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

[Ckpt Engine] feat: new sglang entrypoint support for update (#12216 )

2025-10-30 10:39:27 +08:00

compilation

Custom All Reduce for Piecewise Cuda Graph (#15356 )

2025-12-25 23:55:15 +08:00

configs

fix: change class name of GLM-ASR (#15772 )

2025-12-26 00:15:16 +08:00

connector

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

constrained

Adding tool calling and reasoning parser support for Intern-S1 (#14866 )

2025-12-16 00:01:58 -08:00

debug_utils

Super tiny add more info in dumper (#14145 )

2025-11-30 09:07:39 +08:00

disaggregation

Support PP for zmq_to_scheduler (#15312 )

2025-12-23 17:07:55 +08:00

distributed

[VLM] Support ViT Piecewise CUDA Graph for Qwen3-VL (#15320 )

2025-12-20 21:00:07 +08:00

dllm

Tiny extract ModelRunnerOutput (#15400 )

2025-12-18 22:18:45 +08:00

elastic_ep

[2/N] Added the core structure of elastic EP and the eplb algorithm with faulty rank (#10606 )

2025-10-22 01:13:31 -07:00

entrypoints

[model-gateway]: add gRPC router embeddings endpoint implementation (#15273 )

2025-12-23 09:12:22 -08:00

eplb

Support EPLB balancedness prometheus metric without GPU->CPU synchronize (#15401 )

2025-12-18 22:24:23 +08:00

function_call

Update MiniMax-M2 ToolCall and add MiniMax-M2.1 in Docs (#15538 )

2025-12-23 15:11:52 -08:00

grpc

[model-gateway]: add gRPC router embeddings endpoint implementation (#15273 )

2025-12-23 09:12:22 -08:00

hardware_backend/npu

[NPU] [BUGFIX] [CRITICAL!] Fix NPU inference (torch_npu._npu_reshape_and_cache() crash) (#15484 )

2025-12-20 14:53:46 +08:00

layers

Optimize FP8 MLA KV cache writes with Triton kernel (#15522 )

2025-12-25 12:35:39 -08:00

lora

Feature/Fix multi lora scheduler blocking issue and evict LoRA None lastly (#14795 )

2025-12-12 17:13:05 -08:00

managers

Clean up the __init__ of TokenizerManager and DetokenizerManager (#15796 )

2025-12-25 10:30:27 -08:00

mem_cache

Optimize FP8 MLA KV cache writes with Triton kernel (#15522 )

2025-12-25 12:35:39 -08:00

metrics

Add LoRA metrics for potential auto scaling (#15149 )

2025-12-24 12:13:52 -08:00

model_executor

Custom All Reduce for Piecewise Cuda Graph (#15356 )

2025-12-25 23:55:15 +08:00

model_loader

[Feature] support fastsafetensors (#15091 )

2025-12-23 22:33:56 +08:00

models

fix: change class name of GLM-ASR (#15772 )

2025-12-26 00:15:16 +08:00

multimodal

fix: change class name of GLM-ASR (#15772 )

2025-12-26 00:15:16 +08:00

multiplex

[Feature] PD-Multiplexing Context and Scheduler, lazy import spatial. (#12275 )

2025-11-01 00:40:01 +08:00

parser

fix: prevent points regex from matching checkpoints/endpoints (#15120 )

2025-12-17 12:38:18 -08:00

sampling

Add customized sampler registration (#15423 )

2025-12-18 23:10:23 -08:00

speculative

[2/N] clean duplicate code of logprob processing in spec. (#15593 )

2025-12-24 15:21:13 +08:00

tokenizer

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

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

tracing

[minor] remove debug code in python/sglang/srt/compilation/weak_ref_tensor_jit.py (#13235 )

2025-11-14 12:22:24 -08:00

utils

feat: log request when e2e latency exceeds the specified value (#15759 )

2025-12-25 23:12:35 +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

custom_op.py

Clean up the __init__ of TokenizerManager and DetokenizerManager (#15796 )

2025-12-25 10:30:27 -08:00

environ.py

feat: log request when e2e latency exceeds the specified value (#15759 )

2025-12-25 23:12:35 +08:00

server_args_config_parser.py

[Lint] Add python/sglang to ruff F401 checks and remove unused imports in files (#11685 )

2025-10-17 16:49:46 -07:00

server_args.py

[MiMoV2Flash] fix: respect --swa-full-tokens-ratio arg (#15488 )

2025-12-25 21:02:56 +08:00