[Fork] Code for QuaRot, an end-to-end 4-bit inference of large language models.
active 2024-04-01 → 2026-05-10 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 770 days from 2024-04-01 to 2026-05-10. Pushes: 7 total, peak 2 in a day. Pull requests: 9 total, peak 3 in a day. Issues: 132 total, peak 8 in a day. Comments: 188 total, peak 9 in a day. Stars: 423 total, peak 30 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 |
|---|---|---|---|---|
| sashkboos | 146 | 7 | 2 | 88 |
| mxjmtxrm | 18 | 0 | 0 | 6 |
| YLGH | 10 | 0 | 0 | 7 |
| Tracin | 8 | 0 | 0 | 6 |
| brisker | 8 | 0 | 0 | 8 |
| ChenMnZ | 8 | 0 | 0 | 3 |
| alexztc | 7 | 0 | 0 | 3 |
| wenhuach21 | 7 | 0 | 0 | 7 |
| Coco58323 | 5 | 0 | 0 | 4 |
| zhsky2017 | 4 | 0 | 0 | 3 |
| seamoonlight-YBY | 4 | 0 | 0 | 2 |
| JiangYongYu1 | 4 | 0 | 0 | 2 |
| JeevanBhoot | 4 | 0 | 0 | 3 |
| Niko-zyf | 4 | 0 | 0 | 2 |
| yyfcc17 | 4 | 0 | 0 | 2 |
| CryVeck | 3 | 0 | 3 | 0 |
| ZijianYY | 3 | 0 | 0 | 1 |
| yugezheng | 3 | 0 | 0 | 3 |
| telemorne | 3 | 0 | 0 | 3 |
| Oliver-ss | 3 | 0 | 0 | 2 |
Recent activity
Latest issues, pull requests and releases
- Issue#74sashkboos2026-04-07 10:03questions about feature_dimension in hadamard_utils.get_hadK()
- Issue#82kimjoohyungsd2026-04-07 07:16Question on Decoding Speedup
- Issue comment#79chairman-lu2025-11-13 01:32Hadamard-induced outliers
- Issue#79chairman-lu2025-11-12 09:54Hadamard-induced outliers
- Issue comment#77sashkboos2025-11-03 11:21Question about computational invariance in LayerNorm
- Issue comment#77Wangguanru12025-10-29 09:10Question about computational invariance in LayerNorm
- Issue#75sashkboos2025-06-17 13:19[Error Report] Export intermediate activation values of the model
- Issue comment#74sashkboos2025-05-28 11:35questions about feature_dimension in hadamard_utils.get_hadK()
- Issue comment#36sashkboos2025-05-27 12:31A question regarding the rotation matching pairs
- Issue comment#36HaoKun-Li2025-05-27 03:43A question regarding the rotation matching pairs
- Issue comment#36sashkboos2025-05-26 10:42A question regarding the rotation matching pairs
- Issue comment#73sashkboos2025-04-22 08:01Question about Hadamard transform computation complexity.
- Issue#73sashkboos2025-04-22 08:01Question about Hadamard transform computation complexity.
- Issue comment#73shinezyy2025-04-22 06:56Question about Hadamard transform computation complexity.
- Issue#73ditalker2025-04-21 20:34Question about Hadamard transform computation complexity.
- Issue comment#72Niko-zyf2025-04-16 08:49Precision Loss in bf16 Model with float32 Rotation Calculation
- Issue comment#72sashkboos2025-04-16 08:37Precision Loss in bf16 Model with float32 Rotation Calculation
- Issue#72Niko-zyf2025-04-15 04:28Precision Loss in bf16 Model with float32 Rotation Calculation
- Issue comment#71alexztc2025-04-08 01:49Running Llama7b meets 'NoneType' of attention_mask
- Issue#71alexztc2025-04-08 01:49Running Llama7b meets 'NoneType' of attention_mask
- Issue comment#71sashkboos2025-04-07 06:50Running Llama7b meets 'NoneType' of attention_mask
- Issue#71alexztc2025-04-06 14:53Running Llama7b meets 'NoneType' of attention_mask
- Issue comment#27shuren-z2025-04-05 04:13Does QuaRot only support Llama and OPT style LLM?
- Issue comment#66sashkboos2025-04-03 12:43Question about RTN results of Llama3-70b.
- Issue comment#70sashkboos2025-04-03 12:38Unclear Purpose of Rotating the lm_head in the Code
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 423 stars here means stars gained during the window, not the repo's star count.