Ongoing research training transformer models at scale on linear attention models
Python · active 2025-04-06 → 2026-08-10 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 492 days from 2025-04-06 to 2026-08-10. Pushes: 5,417 total, peak 50 in a day. Pull requests: 2,159 total, peak 47 in a day. Issues: 1,103 total, peak 36 in a day. Comments: 5,460 total, peak 84 in a day. Stars: 1,777 total, peak 33 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 |
|---|---|---|---|---|
| copy-pr-bot[bot] | 4.4K | 3.9K | 0 | 574 |
| ko3n1g | 1.1K | 644 | 246 | 118 |
| github-actions[bot] | 755 | 0 | 185 | 367 |
| svcnvidia-nemo-ci | 748 | 124 | 275 | 257 |
| github-merge-queue[bot] | 613 | 613 | 0 | 0 |
| sbhavani | 533 | 0 | 78 | 184 |
| Phlip79 | 512 | 57 | 118 | 239 |
| tdene | 405 | 0 | 84 | 196 |
| chtruong814 | 336 | 49 | 66 | 32 |
| santhnm2 | 255 | 0 | 44 | 147 |
| claude[bot] | 252 | 0 | 0 | 116 |
| deepakn94 | 195 | 0 | 19 | 82 |
| Victarry | 181 | 0 | 24 | 122 |
| yaox12 | 175 | 1 | 33 | 87 |
| yanring | 163 | 0 | 15 | 84 |
| jaredcasper | 162 | 1 | 3 | 63 |
| dimapihtar | 161 | 0 | 53 | 85 |
| shjwudp | 160 | 0 | 36 | 82 |
| cspades | 149 | 0 | 31 | 58 |
| maanug-nv | 148 | 0 | 48 | 49 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#6312svcnvidia-nemo-ci2026-08-07 22:37Add Nemotron 3.5 Lightning GB200 nightly test
- Pull request#6359wujingyue2026-08-07 22:24
- Pull request#6354maanug-nv2026-08-07 19:54
- Issue comment#6350svcnvidia-nemo-ci2026-08-07 19:29fix(ci): remove community workflow PAT
- Issue comment#6229YangFei19902026-08-07 19:20[NCCL EP] Support eager mode and overflow handling
- Issue comment#6229claude[bot]2026-08-07 18:42[NCCL EP] Support eager mode and overflow handling
- Issue comment#6161wujingyue2026-08-07 18:01[Prototype] MatrixFSDP-style TensorAtomic placement for MFSDP v2 Muon
- Issue comment#6074deepakn942026-08-07 17:53Gated Delta Product (GDP) implementation
- Issue comment#6314cspades2026-08-07 17:45[QUESTION]Megatron-FSDP ZeRO-2 (optim_grads) has higher peak memory than ZeRO-1/ZeRO-3 due to unsharded temporary main_grad bucket allocation
- Pull request#6043github-actions[bot]2026-08-07 17:41
- Issue comment#5972wujingyue2026-08-07 17:37Support HFSDP deferred DP-outer gradient reduction
- Issue comment#6351copy-pr-bot[bot]2026-08-07 17:24cp: `set weights_only=True (6150)` into `core_r0.19.0`
- Issue comment#6351ko3n1g2026-08-07 17:23cp: `set weights_only=True (6150)` into `core_r0.19.0`
- Pull request#6351ko3n1g2026-08-07 17:23
- Pull request#6351ko3n1g2026-08-07 17:23
- Issue comment#5972Achyuthan-S2026-08-07 17:18Support HFSDP deferred DP-outer gradient reduction
- Issue comment#6169svcnvidia-nemo-ci2026-08-07 16:55Encapsulate symmetric-memory AVG in DBuffer
- Pull request#6336svcnvidia-nemo-ci2026-08-07 16:48
- Issue comment#5956copy-pr-bot[bot]2026-08-07 16:35Divide expert gradients by the expert-parallel size
- Pull request#6153svcnvidia-nemo-ci2026-08-07 16:32
- Pull request#6153svcnvidia-nemo-ci2026-08-07 16:31
- Issue comment#6350copy-pr-bot[bot]2026-08-07 16:22fix(ci): remove community workflow PAT
- Pull request#6350ko3n1g2026-08-07 16:22
- Pull request#6338github-actions[bot]2026-08-07 08:01
- Issue comment#6317copy-pr-bot[bot]2026-08-06 08:07[Dev] Make FSDP manual UBR registration deterministic
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 1,777 stars here means stars gained during the window, not the repo's star count.