sglang

4-gpu-models

[Qwen3.5] Enable MTP spec_v2 and add test for nvidia/Qwen3.5-397B-A17B-NVFP4 (#19391 )

2026-03-04 14:01:25 -08:00

8-gpu-models

[Nightly] Replace MiniMax-M2 with MiniMax-M2.5 (#20083 )

2026-03-07 01:15:34 -08:00

amd

[AMD] [DeepSeek-OCR-2 Day 0] Enable DeepSeek-OCR-2 on AMD GPUs and add nightly test (#19732 )

2026-03-10 17:04:35 -07:00

ascend

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

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

attention

fix flaky test for test_kda_kernels (#19864 )

2026-03-04 22:47:29 +08:00

backends

[FlashInfer v0.6.4] [RL] Integrate FlashInfer mxfp8 gemm, MoE, and routed MoE (#19537 )

2026-03-10 15:37:57 -07:00

bench_fn

[Benchmark] Fix generated_shared_prefix attribute naming and remove args dependency (#19363 )

2026-02-25 18:45:54 -08:00

constrained_decoding

…

core

[Utils] Add NetworkAddress abstraction for IPv6-safe address handling (#20306 )

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

debug_utils

Enhance error resilience in dump comparator (#19685 )

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

disaggregation

[Performance] Decode Offload improves the long texts performance 100% through dynamic block offload. (#17216 )

2026-03-08 17:16:53 +08:00

distributed

[Qwen3.5] Support Qwen3.5 Pipeline Parallelism (#19670 )

2026-03-07 23:34:08 +08:00

dllm

[DLLM] Add initial radix cache support (#18724 )

2026-03-04 23:24:09 -08:00

ep

…

eval

Fix nightly VLM accuracy: gemma3n TP fixes + removal, latency thresholds (#19401 )

2026-02-27 14:24:02 -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

hicache

[Benchmark] Remove re-exports from bench_serving.py (#19130 )

2026-02-21 14:30:30 -08:00

kernels

[DSv32] [GLM5] Improve Model Quality by Avoiding FP32 Precision Loss in weights_proj (#19041 )

2026-02-22 16:20:51 +08:00

layers

Return intermediate Mamba states (#19716 )

2026-03-09 16:04:36 +08:00

lora

[Bug] Fix lora tp bug on H200 (#19769 )

2026-03-04 20:11:02 -08:00

metrics

[Bug] Fix missing TTFT histogram for single-batch requests (#20122 )

2026-03-08 15:18:51 -07:00

mla

Relax flaky test thresholds for MLA DeepSeek V3 and AutoRound (#20068 )

2026-03-06 17:26:28 -08:00

model_loading

Fix/nemotron mtp quantaized (#19433 )

2026-03-03 01:07:46 -08:00

models

[AMD] Fix AMD CI - stage-b-small-1-gpu-amd (partition 7) (#20028 )

2026-03-06 23:49:14 -08:00

moe

…

openai_server

[AMD] Fix JSON mode streaming test flake on AMD CI (#20293 )

2026-03-10 17:24:11 -07:00

ops

[AMD] CI - Add MI35x nightly/PR tests for kv-cache-fp8 and allreduce-fusion (DeepSeek) (#19834 )

2026-03-05 07:09:57 -08:00

parser

Add KimiK2Detector with tool interruption support (#19696 )

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

perf

[TestFix] change LoRA tests to use NVIDIA adapter instead of Nutanix (#19642 )

2026-03-02 12:55:41 -08:00

piecewise_cuda_graph

[CI] Refactor PCG related CI (#19994 )

2026-03-10 14:49:25 +08:00

profiling

…

quant

[FlashInfer v0.6.4] [RL] Integrate FlashInfer mxfp8 gemm, MoE, and routed MoE (#19537 )

2026-03-10 15:37:57 -07:00

radix_cache

[Feature] Add SLRU eviction policy & fix RadixCache hit_count bug (#18843 )

2026-03-08 21:30:55 +08:00

rl

[Scheduler] Unify idle checks into is_fully_idle() and fix weight update test (#20296 )

2026-03-10 17:50:23 -07:00

rotary

feat: Support mrope_section with rope_type: "yarn" (#13313 )

2026-02-14 22:51:44 +08:00

sampling

…

scheduler

add mixed chunk unit test and make small refactors (#18776 )

2026-03-08 19:56:09 +08:00

sessions

Enable piecewise-cuda-graph when logprob_start_len = -1 (#19453 )

2026-03-10 12:50:57 -07:00

spec

[CI] Relax Eagle infer_b spec accept length threshold (#20300 )

2026-03-10 15:40:52 -07:00

stress

…

tokenizer

…

utils

[Utils] Add NetworkAddress abstraction for IPv6-safe address handling (#20306 )

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

vlm

[VLM] Replace decord with torchcodec for video decoding (#20055 )

2026-03-09 19:23:49 +08:00

test_hybrid_dp_ep_tp_mtp.py

…

test_srt_backend.py

…