Commits · ca7366d2dbb81751e65f1fa8b7b5fba9165f9f42 · wqshmzh / ktransformers

22 Feb, 2025 2 commits
- fix merge bug, this branch also padding Marlin · f7f10598
  Atream authored Feb 22, 2025
  
  f7f10598
- optimize gguf dequant, save mem, support Q2_K · 5ec33d04
  Atream authored Feb 22, 2025
```
use marlin for lm_head, lm_head only calc last token for prefill
extend context window to 19K for DeepSeek-V3/R1 within 24GB VRAM
```
  5ec33d04
21 Feb, 2025 1 commit
- optimize GPU · 7e1fe256
  Atream authored Feb 21, 2025
  
  7e1fe256
19 Feb, 2025 1 commit
- feat: Support Moore Threads GPU · 2207f6cd
  Xiaodong Ye authored Feb 19, 2025
```
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
```
  2207f6cd
02 Sep, 2024 1 commit
- Support IQ4_XS dequantize · be356c1b
  Yap Sok Ann authored Sep 02, 2024
  
  be356c1b
12 Aug, 2024 2 commits
- [feature] support q2_k & q3_k dequantize on gpu · 7c4cb520
  BITcyman authored Aug 12, 2024
  
  7c4cb520
- [ADD] support multi-gpu qlen>1 q5_k · f5f79f5c
  chenxl authored Aug 12, 2024
  
  f5f79f5c
27 Jul, 2024 1 commit
- Initial commit · 18c42e67
  chenxl authored Jul 27, 2024
  
  18c42e67