V2-Flash day0 support (#15207 )

2025-12-19 11:40:07 +08:00

checkpoint_engine

…

compilation

tiny refactor pcg split op registration (#16863 )

2026-01-10 23:45:28 +08:00

configs

Use swa radix cache and memory pool for gpt-oss model (#17261 )

2026-01-18 13:47:07 +08:00

connector

…

constrained

Fix grammar sync across TP ranks (#17100 )

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

debug_utils

Support multiple engines for router simulation in schedule simulator (#16368 )

2026-01-04 11:46:05 +08:00

disaggregation

[PD] Optimize MHA models pp util calculation logic (#17306 )

2026-01-19 11:23:15 +08:00

distributed

Add clear error message when OOM with symmetric memory (#17038 )

2026-01-16 20:44:37 +08:00

dllm

[DLLM] Implement initial dynamic batching for diffusion LLM (#14883 )

2026-01-17 16:48:15 +08:00

elastic_ep

…

entrypoints

Fix v32 continue_final_message not work (#16567 )

2026-01-19 21:45:39 +08:00

eplb

Re-introduce the unit test of test_mooncake_ep_small (#16019 )

2026-01-12 17:01:24 +08:00

function_call

fix function calling for Trinity (#17364 )

2026-01-19 18:33:17 -08:00

grpc

[gRPC] Add GetLoads RPC for comprehensive load metrics (#17087 )

2026-01-14 12:05:10 -08:00

hardware_backend/npu

[NPU]Support GPT-OSS for NPU (#14197 )

2026-01-19 04:13:41 +08:00

layers

feat: support bitsandbytes quantization algorithm (#15325 )

2026-01-19 18:36:56 -08:00

lora

[Feature] overlap LoRA weight loading with compute (#15512 )

2026-01-19 10:43:17 +08:00

managers

[RadixTree][1/N Refactor]: Support unified match_prefix params (#17142 )

2026-01-19 22:39:40 +08:00

mem_cache

[RadixTree][1/N Refactor]: Support unified match_prefix params (#17142 )

2026-01-19 22:39:40 +08:00

metrics

Code clean up for fp8 quantization (#16982 )

2026-01-13 12:38:39 -08:00

model_executor

Pipe customized_info through CudaGraphRunner output (#17088 )

2026-01-19 15:48:49 -08:00

model_loader

ci: enable offline mode when local cache is complete to avoid HF Hub … (#16121 )

2026-01-15 20:15:33 -08:00

models

[NPU]Support GPT-OSS for NPU (#14197 )

2026-01-19 04:13:41 +08:00

multimodal

[VLM][Reland] Refactor load_mm_data to improve performance (#16152 )

2026-01-18 14:11:17 +08:00

multiplex

Fix deprecation warning of sre_parse for python 3.13 (#16247 )

2025-12-31 21:25:09 -08:00

parser

VLM: enhance VL embedding model with video input support and revise warm-up strategy (#16635 )

2026-01-08 22:12:01 +08:00

sampling

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

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

speculative

Evict swa kv cache during decoding (#17220 )

2026-01-19 22:36:52 +08:00

tokenizer

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

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

tracing

SGLang Tracing: fix attribute errors (header extraction & bootstrap span closing) (#15693 )

2025-12-27 22:44:31 -06:00

utils

[Fix] flashinfer_trtllm intermediate_size assertion with Qwen3 + TP=8 (#16824 )

2026-01-17 09:24:05 +08:00

weight_sync

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

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

constants.py

…

environ.py

[jit-kernel] Add CuTe DSL GDN Decode Kernel (#15631 )

2026-01-18 12:54:36 -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

[Refactor] Set fp4-gemm-backend=auto on SM100 and rename fp4-gemm-backend with flashinfer_ prefix (#17309 )

2026-01-19 20:09:07 +08:00