GithubStar/ktransformers
1
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2026-10-16 21:01:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

docs: restore single-column Figure 2 in KT-FT paper

Browse Source
This commit is contained in:
Peilin Li
2026-10-10 11:58:45 +08:00
parent cde34b75f3
commit 399a5875a0
8 changed files with 20 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+2 -2
View File
@@ -165,9 +165,9 @@ If you use KTransformers in your research, please cite the relevant paper.
title = {{KTransformers-FineTune: Beyond Offload with Layout-Aware and Schedule-Optimized Heterogeneous MoE Fine-Tuning}},
author = {Li, Peilin and Hao, Xingxing and Chen, Hongtao and Xie, Weiyu and Zheng, Yaowei and Wu, Bowen and Yang, Yujie and Shen, Huanming and Ou, Qingliang and Zhang, Boxin and Tang, Jingqi and Yuan, Ziwei and Dong, Jianwei and Kuang, Dongdong and Feng, Zhangchi and Dai, Jiaheng and Yang, Qianrui and Chen, Shaoyuan and Wang, Jiahao and Han, Yaochen and Zhu, Yuening and Liao, Jiaqi and Chen, Xianglin and Ai, Zhiyuan and Wu, Yongwei and Zhang, Mingxing},
year = {2026},
note = {Author preprint, version paper-ktft-v1},
note = {Author preprint, version paper-ktft-v1.1},
howpublished = {Author preprint, GitHub},
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1/papers/kt-ft}
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1.1/papers/kt-ft}
}
```
README_ZH.md
+2 -2
View File
@@ -150,9 +150,9 @@ CUDA_VISIBLE_DEVICES=0,1,2,3 accelerate launch \
title = {{KTransformers-FineTune: Beyond Offload with Layout-Aware and Schedule-Optimized Heterogeneous MoE Fine-Tuning}},
author = {Li, Peilin and Hao, Xingxing and Chen, Hongtao and Xie, Weiyu and Zheng, Yaowei and Wu, Bowen and Yang, Yujie and Shen, Huanming and Ou, Qingliang and Zhang, Boxin and Tang, Jingqi and Yuan, Ziwei and Dong, Jianwei and Kuang, Dongdong and Feng, Zhangchi and Dai, Jiaheng and Yang, Qianrui and Chen, Shaoyuan and Wang, Jiahao and Han, Yaochen and Zhu, Yuening and Liao, Jiaqi and Chen, Xianglin and Ai, Zhiyuan and Wu, Yongwei and Zhang, Mingxing},
year = {2026},
note = {Author preprint, version paper-ktft-v1},
note = {Author preprint, version paper-ktft-v1.1},
howpublished = {Author preprint, GitHub},
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1/papers/kt-ft}
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1.1/papers/kt-ft}
}
```
papers/kt-ft/CHANGELOG.md
+4
View File
@@ -1,5 +1,9 @@
# Paper versions
## paper-ktft-v1.1 — October 10, 2026
Restore Figure 2 to a single column. The figure content, caption, manuscript text, experimental data, and references are unchanged; subsequent pages reflow.
## paper-ktft-v1 — October 8, 2026
Initial GitHub author preprint of KTransformers-FineTune, with the complete manuscript, figures, references, and 26-author citation.
papers/kt-ft/CITATION.bib
+2 -2
View File
@@ -2,7 +2,7 @@
title = {{KTransformers-FineTune: Beyond Offload with Layout-Aware and Schedule-Optimized Heterogeneous MoE Fine-Tuning}},
author = {Li, Peilin and Hao, Xingxing and Chen, Hongtao and Xie, Weiyu and Zheng, Yaowei and Wu, Bowen and Yang, Yujie and Shen, Huanming and Ou, Qingliang and Zhang, Boxin and Tang, Jingqi and Yuan, Ziwei and Dong, Jianwei and Kuang, Dongdong and Feng, Zhangchi and Dai, Jiaheng and Yang, Qianrui and Chen, Shaoyuan and Wang, Jiahao and Han, Yaochen and Zhu, Yuening and Liao, Jiaqi and Chen, Xianglin and Ai, Zhiyuan and Wu, Yongwei and Zhang, Mingxing},
year = {2026},
note = {Author preprint, version paper-ktft-v1},
note = {Author preprint, version paper-ktft-v1.1},
howpublished = {Author preprint, GitHub},
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1/papers/kt-ft}
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1.1/papers/kt-ft}
}
papers/kt-ft/README.md
+4 -4
View File
@@ -1,10 +1,10 @@
# KTransformers-FineTune: Beyond Offload with Layout-Aware and Schedule-Optimized Heterogeneous MoE Fine-Tuning
[中文](README_ZH.md) · [Read PDF](ktransformers-finetune.pdf) · [Download PDF](https://raw.githubusercontent.com/kvcache-ai/ktransformers/paper-ktft-v1/papers/kt-ft/ktransformers-finetune.pdf) · [Fine-tuning cookbook](https://github.com/kvcache-ai/ktransformers/blob/main/doc/en/SFT/KTransformers-Fine-Tuning_Cookbook.md) · [Citation](#citation)
[中文](README_ZH.md) · [Read PDF](ktransformers-finetune.pdf) · [Download PDF](https://raw.githubusercontent.com/kvcache-ai/ktransformers/paper-ktft-v1.1/papers/kt-ft/ktransformers-finetune.pdf) · [Fine-tuning cookbook](https://github.com/kvcache-ai/ktransformers/blob/main/doc/en/SFT/KTransformers-Fine-Tuning_Cookbook.md) · [Citation](#citation)
Peilin Li, Xingxing Hao, Hongtao Chen, Weiyu Xie, Yaowei Zheng, Bowen Wu, Yujie Yang, Huanming Shen, Qingliang Ou, Boxin Zhang, Jingqi Tang, Ziwei Yuan, Jianwei Dong, Dongdong Kuang, Zhangchi Feng, Jiaheng Dai, Qianrui Yang, Shaoyuan Chen, Jiahao Wang, Yaochen Han, Yuening Zhu, Jiaqi Liao, Xianglin Chen, Zhiyuan Ai, Yongwei Wu, and Mingxing Zhang.
Author preprint · Published on GitHub: October 8, 2026 · Version: `paper-ktft-v1`.
Author preprint · Published on GitHub: October 8, 2026 · Updated: October 10, 2026 · Version: `paper-ktft-v1.1`.
The arXiv link will be added when available. Corresponding author: Mingxing Zhang.
@@ -33,9 +33,9 @@ See [CITATION.bib](CITATION.bib) for the full 26-author BibTeX entry.
title = {{KTransformers-FineTune: Beyond Offload with Layout-Aware and Schedule-Optimized Heterogeneous MoE Fine-Tuning}},
author = {Li, Peilin and Hao, Xingxing and Chen, Hongtao and Xie, Weiyu and Zheng, Yaowei and Wu, Bowen and Yang, Yujie and Shen, Huanming and Ou, Qingliang and Zhang, Boxin and Tang, Jingqi and Yuan, Ziwei and Dong, Jianwei and Kuang, Dongdong and Feng, Zhangchi and Dai, Jiaheng and Yang, Qianrui and Chen, Shaoyuan and Wang, Jiahao and Han, Yaochen and Zhu, Yuening and Liao, Jiaqi and Chen, Xianglin and Ai, Zhiyuan and Wu, Yongwei and Zhang, Mingxing},
year = {2026},
note = {Author preprint, version paper-ktft-v1},
note = {Author preprint, version paper-ktft-v1.1},
howpublished = {Author preprint, GitHub},
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1/papers/kt-ft}
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1.1/papers/kt-ft}
}
```
papers/kt-ft/README_ZH.md
+4 -4
View File
@@ -1,10 +1,10 @@
# KTransformers-FineTune: Beyond Offload with Layout-Aware and Schedule-Optimized Heterogeneous MoE Fine-Tuning
[English](README.md) · [阅读 PDF](ktransformers-finetune.pdf) · [下载 PDF](https://raw.githubusercontent.com/kvcache-ai/ktransformers/paper-ktft-v1/papers/kt-ft/ktransformers-finetune.pdf) · [微调 Cookbook](https://github.com/kvcache-ai/ktransformers/blob/main/doc/zh/KTransformers-Fine-Tuning_Cookbook_zh.md) · [引用](#引用)
[English](README.md) · [阅读 PDF](ktransformers-finetune.pdf) · [下载 PDF](https://raw.githubusercontent.com/kvcache-ai/ktransformers/paper-ktft-v1.1/papers/kt-ft/ktransformers-finetune.pdf) · [微调 Cookbook](https://github.com/kvcache-ai/ktransformers/blob/main/doc/zh/KTransformers-Fine-Tuning_Cookbook_zh.md) · [引用](#引用)
Peilin Li, Xingxing Hao, Hongtao Chen, Weiyu Xie, Yaowei Zheng, Bowen Wu, Yujie Yang, Huanming Shen, Qingliang Ou, Boxin Zhang, Jingqi Tang, Ziwei Yuan, Jianwei Dong, Dongdong Kuang, Zhangchi Feng, Jiaheng Dai, Qianrui Yang, Shaoyuan Chen, Jiahao Wang, Yaochen Han, Yuening Zhu, Jiaqi Liao, Xianglin Chen, Zhiyuan Ai, Yongwei Wu, Mingxing Zhang。
作者预印本 · GitHub 发布日期：2026 年 10 月 8 日 · 版本：`paper-ktft-v1`。
作者预印本 · GitHub 发布日期：2026 年 10 月 8 日 · 更新日期：2026 年 10 月 10 日 · 版本：`paper-ktft-v1.1`。
公开 arXiv 链接可用后将补充。通讯作者：Mingxing Zhang。
@@ -33,9 +33,9 @@ LoRA Experts 将剩余显存用于共享的模型适配路径，通过系统-算
title = {{KTransformers-FineTune: Beyond Offload with Layout-Aware and Schedule-Optimized Heterogeneous MoE Fine-Tuning}},
author = {Li, Peilin and Hao, Xingxing and Chen, Hongtao and Xie, Weiyu and Zheng, Yaowei and Wu, Bowen and Yang, Yujie and Shen, Huanming and Ou, Qingliang and Zhang, Boxin and Tang, Jingqi and Yuan, Ziwei and Dong, Jianwei and Kuang, Dongdong and Feng, Zhangchi and Dai, Jiaheng and Yang, Qianrui and Chen, Shaoyuan and Wang, Jiahao and Han, Yaochen and Zhu, Yuening and Liao, Jiaqi and Chen, Xianglin and Ai, Zhiyuan and Wu, Yongwei and Zhang, Mingxing},
year = {2026},
note = {Author preprint, version paper-ktft-v1},
note = {Author preprint, version paper-ktft-v1.1},
howpublished = {Author preprint, GitHub},
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1/papers/kt-ft}
url = {https://github.com/kvcache-ai/ktransformers/tree/paper-ktft-v1.1/papers/kt-ft}
}
```
papers/kt-ft/SHA256SUMS
+2 -2
View File
@@ -1,2 +1,2 @@
3f696f0a1de2a21de6221a8ced9e992ac326bf20cf2b2ad28204730e1e1e9728 ktransformers-finetune.pdf
006d89e60e1cf9d3fb3be17292de1d1322d4a05c6cccb4d6f0fd964c224cdb77 CITATION.bib
7cf0707f508c25218f5fe76ac0f21c9d7fa21ac4278aa588cc431796329ce137 ktransformers-finetune.pdf
e90e5c21c2d9ff690973096f51bd7241b63b0ddec3ca8a84ee67507ee959d605 CITATION.bib
papers/kt-ft/ktransformers-finetune.pdf
BIN
View File
Binary file not shown.