Default Branch

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

ZIP  TAR.GZ

Branches

ik/fuse_moe_up_gate a6a2c89d19 · Merge remote-tracking branch 'origin/main' into ik/fuse_moe_up_gate · Updated 2025-02-22 08:39:46 +01:00    git	4211 3570	Compare	ZIP  TAR.GZ
ik/attn_gemm af790bb5fa · Cleanup · Updated 2025-02-22 05:09:54 +01:00    git	4211 3567	Compare	ZIP  TAR.GZ
ik/fix_avx512_vs_fancy_simd ffe65511cf · Hopefully this really fixes the confusion between AVX512 and FANCY_SIMD · Updated 2025-02-21 11:30:33 +01:00    git	4211 3565	Compare	ZIP  TAR.GZ
ik/issue_214 9ccc810b08 · Trying to fix confusion betweem HAVE_FANCY_SIMD and AVX512 · Updated 2025-02-21 08:06:57 +01:00    git	4211 3565	Compare	ZIP  TAR.GZ
ik/fix_neon_legacy_quants 75e11382a6 · Fix typo · Updated 2025-02-20 12:54:28 +01:00    git	4211 3564	Compare	ZIP  TAR.GZ
ik/gemm_iq1s 74c26d05c2 · iq1s: NEON · Updated 2025-02-20 10:33:36 +01:00    git	4211 3565	Compare	ZIP  TAR.GZ
ik/q8_KV 4502eab09e · Minor · Updated 2025-02-19 09:42:15 +01:00    git	4211 3575	Compare	ZIP  TAR.GZ
ik/repack_also_experts 7d020d8681 · Repack also experts · Updated 2025-02-19 08:54:48 +01:00    git	4211 3560	Compare	ZIP  TAR.GZ
ik/gemm_4d bdc882bfac · Moving 4D gemm logic from ggml.c to iqk_mul_mat.cpp · Updated 2025-02-14 17:03:19 +01:00    git	4211 3558	Compare	ZIP  TAR.GZ
ik/mla_q80 a0ee859784 · MLA: allow Q8_0 K-cache for MLA · Updated 2025-02-13 11:44:05 +01:00    git	4211 3557	Compare	ZIP  TAR.GZ
ik/mla_fixes f875ed00e8 · MLA: compile time option to not use transposed KV cache · Updated 2025-02-13 09:54:42 +01:00    git	4211 3562	Compare	ZIP  TAR.GZ
ik/fix_missing_bf16_avx512 8861e7a4ef · One more · Updated 2025-02-12 12:56:54 +01:00    git	4211 3556	Compare	ZIP  TAR.GZ
ik/fix_imatrix_nonsense 1044af2d95 · Fix imatrix overprotectiveness · Updated 2025-02-11 17:05:57 +01:00    git	4211 3554	Compare	ZIP  TAR.GZ
ik/fattn_Dk_Dv 4066235b8f · FA: very slightly faster for nq = 1 (TG) · Updated 2025-02-10 17:25:44 +01:00    git	4211 3555	Compare	ZIP  TAR.GZ
s6/warmup 370274317b · Unify warmup to one token · Updated 2025-02-09 23:05:16 +01:00    git	4211 3553	Compare	ZIP  TAR.GZ
ik/try_trellis c13027bcaf · Merge remote-tracking branch 'origin/main' into ik/try_trellis · Updated 2025-02-09 19:00:41 +01:00    git	4211 3605	Compare	ZIP  TAR.GZ
ik/iqk_fattn_all_quants 01e2b0c2ce · FA: Add option to build all FA kernels · Updated 2025-02-09 17:50:50 +01:00    git	4211 3550	Compare	ZIP  TAR.GZ
ik/mla c6cfbc79a9 · Better gemm strategy when nth > nhead · Updated 2025-02-09 12:03:18 +01:00    git	4211 3554	Compare	ZIP  TAR.GZ
s6/mla bf1d056125 · Make sure we do have wk_b and wv_b before enabling MLA · Updated 2025-02-09 08:24:52 +01:00    git	4211 3555	Compare	ZIP  TAR.GZ
ik/iq1_s_r4_k128 03dc7bd787 · Cleanup · Updated 2025-02-09 08:14:12 +01:00    git	4211 3554	Compare	ZIP  TAR.GZ