Default Branch

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

Branches

ef79941ac9 · llama : disable FA if KV head size do not match · Updated 2024-06-17 16:20:24 +00:00    kanshan

6708
1

a235b7c532 · Vectorize q load · Updated 2024-06-17 09:30:40 +00:00    kanshan

6738
11

98f948b9d0 · unicode : avoid char32_t · Updated 2024-06-16 10:18:46 +00:00    kanshan

6723
1

28f7a4d028 · ggml : fix handling of zero blocks in IQ quants · Updated 2024-06-16 07:41:53 +00:00    kanshan

6724
1

e9f2abfc8c · bitnet : pad tensors to 256 · Updated 2024-06-15 16:01:03 +00:00    kanshan

6742
25

34bdbed481 · rpc : fix load/store misaligned addresses · Updated 2024-06-15 11:39:20 +00:00    kanshan

6726
1

eaf34ba0cd · metal : utilize max shared memory for mul_mat_id · Updated 2024-06-14 10:02:25 +00:00    kanshan

6733
1

18133cab40 · Revert "use the correct SYCL context for host USM allocations" · Updated 2024-06-13 11:08:27 +00:00    kanshan

6738
4

46325233c9 · Revert 7777 · Updated 2024-06-12 15:22:55 +00:00    kanshan

6738
1

8412561c4b · ggml : update unary asserts and "supports_op" · Updated 2024-06-12 12:25:14 +00:00    kanshan

6739
2

cd026b48ef · ggml : support more contiguous cases · Updated 2024-06-12 12:12:32 +00:00    kanshan

6754
2

4356325ef5 · tests : check the Python version · Updated 2024-06-11 06:05:15 +00:00    kanshan

6750
1

4bb03cade0 · ci : disable server-windows workflow · Updated 2024-06-10 09:30:18 +00:00    kanshan

6759
1

9e4d62e6ab · server : improve "prompt" handling · Updated 2024-06-10 06:31:50 +00:00    kanshan

6759
1

956bb14595 · examples : remove --instruct remnants · Updated 2024-06-10 05:37:47 +00:00    kanshan

6759
1

d857e5192e · quantize : check imatrix for nan/inf values · Updated 2024-06-06 20:44:24 +00:00    kanshan

6777
2

731e7528be · server : fix --threads-http arg · Updated 2024-06-06 13:37:12 +00:00    kanshan

6778
1

f7d4b7c343 · build only main and server in their docker images · Updated 2024-06-05 22:13:01 +00:00    kanshan

6785
2

3d2e79da7f · add openmp lib to dockerfiles · Updated 2024-06-05 22:05:25 +00:00    kanshan

6785
1

0085f94936 · server : add /v1/completion endpoint · Updated 2024-06-04 12:58:14 +00:00    kanshan

6795
1