逐步从零开始在 PyTorch 中实现一个类似 ChatGPT 的大语言模型
Jupyter Notebook · active 2025-05-05 → 2026-08-10 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 463 days from 2025-05-05 to 2026-08-10. Pushes: 172 total, peak 9 in a day. Pull requests: 164 total, peak 7 in a day. Issues: 86 total, peak 9 in a day. Comments: 234 total, peak 17 in a day. Stars: 20,759 total, peak 878 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 |
|---|---|---|---|---|
| rasbt | 429 | 171 | 122 | 95 |
| review-notebook-app[bot] | 48 | 0 | 0 | 48 |
| d-kleine | 35 | 0 | 5 | 23 |
| murphyk | 12 | 0 | 0 | 5 |
| labdmitriy | 9 | 0 | 0 | 8 |
| PrinceSajjadHussain | 9 | 0 | 1 | 5 |
| weezymatt | 8 | 0 | 4 | 1 |
| casinca | 8 | 0 | 3 | 5 |
| RahulYadav-tech-gif | 5 | 0 | 0 | 0 |
| VINEAY | 4 | 0 | 0 | 4 |
| aetrstrstesterst-hash | 3 | 0 | 0 | 0 |
| jeremy-feng | 3 | 0 | 0 | 0 |
| yuedajiong | 3 | 0 | 0 | 3 |
| kira-offgrid | 3 | 0 | 3 | 0 |
| omifounder | 3 | 0 | 0 | 0 |
| joselpart | 3 | 0 | 0 | 3 |
| 17Swagat | 3 | 0 | 0 | 0 |
| gdeisat | 3 | 0 | 0 | 3 |
| gziz | 3 | 0 | 2 | 0 |
| LokeshReddy001 | 2 | 0 | 0 | 1 |
Recent activity
Latest issues, pull requests and releases
- Issue#1036TentenMarchhhh2026-06-12 23:44ask about project
- Issue#1020rasbt2026-05-17 16:29Chapter 5 `train_model_simple` should control randomness
- Issue comment#1020rasbt2026-05-17 16:29Chapter 5 `train_model_simple` should control randomness
- Pull request#1026rasbt2026-05-17 16:13
- Issue comment#1014chtarunkumar2026-05-05 03:54Add Deepseek Sparse Attention (DSA) implementation (ch04/09_dsa)
- Issue comment#1021rasbt2026-05-03 08:13Architecture Gallery:
- Issue#1011aetrstrstesterst-hash2026-04-15 05:05关于免费额度的替代方案
- Issue#1011aetrstrstesterst-hash2026-04-15 05:05关于免费额度的替代方案
- Issue#1011aetrstrstesterst-hash2026-04-15 05:05关于免费额度的替代方案
- Issue comment#1004rasbt2026-04-11 00:31Handle short-token edge case in BPE pair selection
- Issue comment#1005review-notebook-app[bot]2026-04-11 00:17Add Gemma 4 E2B and E4B (from scratch)
- Pull request#1000rasbt2026-04-06 01:46
- Pull request#999rasbt2026-04-04 17:31
- Pull request#998rasbt2026-04-04 17:06
- Issue comment#996rasbt2026-04-04 17:04Please revert `--exclude-newer`
- Issue comment#978brandonedwards12314-collab2026-04-03 15:22Support DeepSeek Sparse Attention (DSA) like DeepSeek V3.2
- Issue comment#995brandonedwards12314-collab2026-04-03 15:21😂 AI Hallucination: When ChatGPT Makes Up a Python Library That Doesn't Exist
- Issue comment#991rasbt2026-03-31 00:36Actions
- Issue#990Jack14472026-03-30 03:43A latent mistake about the semantic representation of the message within predicted label computation in LLM classification fine-tuning
- Issue#990Jack14472026-03-30 03:43A latent mistake about the semantic representation of the message within predicted label computation in LLM classification fine-tuning
- Issue comment#989RocMarshal2026-03-29 13:23Fix the ssl issue for MacOS M1 when downloading gpt model related resources.
- Issue#984rasbt2026-03-21 00:35Architecture Gallery: GOODSAM
- Issue comment#979casinca2026-03-15 17:56Architecture Gallery: The context length for Qwen3 may be incorrect
- Issue comment#978rasbt2026-03-14 21:18Support DeepSeek Sparse Attention (DSA) like DeepSeek V3.2
- Issue comment#976review-notebook-app[bot]2026-03-08 04:04feat:完成了torch注释的添加
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 20,759 stars here means stars gained during the window, not the repo's star count.