rl from zero pretrain, can it be done? šĢ¶š¾Ģ¶'Ģ¶š Ģ¶š Ģ¶ Ģ¶šĢ¶š¾Ģ¶š¾Ģ¶ yes.
active 2025-05-30 ā 2026-03-27 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 302 days from 2025-05-30 to 2026-03-27. Pushes: 90 total, peak 10 in a day. Pull requests: 29 total, peak 5 in a day. Issues: 38 total, peak 10 in a day. Comments: 27 total, peak 6 in a day. Stars: 184 total, peak 38 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 ā ā95% 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 ā stars and forks excluded, so this is contribution rather than popularity
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| tokenbender | 166 | 89 | 23 | 18 |
| ChinmayK0607 | 6 | 0 | 2 | 2 |
| KShivendu | 2 | 0 | 0 | 1 |
| huaishengzhu | 2 | 0 | 0 | 1 |
| devin-ai-integration[bot] | 2 | 0 | 1 | 1 |
| NaOH678 | 1 | 0 | 0 | 0 |
| Occupying-Mars | 1 | 0 | 1 | 0 |
| goyalayus | 1 | 0 | 0 | 1 |
| linear[bot] | 1 | 0 | 0 | 0 |
Recent activity
Latest issues, pull requests and releases
- Issue#44linear[bot]2026-02-20 03:35Epoch math + eval interval
- Issue comment#48goyalayus2025-12-14 08:15Critic model training
- Issue#47NaOH6782025-09-20 14:37The hyperparameters to reproduce
- Issue comment#45KShivendu2025-08-20 21:59Missing TokenBender/avataRL dataset
- Issue#45KShivendu2025-08-20 21:59Missing TokenBender/avataRL dataset
- Issue comment#45tokenbender2025-08-19 21:53Missing TokenBender/avataRL dataset
- Issue#42tokenbender2025-08-17 13:13optimize avatarl loss computation - eliminate python loops and gpu starvation
- Issue#38tokenbender2025-08-17 13:12Questions about training critic model or teacher model
- Pull request#43tokenbender2025-08-16 22:29
- Pull request#43tokenbender2025-08-16 22:27
- Issue comment#40tokenbender2025-08-16 11:19Centralized Critic Model with Pipelined Inference for Memory Optimization
- Issue#40tokenbender2025-08-16 11:19Centralized Critic Model with Pipelined Inference for Memory Optimization
- Pull request#41tokenbender2025-08-15 12:09
- Pull request#41tokenbender2025-08-15 12:09
- Issue comment#38tokenbender2025-08-15 04:36Questions about training critic model or teacher model
- Issue comment#38huaishengzhu2025-08-14 18:35Questions about training critic model or teacher model
- Issue comment#38tokenbender2025-08-12 08:15Questions about training critic model or teacher model
- Pull request#39Occupying-Mars2025-08-11 09:26
- Issue#38huaishengzhu2025-08-10 06:17Questions about training critic model or teacher model
- Issue#31tokenbender2025-08-09 23:38Enable Fused AdamW Optimizer
- Issue#32tokenbender2025-08-09 23:38Add Pre-tokenized Data Loading Support
- Issue#33tokenbender2025-08-09 23:38Update README to reflect current project status
- Issue#35tokenbender2025-08-09 23:38Modal training artifacts not persisted - checkpoints lost after training
- Issue#28tokenbender2025-08-09 23:38Implement Separate Weight Decay Groups
- Issue#29tokenbender2025-08-09 23:38Add MFU (Model FLOPs Utilization) Tracking
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals ā 184 stars here means stars gained during the window, not the repo's star count.