Train speculative decoding models effortlessly and port them smoothly to SGLang serving.
Python · active 2025-07-25 → 2026-08-14 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 389 days from 2025-07-22 to 2026-08-14. Pushes: 396 total, peak 18 in a day. Pull requests: 231 total, peak 10 in a day. Issues: 112 total, peak 6 in a day. Comments: 809 total, peak 33 in a day. Stars: 360 total, peak 72 in a day.
- Pushes
- Pull requests
- Issues
- Comments
- Stars
Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.
Top contributors
Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| gemini-code-assist[bot] | 439 | 0 | 0 | 322 |
| FrankLeeeee | 330 | 137 | 37 | 94 |
| sleepcoo | 144 | 61 | 29 | 20 |
| jiapingW | 118 | 47 | 12 | 53 |
| yubofredwang | 111 | 19 | 23 | 38 |
| zyksir | 102 | 64 | 10 | 18 |
| maocheng23 | 33 | 30 | 3 | 0 |
| FlamingoPg | 25 | 1 | 2 | 17 |
| xiaomin-D | 24 | 0 | 7 | 14 |
| Ximingwang-09 | 17 | 0 | 8 | 8 |
| xiaoxi-s | 16 | 8 | 4 | 2 |
| ggg-s | 16 | 0 | 1 | 11 |
| github-actions[bot] | 15 | 15 | 0 | 0 |
| uygnef | 14 | 5 | 3 | 5 |
| ZhengHSI | 14 | 1 | 3 | 7 |
| hanq-moreh | 13 | 0 | 3 | 6 |
| fzyzcjy | 12 | 0 | 8 | 3 |
| fan-niu | 12 | 0 | 0 | 12 |
| Copilot | 12 | 0 | 0 | 8 |
| shimizust | 11 | 0 | 6 | 3 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#681gemini-code-assist[bot]2026-07-15 03:09[PR678 stack 00/14] fix(bench): correct HumanEval scoring
- Pull request#653maocheng232026-07-06 07:33
- Issue comment#647gemini-code-assist[bot]2026-07-03 23:49[DataFlow runtime] fix: NameError on device/device_type in scripts/train_dflash.py main()
- Issue comment#631gemini-code-assist[bot]2026-07-01 03:07[DataFlow runtime] Phase B1 — TargetEngine ABC + de-EAGLE3 the target boundary
- Issue comment#533moehanabi2026-06-29 10:11[Bug] There is a gap between the acceptance rates of training and inference
- Issue comment#612gemini-code-assist[bot]2026-06-28 00:33[DataFlow runtime · M6 4/4] MooncakeFeatureStore — RDMA fast-path backend
- Pull request#607maocheng232026-06-28 00:33
- Pull request#606maocheng232026-06-28 00:33
- Pull request#581Hayden7272026-06-14 03:38
- Issue comment#563jiapingW2026-06-11 04:59fix: correct 9 critical/high bugs found via multi-agent audit
- Issue comment#551Dogacel2026-05-31 01:19[Bug] Chat Template Not Applied Correctly to Recent Models
- Issue comment#563gemini-code-assist[bot]2026-05-26 11:03fix: correct 9 critical/high bugs found via multi-agent audit
- Issue comment#558jiapingW2026-05-26 00:12Support sharded target logits for EAGLE3 online training
- Issue comment#496yangshen88-rgb2026-05-14 03:35upgrade to sglang==0.5.9 and support qwen3.5 eagle3
- Issue comment#194Rachelcoll2026-05-07 17:49[Bug] train error
- Pull request#543lianakoleva2026-04-24 00:42
- Issue#539jjtql2026-04-17 03:20[Bug] Segfault encountered in training draft model
- Issue comment#538gemini-code-assist[bot]2026-04-17 00:45Fix/gemma3 eagle3 hooks
- Pull request#538tcligg2026-04-17 00:45
- Issue comment#524laoconeth2026-04-14 04:51feat: reduce Eagle3 training memory spike via all-to-all sharding
- Issue comment#531gemini-code-assist[bot]2026-04-13 20:48Fix: Ensure aux hidden states are correctly set for EAGLE3 in Gemma3.…
- Issue comment#528gemini-code-assist[bot]2026-04-09 04:25Reduce peak GPU memory in Eagle3 online target generation by avoiding an extra logits copy
- Pull request#527liusy582026-04-08 09:16
- Pull request#525liusy582026-04-08 08:00
- Issue comment#466laoconeth2026-04-06 11:03[Bug] Abnormal memory usage and Out-of-Memory in eagle3 training.
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 360 stars here means stars gained during the window, not the repo's star count.