🎯An accuracy-first, highly efficient quantization toolkit for LLMs, designed to minimize quality degradation across Weight-Only Quantization, MXFP4, NVFP4, GGUF, and adaptive schemes.
Python · active 2025-03-08 → 2026-08-09 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 520 days from 2025-03-08 to 2026-08-09. Pushes: 3,193 total, peak 38 in a day. Pull requests: 628 total, peak 10 in a day. Issues: 569 total, peak 20 in a day. Comments: 1,659 total, peak 26 in a day. Stars: 270 total, peak 14 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 |
|---|---|---|---|---|
| wenhuach21 | 2.3K | 740 | 217 | 620 |
| n1ck-guo | 758 | 473 | 92 | 81 |
| xin3he | 716 | 332 | 63 | 135 |
| yiliu30 | 614 | 283 | 48 | 134 |
| Copilot | 559 | 56 | 19 | 304 |
| pre-commit-ci[bot] | 461 | 459 | 2 | 0 |
| WeiweiZhang1 | 291 | 152 | 51 | 29 |
| chensuyue | 270 | 163 | 19 | 25 |
| XuehaoSun | 264 | 167 | 24 | 19 |
| mengniwang95 | 206 | 102 | 18 | 46 |
| lvliang-intel | 157 | 103 | 10 | 26 |
| Kaihui-intel | 66 | 36 | 11 | 10 |
| azure-pipelines[bot] | 62 | 0 | 0 | 62 |
| luoyu-intel | 33 | 22 | 3 | 3 |
| Zhenzhong1 | 32 | 29 | 1 | 2 |
| hshen14 | 31 | 0 | 0 | 11 |
| lkk12014402 | 29 | 0 | 2 | 15 |
| thuang6 | 27 | 3 | 0 | 5 |
| ZaneMark | 22 | 14 | 2 | 3 |
| a32543254 | 16 | 8 | 0 | 2 |
Recent activity
Latest issues, pull requests and releases
- Pull request#2106xin3he2026-07-31 06:17
- Issue comment#1889xin3he2026-07-13 05:43[Regresion]: multi-cards tuning+eval causes exceptions at eval phase
- Issue comment#2014AutoRoundBot2026-07-08 02:12fix the mismatch between weight and scale in gguf format
- Issue#2016xin3he2026-07-07 08:21[Bug]: Tuning MXFP uses higher memory than expected
- Issue#1689chensuyue2026-07-07 04:32[Bug]: dry-run estimation should fall back to text_config for Qwen3.5-MoE
- Issue#1996XuehaoSun2026-07-06 07:05KeyError: '.weight_scale$' Error: DecompressExperts on tensors destined for model.embed_tokens.weight. Ckpt contains: 1
- Pull request#1982xin3he2026-07-02 03:26
- Issue comment#1979ramkrishs2026-07-01 21:58fix: promote None-initialized kwarg to list in forward_capture
- Issue comment#1679wenhuach212026-07-01 02:26[Bug]: torch._dynamo hit config.recompile_limit (8) ... tensor 'v' size mismatch at index 0 ...
- Issue comment#1745wenhuach212026-07-01 02:20[Feature]: support AutoScheme in model free mode
- Pull request#1949xin3he2026-07-01 02:19
- Issue#1788wenhuach212026-07-01 02:18[Feature]: Refine device operations into a unified class to simplify support for new devices
- Issue comment#1949azure-pipelines[bot]2026-06-30 12:38Fix layer_config bug and CI unstable issue
- Pull request#1957luoyu-intel2026-06-30 06:27
- Issue comment#1956lkk120144022026-06-26 07:11rename the base rotation (deterministic/random hadamard) directory name.
- Issue#1591github-actions[bot]2026-06-20 22:49[Feature]: add --dry-run estimation mode
- Issue comment#1897azure-pipelines[bot]2026-06-18 08:55DDP for multiple cards
- Issue#1915thuang62026-06-17 12:53[Bug]: Use auto round quant minimax m2.7, the moe gate weight will be quanted as default.
- Issue comment#1927chensuyue2026-06-17 04:59Fix UT coverage
- Issue comment#1897azure-pipelines[bot]2026-06-17 03:05DDP for multiple cards
- Issue#1925XuehaoSun2026-06-17 02:04[Bug]: Llama-3.3-70B-Instruct q4_k_m q2_k_mixed inference OOM
- Pull request#1920n1ck-guo2026-06-13 01:22
- Pull request#1920n1ck-guo2026-06-13 01:22
- Issue#1112github-actions[bot]2026-06-12 22:57NVFP4 RTN is missing act_max for lm_head
- Issue#1908chensuyue2026-06-11 01:22[Bug] GGUF Q6_K export writes NaN block scales for all-zero blocks (0/0 in make_qx_quants_chunk) — NaN logits in llama.cpp
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 270 stars here means stars gained during the window, not the repo's star count.