sglang

batch_invariant_ops

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

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

batch_overlap

[Score API] Add return_pooled_hidden_states to Scoring API for SequenceClassification / RewardModel (#22427 )

2026-04-15 14:58:56 -07:00

checkpoint_engine

update pre-commit config (#18860 )

2026-02-16 00:18:31 +08:00

compilation

[bugfix] avoid attention padding tokens computation in pcg (#17706 )

2026-04-14 16:08:23 +08:00

configs

Revert "feat: Support MXFP4 quantized dense models on AMD CDNA2/CDNA3 GPUs (#19143 )" (#23031 )

2026-04-16 21:53:25 -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] eagle/eagle3 speculative decoding conflicts with xgrammar in NPU (#20989 )

2026-04-15 23:34:23 -07:00

debug_utils

Add e2e demo test in dump comparator (#21031 )

2026-03-20 22:41:01 +08:00

disaggregation

[Bug Fix] Ensure prefill_info_table is populated before honoring disagg_prefill_dp_rank (#22990 )

2026-04-17 11:10:31 +08:00

distributed

[Feature][JIT Kernel] Fused TP QK norm For Minimax (#20673 )

2026-04-13 20:29:47 +08:00

dllm

Decouple scheduler log printing from metrics collection (#20107 )

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

elastic_ep

[5/N] (Elastic EP) Use GPU P2P to exchange expert weights during EPLB as much as possible (#12068 )

2026-03-16 18:40:58 +08:00

entrypoints

add check for none status code in FinishAbort (#22535 )

2026-04-16 16:21:07 -07:00

eplb

[sgl] perf optimization for eplb (#21232 )

2026-04-14 22:52:17 +08:00

function_call

Fix tool call constrained decoding and parsing for models with native formats (#21593 )

2026-04-10 20:37:23 -07:00

grpc

[gRPC] Extract gRPC servicer into standalone package (#20478 )

2026-03-13 09:13:29 -07:00

hardware_backend

[MUSA][9/N] Add FA3 attention backend support through MATE (MUSA AI Tensor Engine) (#22051 )

2026-04-10 14:18:39 -07:00

layers

Revert "feat: Support MXFP4 quantized dense models on AMD CDNA2/CDNA3 GPUs (#19143 )" (#23031 )

2026-04-16 21:53:25 -07:00

lora

[lora] Speedup triton backend sgemm calls with better grid (#22386 )

2026-04-15 13:47:07 -07:00

managers

[sgl] improve accuracy of additional page requirement during spec decode (#22406 )

2026-04-16 15:50:51 -07:00

mem_cache

[UnifiedRadixTree]: Add HiCache hook interface for TreeComponent (#22924 )

2026-04-17 12:09:41 +08:00

model_executor

[Score API] Add return_pooled_hidden_states to Scoring API for SequenceClassification / RewardModel (#22427 )

2026-04-15 14:58:56 -07:00

model_loader

feat: add coordinated checkpoint prefetch for network filesystem loading (#20843 )

2026-04-16 20:08:19 -07:00

models

[AMD] Qwen3.5 MXFP4 breaks after shared expert fusion is enabled (#22948 )

2026-04-16 15:25:33 -07:00

multimodal

[VLM] Reduce GPU memory footprint of CUDA IPC MM feature transport (#22662 )

2026-04-17 10:38:36 +08:00

multiplex

[mem] Flatten memory checkers into composable per-pool invariant checks (#22562 )

2026-04-11 02:56:22 -07:00

observability

Rename _alive_streaming_session_count; use _is_streaming helper (#22755 )

2026-04-14 13:26:03 -07:00

parser

[New Model] Gemma 4 (#21952 )

2026-04-06 20:24:44 -07:00

ray

[Ray] Bind scheduler actors to GPU-local NUMA node (#22989 )

2026-04-16 14:52:15 -07:00

sampling

fix: server crash when stop_token_ids contains null (#22175 )

2026-04-10 11:42:23 -07:00

speculative

streaming session: spec v2 bonus accounting + comprehensive test matrix (#22651 )

2026-04-15 17:12:41 -07:00

tokenizer

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

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

utils

[VLM] Reduce GPU memory footprint of CUDA IPC MM feature transport (#22662 )

2026-04-17 10:38:36 +08:00

weight_sync

…

constants.py

Clean up detokenizer and remove dead multimodal_gen code (#21588 )

2026-03-28 21:44:40 -07:00

environ.py

feat: add coordinated checkpoint prefetch for network filesystem loading (#20843 )

2026-04-16 20:08:19 -07:00

server_args_config_parser.py

[ConfigArgumentMerger] Improve ConfigArgumentMerger compatibility with external callers (#17051 )

2026-01-16 15:32:42 +08:00

server_args.py

Revert "feat: Support MXFP4 quantized dense models on AMD CDNA2/CDNA3 GPUs (#19143 )" (#23031 )

2026-04-16 21:53:25 -07:00