Default Branch

b2cb4512c5 · Create parameters overview (#1269) · Updated 2026-02-20 07:20:56 +01:00

Branches

07516cec2d · This appears to work · Updated 2026-02-20 08:53:06 +01:00    git

0
2

d2e193d711 · More standard attn for Qwen3-Next · Updated 2026-02-18 15:55:31 +01:00    git

6
4

4f62026fb5 · Fix very low bpw missing imatrix check · Updated 2026-02-18 09:08:48 +01:00    git

7
1

de8ae4b491 · Improve ggml_compute_forward_dup_bytes · Updated 2026-02-18 08:49:19 +01:00    git

7
2

88f5258ed3 · Don't disable CUDA graphs for Qwen3-Next · Updated 2026-02-17 16:19:51 +01:00    git

8
1

f029c3d092 · Avoid some more repeats · Updated 2026-02-17 11:46:43 +01:00    git

10
8

40674c4c31 · Faster CPU PP performance for Qwen3-Next - optimize concat · Updated 2026-02-16 11:48:09 +01:00    git

11
1

400efc23b6 · Faster Qwen3-Next PP on CUDA - optimize concat · Updated 2026-02-16 11:22:38 +01:00    git

12
1

46c40571a6 · Merge remote-tracking branch 'origin/main' into ik/qwen3next · Updated 2026-02-15 18:44:08 +01:00    git

15
76

908aa6493f · fix build error · Updated 2026-02-15 16:39:45 +01:00    git

15
1

9d9c6261b5 · GLM-5 support · Updated 2026-02-14 08:07:08 +01:00    git

19
1

34731cc9ff · spec: show warnings instead of abort · Updated 2026-02-13 00:41:20 +01:00    git

20
3

1cd03aaab7 · Typo · Updated 2026-02-08 16:15:44 +01:00    git

21
2

4d4e33a6c5 · Be able to read uint32_t and bool arrays from GGUFs · Updated 2026-02-07 17:58:44 +01:00    git

23
1

e0396aee0f · fix model name missing in final response · Updated 2026-02-07 16:40:38 +01:00    git

25
1

5bd4f8c419 · Also read rope_freq_base_train_swa from the GGUF · Updated 2026-02-07 06:44:13 +01:00    git

29
2

927593d424 · It works now, but performance gain is very minor · Updated 2026-02-06 13:55:11 +01:00    git

27
19

cb29b9a15c · Cleanup · Updated 2026-02-06 10:47:30 +01:00    git

28
3

9e1e1c0b5a · Bespoke ggml_repeat for Step3.5-Flash · Updated 2026-02-05 17:51:38 +01:00    git

31
1

e60a56c317 · Fix #1237 · Updated 2026-02-05 17:19:55 +01:00    git

33
1