Default Branch

2da6686176 · Fix stale tensor-split params for draft models (#24814) · Updated 2026-07-05 18:39:36 +00:00

Branches

ced26486ff · cont · Updated 2025-04-07 12:24:01 +00:00    kanshan

4816
6

fe564b0dfb · ci : rename job MSVC -> MinGW · Updated 2025-04-04 10:51:48 +00:00    kanshan

4831
1

43ab09b85d · ci : testing (wip) · Updated 2025-04-04 10:43:43 +00:00    kanshan

4831
1

7a73e861a7 · cont · Updated 2025-04-04 09:02:20 +00:00    kanshan

4840
4

c875e03f96 · rpc : update README for cache usage · Updated 2025-03-28 07:41:47 +00:00    kanshan

4897
1

efe0222130 · media : add SVG logo [no ci] · Updated 2025-03-27 21:07:46 +00:00    kanshan

4900
1

70b063a550 · metal : reduce register pressure · Updated 2025-03-26 19:24:28 +00:00    kanshan

4928
8

20b256e0fd · convert : match ssm_conv tensors by type · Updated 2025-03-25 18:29:22 +00:00    kanshan

4921
2

e94c2bd360 · ggml : improve repack templates · Updated 2025-03-25 07:42:31 +00:00    kanshan

4933
2

b8b7885484 · SYCL: disable Q4_0 reorder optimization · Updated 2025-03-25 04:44:11 +00:00    kanshan

4926
1

a5b1943912 · ggml-quants : fix some edge cases in make_qkxh_nl_quants · Updated 2025-03-23 21:59:37 +00:00    kanshan

4937
13

35c2f8b9ff · llama-vocab : add SuperBPE pre-tokenizer · Updated 2025-03-23 20:19:11 +00:00    kanshan

4934
1

b8b173274d · server : remove old commented code [no ci] · Updated 2025-03-20 16:20:54 +00:00    kanshan

4964
45

7a3c178d78 · speculative : adapt to new llama API · Updated 2025-03-18 20:05:44 +00:00    kanshan

4964
36

29acf2cf05 · context : move the change to llama_context::encode() · Updated 2025-03-18 09:55:19 +00:00    kanshan

4968
2

f6711cef44 · CUDA: determine FA parallel blocks at runtime · Updated 2025-03-16 13:36:57 +00:00    kanshan

5036
1

c4aca65582 · hparams : add SWA rope parameters · Updated 2025-03-13 17:26:09 +00:00    kanshan

4995
1

21fe0ce4eb · hparams : add comment [no ci] · Updated 2025-03-13 15:56:38 +00:00    kanshan

4996
2

ed58975f51 · server : improve infill stop criteria · Updated 2025-03-12 13:28:48 +00:00    kanshan

5007
1

87dae2fd15 · Vulkan: Print coopmat shapes, then exit · Updated 2025-03-09 10:53:55 +00:00    kanshan

5021
1