Default Branch

319146247e · vulkan: improve partial offloading performance on AMD (#19976) · Updated 2026-03-01 17:32:14 +01:00

Branches

98a040561c · vulkan: tune MMVQ for Intel Windows · Updated 2026-02-28 16:48:01 +01:00

1
1

4687e98072 · add command and file auto-completion · Updated 2026-02-28 15:35:40 +01:00

2
1

fa3e83dbac · fix inplace error · Updated 2026-02-28 11:42:04 +01:00

29
12

4b436e4e5e · flake8 fix · Updated 2026-02-23 11:48:01 +01:00

75
20

5d45884106 · metal : fix build · Updated 2026-02-18 08:14:31 +01:00

182
23

c0c3e428dd · refactor · Updated 2026-02-16 22:02:45 +01:00

119
49

5da56dc1d8 · args : add -kvu to llama-parallel · Updated 2026-02-12 20:50:01 +01:00

182
17

e7fbfc9b80 · ci : tmp fixes · Updated 2026-02-11 14:48:40 +01:00

232
22

5372fc6461 · wip · Updated 2026-02-10 22:44:42 +01:00

197
18

b9b56b017e · Apply suggestion from @ggerganov (src->buffer to buf_src) v2 · Updated 2026-02-10 12:00:44 +01:00

197
13

5144018e7b · cont : simplify · Updated 2026-02-07 13:50:05 +01:00

218
4

1213a03564 · qwen3next : fix chunking · Updated 2026-02-04 09:06:38 +01:00

251
1

5b01d8575d · examples : add compare-mlx · Updated 2026-01-31 08:57:35 +01:00

288
1

6c8a04576e · experiments · Updated 2026-01-28 08:45:07 +01:00

339
29

8b407e3978 · quant : manual overrides of tensor types take precedence · Updated 2026-01-20 10:20:24 +01:00

403
1

3bfbbcc5fc · winget : update komac version · Updated 2026-01-18 09:29:03 +01:00

413
1

e2751545b9 · cont : inline verification · Updated 2026-01-17 13:33:07 +01:00

425
5

36f0132464 · CUDA: Factor out and re-use block_reduce function (#18785) · Updated 2026-01-15 03:44:54 +01:00

444
0
Included

60864997fe · fit-params : print signed int for -ngl param · Updated 2026-01-14 18:59:23 +01:00

447
1

5292965711 · Merge branch 'master' into xsn/lora_keep_track · Updated 2026-01-13 13:44:22 +01:00

462
4