PyTorch implementation of the Differential-Transformer architecture for sequence modeling, specifically tailored as a decoder-only model similar to large language models (LLMs). The architecture incorporates a novel Differential Attention mechanism, Multi-Head structure, RMSNorm, and SwiGLU.
active 2024-10-08 → 2025-12-23 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 442 days from 2024-10-08 to 2025-12-23. Pushes: 5 total, peak 2 in a day. Pull requests: 2 total, peak 1 in a day. Issues: 4 total, peak 1 in a day. Comments: 4 total, peak 1 in a day. Stars: 70 total, peak 4 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 |
|---|---|---|---|---|
| nanowell | 9 | 5 | 1 | 2 |
| KernelInterrupt | 3 | 0 | 1 | 2 |
| zziC7 | 1 | 0 | 0 | 0 |
| Taning-by | 1 | 0 | 0 | 0 |
| xuanyin777 | 1 | 0 | 0 | 0 |
Recent activity
Latest issues, pull requests and releases
- Issue#4Taning-by2025-04-20 05:40仅仅训练一个鸟类分类数据集，很难收敛
- Issue comment#3nanowell2024-10-28 15:14the shape of q,k,v
- Issue#3zziC72024-10-28 07:16the shape of q,k,v
- Issue comment#1nanowell2024-10-27 19:43形状冲突
- Issue#1nanowell2024-10-27 19:43形状冲突
- Pull request#2nanowell2024-10-27 19:43
- Issue comment#1KernelInterrupt2024-10-25 04:15形状冲突
- Pull request#2KernelInterrupt2024-10-25 04:14
- Issue comment#1KernelInterrupt2024-10-24 18:57形状冲突
- Issue#1xuanyin7772024-10-12 09:20形状冲突
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 70 stars here means stars gained during the window, not the repo's star count.