PyTorch-native framework for fault-tolerant distributed training of foundation models on AI clusters
active 2026-04-21 → 2026-08-08 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 110 days from 2026-04-21 to 2026-08-08. Pushes: 71 total, peak 6 in a day. Pull requests: 7 total, peak 2 in a day. Issues: 7 total, peak 2 in a day. Comments: 6 total, peak 2 in a day. Stars: 1 total, peak 1 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 |
|---|---|---|---|---|
| amazloumi | 44 | 38 | 1 | 0 |
| mmshad | 24 | 13 | 5 | 1 |
| camilobrownpinilla | 16 | 16 | 0 | 0 |
| Naeemkh | 10 | 3 | 0 | 1 |
| codecov[bot] | 3 | 0 | 0 | 3 |
| dependabot[bot] | 2 | 0 | 1 | 1 |
| Copilot | 1 | 1 | 0 | 0 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#151mmshad2026-07-18 23:14Bump bleach from 6.3.0 to 6.4.0
- Issue comment#152dependabot[bot]2026-07-18 23:14Bump soupsieve from 2.8.3 to 2.8.4
- Pull request#153dependabot[bot]2026-07-18 23:05
- Pull request#153mmshad2026-07-18 23:05
- Issue#121amazloumi2026-06-13 22:49Add LatentMoE support (latent-space routed experts, moe_latent_dim)
- Issue#104mmshad2026-05-16 06:15Async checkpoint save can deadlock: DCP background save thread and `save()` barrier use the same process group
- Issue comment#101codecov[bot]2026-05-15 22:15Lift coverage on metrics/* and training/grad.py to 100%
- Issue#92mmshad2026-05-13 03:22Document Mode, tokenizer pairing, and max_seq_len rule in VLM configs
- Issue comment#89codecov[bot]2026-05-13 01:13Promote Adapter to a registered component
- Issue comment#85codecov[bot]2026-05-13 00:05Replace dead nlphuji/coco_karpathy default with sayakpaul/coco-30-val-2014
- Issue#82mmshad2026-05-12 23:35Trim default-restating lines and fix stale inline comments in VLM configs
- Pull request#81mmshad2026-05-12 23:16
- Pull request#75amazloumi2026-04-30 22:25
- Issue#71amazloumi2026-04-28 19:04Docs update to suggest running `uv python install 3.12` before `uv sync`
- Issue comment#62Naeemkh2026-04-26 17:51Gate train_state.pt load behind an ownership check
- Pull request#58mmshad2026-04-26 15:56
- Issue#61mmshad2026-04-22 01:43train_state.pt is loaded with weights_only=False without a trust check
- Issue#55mmshad2026-04-22 01:43MemoryMappedDataset leaks partial mmaps when init fails
- Pull request#60mmshad2026-04-22 01:11
- Pull request#50mmshad2026-04-22 00:09
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 1 stars here means stars gained during the window, not the repo's star count.