Default Branch

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

Branches

9af90481d0 · Vulkan: Add renderdoc tracing support · Updated 2025-01-12 13:47:36 +00:00    kanshan

5416
1

fbddb26250 · ggml-cuda : use i and j instead of i0 and i in vec_dot_tq2_0_q8_1 · Updated 2025-01-12 02:06:49 +00:00    kanshan

5423
7

9605c5fb28 · cmake : remove explicit _XOPEN_SOURCE · Updated 2025-01-06 11:02:48 +00:00    kanshan

5468
2

aa014d7e89 · Use mutex instead of atomics for vk_instance counters · Updated 2024-12-30 05:14:58 +00:00    kanshan

5481
2

fe9235d795 · Force max subgroup size for coopmat shaders · Updated 2024-12-18 07:26:27 +00:00    kanshan

5527
1

4fbb801a9d · ggml : update ggml_backend_cpu_device_supports_op · Updated 2024-12-17 16:09:02 +00:00    kanshan

5537
3

3e92f4ecbe · cont [no ci] · Updated 2024-12-15 10:36:03 +00:00    kanshan

5549
2

7e9208e408 · scripts : change build path to "build-bench" for compare-commits.sh · Updated 2024-12-15 09:47:30 +00:00    kanshan

5549
1

fb18934a97 · gguf-py : bump version to 0.11.0 · Updated 2024-12-11 21:13:31 +00:00    kanshan

5570
0
Included

4f3a7e279b · Force max subgroup size for coopmat shaders · Updated 2024-12-10 20:27:04 +00:00    kanshan

5578
2

b8d1b1a5e1 · server : fix infill prompt format · Updated 2024-12-08 20:12:11 +00:00    kanshan

5588
1

a6648b9df7 · server : chunked prefill support · Updated 2024-12-08 07:48:18 +00:00    kanshan

5592
1

a8046c888a · use calloc instead of malloc · Updated 2024-12-04 16:24:35 +00:00    kanshan

5623
3

81611bef72 · server : add tests · Updated 2024-12-04 11:11:26 +00:00    kanshan

5623
3

33d7b70c88 · server : do not speculate during prompt processing · Updated 2024-12-03 08:58:43 +00:00    kanshan

5636
1

3c8a2a83fe · shmem experiments · Updated 2024-11-26 13:17:38 +00:00    kanshan

5699
3

dafedd33d2 · 4x4 -> 4x · Updated 2024-11-26 12:54:02 +00:00    kanshan

5699
2

bf3494345e · metal : some mul_mv experiments · Updated 2024-11-26 12:48:50 +00:00    kanshan

5699
1

b83cae088c · speculative : add infill mode · Updated 2024-11-26 09:14:17 +00:00    kanshan

5704
1

4ff0831ce6 · metal : use F16 math in mul_mat kernels · Updated 2024-11-25 13:15:26 +00:00    kanshan

5717
1