Default Branch

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

ZIP  TAR.GZ

Branches

ik/fix_vulkan_required a4a7334aad · Minor · Updated 2025-07-02 16:08:58 +02:00    git	4211 3780	Compare	ZIP  TAR.GZ
ik/iq3_ks_v2 59967f3d64 · iq3_ks: Metal gemv - pathetic performance · Updated 2025-07-02 09:14:52 +02:00    git	4211 3789	Compare	ZIP  TAR.GZ
ik/vulkan_disable_fused_ops d9fd346cb6 · Conditionally disable fused ops when building with Vulkan enabled · Updated 2025-07-02 08:56:58 +02:00    git	4211 3776	Compare	ZIP  TAR.GZ
ik/improve_mmq 036b5bb822 · Minor · Updated 2025-06-27 18:00:58 +02:00    git	4211 3776	Compare	ZIP  TAR.GZ
ik/cuda_large_cpy 409bfe6648 · Remove what appears to be unnecessary asserts in ggml_cuda_cpy · Updated 2025-06-26 19:27:50 +02:00    git	4211 3773	Compare	ZIP  TAR.GZ
ik/mmq_to_cublas 3dbc84377f · Use cuBLAS for large batches and quants with block size 16 · Updated 2025-06-26 13:13:00 +02:00    git	4211 3773	Compare	ZIP  TAR.GZ
ik/cuda_iqk_r4 b3417c9366 · iqk_r4 quants: use MMQ only for batches < 1024 tokens · Updated 2025-06-25 13:47:59 +02:00    git	4211 3775	Compare	ZIP  TAR.GZ
ik/falcon_edge b74bd33b6a · Add Falcon-Edge support · Updated 2025-06-25 09:11:37 +02:00    git	4211 3771	Compare	ZIP  TAR.GZ
ik/gemm_neon_1bit e5e5acfdda · iq1_m · Updated 2025-06-24 14:05:30 +02:00    git	4211 3771	Compare	ZIP  TAR.GZ
ik/gemm_neon_kquants e18b10bc48 · Merge remote-tracking branch 'origin/main' into ik/gemm_neon_kquants · Updated 2025-06-24 13:03:02 +02:00    git	4211 3780	Compare	ZIP  TAR.GZ
ik/gemm_neon_iquants 548a5f3f0d · iq3_s · Updated 2025-06-23 15:34:41 +02:00    git	4211 3772	Compare	ZIP  TAR.GZ
ik/gemm_neon_iqk ec4b4536ea · iq2_k · Updated 2025-06-23 11:15:25 +02:00    git	4211 3773	Compare	ZIP  TAR.GZ
ik/gemm_neon_legacy aaa164773d · q5_1 · Updated 2025-06-21 16:00:22 +02:00    git	4211 3770	Compare	ZIP  TAR.GZ
ik/metal_new_trellis a0ba58e9b9 · iq2_kt and iq3_kt work with new int trellis · Updated 2025-06-20 09:47:22 +02:00    git	4211 3763	Compare	ZIP  TAR.GZ
ik/trellis_opt 5b677c3caf · Enable next_128() also on AVX2 · Updated 2025-06-20 08:41:22 +02:00    git	4211 3768	Compare	ZIP  TAR.GZ
ik/neon_iq3_kt a45e368444 · iq3_kt is now working on NEON · Updated 2025-06-20 06:25:48 +02:00    git	4211 3763	Compare	ZIP  TAR.GZ
ik/fix_neon_build 1e534df8ea · Fix NEON build · Updated 2025-06-19 17:35:16 +02:00    git	4211 3761	Compare	ZIP  TAR.GZ
ik/fix_538 19ac0b595c · Fix missed block_q8_x2 bf16 -> i16 change · Updated 2025-06-19 08:02:42 +02:00    git	4211 3757	Compare	ZIP  TAR.GZ
s6/readme-minor1 829a4d7177 · move thing fix · Updated 2025-06-18 19:36:41 +02:00    git	4211 3759	Compare	ZIP  TAR.GZ
ik/new_iq2kt_v2 6da5afa3f6 · The trellis quants now need super-blocks of 256, so we need a check · Updated 2025-06-18 14:35:54 +02:00    git	4211 3786	Compare	ZIP  TAR.GZ