Default Branch

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

ZIP  TAR.GZ

Branches

ik/fix_flash_attn de91911d7a · Fix Zen4 Flash Attention · Updated 2024-09-02 14:51:10 +02:00    git	4211 3408	Compare	ZIP  TAR.GZ
ik/fix_multiple_choice 6bc273c1d6 · Do not process prompts containing binary data for escapes · Updated 2024-09-02 08:12:08 +02:00    git	4211 3407	Compare	ZIP  TAR.GZ
ik/zen4_flash_attn a66d1fc562 · Update FlashAttn comment · Updated 2024-09-01 11:47:50 +02:00    git	4211 3409	Compare	ZIP  TAR.GZ
ik/kq_fused_softmax 1b834ac6e4 · Flash attention: templated implementation · Updated 2024-08-31 12:10:36 +02:00    git	4211 3421	Compare	ZIP  TAR.GZ
ik/fix_no_iqk_build eb046ae5f3 · Fix build when iqk_mul_mat is disabled · Updated 2024-08-31 07:52:08 +02:00    git	4211 3405	Compare	ZIP  TAR.GZ
ik/kq_mask 3b4fe65e1c · Minor · Updated 2024-08-28 17:31:27 +02:00    git	4211 3417	Compare	ZIP  TAR.GZ
ik/fused_softcap_softmax e4f200098b · Flash attention with softcap: Metal · Updated 2024-08-26 18:34:43 +02:00    git	4211 3409	Compare	ZIP  TAR.GZ
ik/softcap_minor c9116e9eca · softcap: minor improvement · Updated 2024-08-21 11:34:35 +02:00    git	4211 3403	Compare	ZIP  TAR.GZ
ik/softcap cc3d42e60b · softcap, tanh: avoid NaNs for large arguments (NEON) · Updated 2024-08-20 14:46:30 +02:00    git	4211 3408	Compare	ZIP  TAR.GZ
ik/iq4_k_tweaks 3e97ec87a2 · iq4_k: use iq5_k also when n_gqa = 2 · Updated 2024-08-20 08:29:45 +02:00    git	4211 3401	Compare	ZIP  TAR.GZ
ik/quantize_q8k_avx2 6b6e2f2dbc · AVX2 quantization for Q8_K · Updated 2024-08-19 14:31:55 +02:00    git	4211 3400	Compare	ZIP  TAR.GZ
ik/quantize_stats ff471dfd61 · quantize_stats: print rmse and max error as fraction of <x> · Updated 2024-08-19 12:47:19 +02:00    git	4211 3399	Compare	ZIP  TAR.GZ
ik/iq2_k_tweak b2212f170c · iq2_k: slightly better bpw - accuracy compromise · Updated 2024-08-19 12:33:19 +02:00    git	4211 3398	Compare	ZIP  TAR.GZ
ik/skip_noop_barriers 686e75650e · Skip barriers of noops · Updated 2024-08-14 08:49:12 +02:00    git	4211 3397	Compare	ZIP  TAR.GZ
ik/merge_Aug_12_2024 28bb16556d · Remove CI check · Updated 2024-08-12 11:28:44 +02:00    git	4211 3397	Compare	ZIP  TAR.GZ
ik/fix_Makefile 2c9aaae809 · Fix Makefile · Updated 2024-08-09 16:29:32 +02:00    git	4211 3394	Compare	ZIP  TAR.GZ
ik/iq6_k bf745357ec · Fix Zen4 implementation of iq3_k, iq4_k, iq5_k · Updated 2024-08-09 09:32:07 +02:00    git	4211 3393	Compare	ZIP  TAR.GZ
ik/iq2_tn 8178075f84 · iq2_tn: small NEON improvement · Updated 2024-08-06 12:08:22 +02:00    git	4211 3393	Compare	ZIP  TAR.GZ
ik/trinet 54481695a0 · q2_K: allow it to detect ternary nets and quantize accordingly · Updated 2024-08-05 10:59:36 +02:00    git	4211 3382	Compare	ZIP  TAR.GZ
ik/faster_iq3_iq5_quantize de818b77d6 · iq3_k, iq5_k: faster quantization · Updated 2024-08-05 07:13:53 +02:00    git	4211 3380	Compare	ZIP  TAR.GZ