MoBA：长上下文 LLM 的块注意混合。开源版deepseek的NSA
active 2025-02-18 → 2026-05-11 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 448 days from 2025-02-18 to 2026-05-11. Pushes: 9 total, peak 3 in a day. Pull requests: 3 total, peak 1 in a day. Issues: 54 total, peak 5 in a day. Comments: 58 total, peak 7 in a day. Stars: 1,754 total, peak 458 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 |
|---|---|---|---|---|
| whitelez | 47 | 8 | 0 | 22 |
| wxthu | 7 | 0 | 0 | 5 |
| ZetangForward | 6 | 0 | 0 | 3 |
| hewr2010 | 6 | 1 | 1 | 3 |
| 1180301012 | 4 | 0 | 0 | 4 |
| CacatuaAlan | 4 | 0 | 0 | 2 |
| yibozhong | 3 | 0 | 1 | 2 |
| Jiangfeng-Xiong | 3 | 0 | 0 | 2 |
| monellz | 3 | 0 | 0 | 1 |
| yliuls | 3 | 0 | 0 | 0 |
| bestfleer | 3 | 0 | 0 | 1 |
| haochengxia | 3 | 0 | 0 | 1 |
| jianzongwu | 2 | 0 | 0 | 1 |
| yinian-lw | 2 | 0 | 0 | 2 |
| martyLY | 2 | 0 | 0 | 1 |
| vefalun | 2 | 0 | 0 | 1 |
| bubbliiiing | 1 | 0 | 0 | 1 |
| weiye666 | 1 | 0 | 0 | 1 |
| BILLYisavailable | 1 | 0 | 0 | 0 |
| weixuansun | 1 | 0 | 0 | 0 |
Recent activity
Latest issues, pull requests and releases
- Pull request#39ReyJerry2025-12-17 17:38
- Issue#37BILLYisavailable2025-08-26 13:29Question: Triton Kernel Implementation in MoBA
- Issue#35YuHuiGao2025-06-16 04:01moba注意力机制是否可以不经过训练直接插入其他模型中使用？
- Issue comment#22derby-ding2025-05-30 07:38Qwen2.5使用moba之后显示sliding window attention相关问题
- Issue comment#34weiye6662025-05-28 08:37Question on Implementing Context Parallelism with MoBA Efficient Attention
- Issue comment#34whitelez2025-05-27 16:12Question on Implementing Context Parallelism with MoBA Efficient Attention
- Issue#33yliuls2025-05-19 04:250519
- Issue#33yliuls2025-05-19 04:150519
- Issue comment#32ZetangForward2025-05-02 12:51mismatch max_seqlen in _flash_attn_varlen_forward
- Issue#32ZetangForward2025-05-02 12:50mismatch max_seqlen in _flash_attn_varlen_forward
- Issue#27ZetangForward2025-05-02 12:41请问moba_topk = min(moba_topk - 1, num_filtered_chunk) 这里设计的原因是什么？
- Issue#31wxthu2025-04-26 06:35Trained MoBA Llama model
- Issue comment#30monellz2025-04-21 10:37Question about decode stage
- Issue#30monellz2025-04-21 10:37Question about decode stage
- Issue comment#31wxthu2025-04-21 06:32Trained MoBA Llama model
- Issue comment#3011803010122025-04-21 04:48Question about decode stage
- Issue comment#30wxthu2025-04-21 02:13Question about decode stage
- Issue comment#30whitelez2025-04-20 11:12Question about decode stage
- Issue#19whitelez2025-04-20 11:08Question about paper Figure 3 (a) and (b)
- Issue comment#31whitelez2025-04-20 11:08Trained MoBA Llama model
- Issue#31wxthu2025-04-19 13:52Trained MoBA Llama model
- Issue comment#3011803010122025-04-18 09:41Question about decode stage
- Issue comment#30wxthu2025-04-18 06:38Question about decode stage
- Issue comment#3011803010122025-04-18 05:58Question about decode stage
- Issue comment#30wxthu2025-04-18 01:16Question about decode stage
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 1,754 stars here means stars gained during the window, not the repo's star count.