Default Branch

master ee4c505a4f · server: add dedup-cache-models preset option (#27346) · Updated 2026-08-19 17:04:26 +08:00

ZIP  TAR.GZ

Branches

ceb/bert-tokenizer-fixes a37696d4f1 · speculative : more robust tokenizer comparison · Updated 2024-04-05 06:28:13 +08:00    tqcq	7886 9		ZIP  TAR.GZ
gg/flash-attn-a 4c190ba676 · cuda : reduce registers · Updated 2024-03-29 03:17:08 +08:00    tqcq	7928 77		ZIP  TAR.GZ
compilade/fix-command-r 64b7d85891 · llama : fix command-r inference · Updated 2024-03-28 18:22:24 +08:00    tqcq	7933 1		ZIP  TAR.GZ
gg/flash-attn-wip 6be02b5969 · cuda : fix build · Updated 2024-03-27 16:31:52 +08:00    tqcq	7950 72		ZIP  TAR.GZ
ceb/wpm-portable-tolower 87a6088ffe · rename unicodedata.{cpp,h} to unicode-data.{cpp,h} · Updated 2024-03-26 22:52:33 +08:00    tqcq	7965 7		ZIP  TAR.GZ
ik/quantize_with_kv_overrides 9c5fd6be14 · minor : spacing · Updated 2024-03-26 20:09:02 +08:00    tqcq	7963 2		ZIP  TAR.GZ
ik/test_quantize_fns 6f20e2672f · Include IQ2_XXS and IQ2_XS in teet-quantize-fns · Updated 2024-03-26 01:01:20 +08:00    tqcq	7967 1		ZIP  TAR.GZ
sl/cuda-f16-fix3 210e469114 · cuda : fix LLAMA_CUDA_F16 build · Updated 2024-03-25 22:31:10 +08:00    tqcq	7969 1		ZIP  TAR.GZ
ceb/fix-win-unicode-fpaths d05c13b3b9 · llama : fix BPE LF token on MSVC · Updated 2024-03-24 02:03:16 +08:00    tqcq	7989 3		ZIP  TAR.GZ
gg/flash-attn-rebase 3a468e6f9f · llama : fix type of KQ_mask and KQ_pos · Updated 2024-03-22 23:12:17 +08:00    tqcq	7994 68		ZIP  TAR.GZ
ik/quantize_not_repeating 0e826d12a5 · quantize: be able to specify the token embedding tensor type · Updated 2024-03-22 22:27:34 +08:00    tqcq	8003 2		ZIP  TAR.GZ
gg/hf-args 8c3d5b5a79 · common : remove defaults · Updated 2024-03-22 21:33:24 +08:00    tqcq	7999 2		ZIP  TAR.GZ
patch-1 12aa74ba7d · minor : spacing · Updated 2024-03-22 21:24:57 +08:00    tqcq	8224 6		ZIP  TAR.GZ
gg/metal-dequant-align 072c56fcdb · metal : fix the fix · Updated 2024-03-22 15:58:22 +08:00    tqcq	8004 3		ZIP  TAR.GZ
ik/try_fix_rocm_k_cache a710d58d88 · Try fix quantized k-cache on ROCm · Updated 2024-03-22 02:18:50 +08:00    tqcq	8009 1		ZIP  TAR.GZ
ik/fix_k_cache_backend_tests 68e4fed4d9 · Now fix test-quantize-fns · Updated 2024-03-21 19:18:03 +08:00    tqcq	8016 3		ZIP  TAR.GZ
compilade/fix-server-tests-penalty 9a424a3872 · server : fix tests expecting old repeat penalty · Updated 2024-03-20 05:12:28 +08:00    tqcq	8032 1		ZIP  TAR.GZ
gg/repeng 0a9bc301ac · control-vectors : minor code style updates · Updated 2024-03-14 22:43:37 +08:00    tqcq	8071 3		ZIP  TAR.GZ
gg/metal-embed abf0afd0d6 · ci : fix iOS builds to use embedded library · Updated 2024-03-14 17:34:22 +08:00    tqcq	8089 4		ZIP  TAR.GZ
ik/try_fix_iq1s_sycl 9f805264dc · Attempt 2 · Updated 2024-03-13 00:40:13 +08:00    tqcq	8089 3		ZIP  TAR.GZ