Default Branch

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

ZIP  TAR.GZ

Branches

ik/poc_tp ec45020e37 · Leave FFN partial results as f16 · Updated 2025-11-28 08:25:20 +01:00    git	182 15	Compare	ZIP  TAR.GZ
ik/try_fix_1014 4c4c84ba7f · Attempt to fix #1014 · Updated 2025-11-27 10:11:23 +01:00    git	4211 4028	Compare	ZIP  TAR.GZ
ik/fix_1015 0a6e650e29 · Fix llama-bench mla parameter · Updated 2025-11-27 09:30:58 +01:00    git	4211 4028	Compare	ZIP  TAR.GZ
fcp/fix_rpc_device 2339d41d2e · Change default RPC order and fix wrong RPC order in --device arg · Updated 2025-11-26 03:32:00 +01:00    git	187 1	Compare	ZIP  TAR.GZ
ik/gguf_py_add_maxfp4 43f9f342dd · Add MXFP4 to gguf-py constants · Updated 2025-11-24 15:42:33 +01:00    git	4211 4025	Compare	ZIP  TAR.GZ
ik/iq4_nl_cache 422585d726 · Enable iq4_nl KV cache on CUDA · Updated 2025-11-24 09:39:14 +01:00    git	4211 4024	Compare	ZIP  TAR.GZ
ik/fix_q6_0_dequantize 8297d10111 · Fix q6_0 dequantize · Updated 2025-11-24 09:04:46 +01:00    git	4211 4023	Compare	ZIP  TAR.GZ
ik/disable_rope_cache 99e0e334a5 · Disable RoPE cache · Updated 2025-11-24 07:08:07 +01:00    git	4211 4021	Compare	ZIP  TAR.GZ
ik/support_gigachat 0369d2ba44 · Gigachat: CPU FA (needs 192 x 192 for MLA = 3) · Updated 2025-11-21 10:44:34 +01:00    git	4211 4018	Compare	ZIP  TAR.GZ
ik/wip_sync_llama 2e4bfed583 · WIP: try syncing - not working yet · Updated 2025-11-20 14:30:43 +01:00    git	196 1	Compare	ZIP  TAR.GZ
ik/fix_requantize_interleaved b9d25dc35b · Fix requatizing from row-interleaved quants · Updated 2025-11-20 11:45:56 +01:00    git	197 1	Compare	ZIP  TAR.GZ
ik/gguf_py_changes_for_np2.0 8f7dd2f06b · Make gguf-py stuff work with numpy 2.0 · Updated 2025-11-20 10:11:01 +01:00    git	199 1	Compare	ZIP  TAR.GZ
ik/fix_kimi2_parse 4b731fe333 · Fix junja -> junja · Updated 2025-11-20 09:01:21 +01:00    git	199 3	Compare	ZIP  TAR.GZ
ik/disable_sm_row 00259c14a7 · Also llama-bench · Updated 2025-11-19 16:14:52 +01:00    git	200 2	Compare	ZIP  TAR.GZ
ik/try_fix_974 810c47fc38 · Attempt to fix #974 · Updated 2025-11-19 13:50:35 +01:00    git	202 1	Compare	ZIP  TAR.GZ
ik/deepseek_guarantee_rope_fusion c1d0738a1b · Make sure we can fuse Q and K RoPE for DeepSeek models · Updated 2025-11-19 13:39:34 +01:00    git	204 1	Compare	ZIP  TAR.GZ
ik/topk_moe_with_norm f514891418 · Fuse sum_rows and div with topk-moe · Updated 2025-11-19 11:14:33 +01:00    git	204 1	Compare	ZIP  TAR.GZ
ik/fused_rope_rope 5195e38d47 · Fuse Q and K RoPE · Updated 2025-11-18 13:05:15 +01:00    git	206 1	Compare	ZIP  TAR.GZ
ik/add_vq_help a1c32c1d39 · Add usage for -vq, --validate-quants · Updated 2025-11-17 16:00:51 +01:00    git	4211 4004	Compare	ZIP  TAR.GZ
fcp/context_shift_fix 415015f386 · Handle context shift better to reduce pp · Updated 2025-11-17 02:35:25 +01:00    git	211 1	Compare	ZIP  TAR.GZ