Default Branch

9cf3bf256b · py : bump numpy to 2.4.6 (#28649) · Updated 2026-09-09 20:59:37 +08:00

Branches

2fcdf869cd · batched-bench : add mmq CLI arg · Updated 2023-10-12 00:42:33 +08:00    tqcq

9132
7

ee7456926e · ggml-alloc : fix assert in debug builds · Updated 2023-10-09 20:33:12 +08:00    tqcq

9141
1

ee268b5446 · llama : no longer perform uninitialized access to the KV cache · Updated 2023-10-08 16:49:38 +08:00    tqcq

9148
5

acead654d2 · Merge branch 'master' into fix-refact · Updated 2023-10-08 16:25:16 +08:00    tqcq

9148
4

6b9554a740 · metal : print more GPU info + disable mul_mm for MTLGPUFamiliy < Apple7 · Updated 2023-10-08 14:55:13 +08:00    tqcq

9155
5

ba44776dc2 · bump version · Updated 2023-10-08 02:47:48 +08:00    tqcq

9154
6

5ab6c2132a · server-parallel : add "--reverse-prompt" + compiler warning fixes · Updated 2023-10-06 19:32:19 +08:00    tqcq

9167
4

5418932b71 · llama : fix comments for llama_kv_cache API · Updated 2023-10-04 02:01:52 +08:00    tqcq

9192
5

c5650ed470 · server : avoid context swaps by shifting the KV cache · Updated 2023-09-29 00:03:36 +08:00    tqcq

9216
57

72e7ef4e53 · simple : fixes · Updated 2023-09-27 05:19:36 +08:00    tqcq

9242
48

784d14ed31 · llama : store non-RoPEd K cache (WIP) · Updated 2023-09-18 04:43:07 +08:00    tqcq

9254
5

92a4f86879 · llama : make starcoder graph build more consistent with others · Updated 2023-09-15 22:57:10 +08:00    tqcq

9264
20

e7e7b11455 · llama : remove experimental stuff · Updated 2023-09-15 03:52:01 +08:00    tqcq

9276
3

2f689dee06 · metal : minor · Updated 2023-09-07 20:33:21 +08:00    tqcq

9309
5

30ac7a4117 · gitignore : metal · Updated 2023-09-05 03:23:16 +08:00    tqcq

9321
12

f3a84b2e0d · llama : better express the KV cache dependencies in the graph · Updated 2023-09-05 02:44:48 +08:00    tqcq

9321
5

c79d130f74 · make : fix speculative build · Updated 2023-09-04 20:50:04 +08:00    tqcq

9322
9

847896aba7 · speculative : add --draft CLI arg · Updated 2023-09-03 18:51:07 +08:00    tqcq

9328
3

8c2b881281 · cuda : poc for norm quants (only -b 1 works) · Updated 2023-08-31 02:42:28 +08:00    tqcq

9369
3

b4e70822f6 · metal : add poc for normalized Q4_0 and Q4_1 · Updated 2023-08-30 23:47:16 +08:00    tqcq

9369
7