Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-08-25 22:21:14 +02:00
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • 7b38cb71b9 Fixed gating logic for problematic Intel driver version Masato Nakasaka 2026-08-14 10:39:51 +09:00
  • 9d57ce456c mtmd: fix Granite4 Vision image sequence assembly (#26653) Hemanth Battu 2026-08-14 15:25:26 -07:00
  • 16d222fc5e model : add support for MiniMaxText01ForCausalLM and MiniMaxM1ForCausalLM (#27018) fairydreaming 2026-08-15 00:02:38 +02:00
  • 6fed9f6ff7 mtmd, common: various fixes (#27071) Xuan-Son Nguyen 2026-08-14 23:34:56 +02:00
  • 9e40df63ba jinja : fix quadratic cost in gather_string_parts (#27034) 0 2026-08-14 17:34:40 -04:00
  • 7e4c0a9688 chat : pass reasoning_effort to template Andy Williams 2026-08-15 03:53:11 +09:30
  • 9b05354ec6 sync : ggml Georgi Gerganov 2026-08-14 19:05:40 +03:00
  • 06ae2326ba ggml : bump version to 0.20.0 (ggml/1584) Georgi Gerganov 2026-08-14 18:13:36 +03:00
  • 531cac9cd3 fixed dflash Concedo 2026-08-14 22:38:55 +08:00
  • 22806c84d3 fixed prints Concedo 2026-08-14 22:27:12 +08:00
  • 1692f9e50b ggml : recurrent state rollback for ggml_ssm_scan (#26623) lnigam 2026-08-14 19:50:40 +05:30
  • d81ae290d2 move a print to debug Concedo 2026-08-14 21:22:42 +08:00
  • 4c1a0af40d llama : allow virtual igpu devices (#26953) Georgi Gerganov 2026-08-14 15:14:19 +03:00
  • 77918caf30 server: allow accessing /metrics and /slots during llama_decode() (#27041) Xuan-Son Nguyen 2026-08-14 13:23:10 +02:00
  • 886a42d446 Merge branch 'upstream' into concedo_experimental Concedo 2026-08-14 18:10:18 +08:00
  • 885c5bbe8e tests : replace personal home directory paths with generic placeholders (#27043) Jim Wu 2026-08-14 01:32:59 -07:00
  • 6509138622 sycl: fuse mul_mat(gate) + mul_mat(up) + GLU for q4_K dense FFN (#26779) Titaniumtown 2026-08-13 23:26:23 -07:00
  • c6f6a92c55 ggml: force single thread on wasi (#25686) Mendy Berger 2026-08-14 02:16:20 -04:00
  • 3d93885352 sycl: fuse the gated-delta-net state writeback cpy (#26643) Titaniumtown 2026-08-13 23:00:36 -07:00
  • da91226fbc updated sdui Concedo 2026-08-14 13:54:20 +08:00
  • 07394b0096 build: group objects common to all koboldcpp_* libraries (#2397) Wagner Bruna 2026-08-14 02:31:56 -03:00
  • 2bacf9ea5c dflash : clarify output logging of target_layer_ids (#27013) Daniel Bevenius 2026-08-14 06:57:05 +02:00
  • a94d563ed8 common: apply CPU parameters across tools (#27026) Niklas Wenzel 2026-08-13 20:40:59 +02:00
  • bdffafa5df ui: Refactor data-attrs constants, enum for bool strings (#27002) Aleksander Grygier 2026-08-13 20:01:12 +02:00
  • fa4ec4590c refactor: Naming (#27001) Aleksander Grygier 2026-08-13 19:45:32 +02:00
  • 9c5531e2bf ui: fix VITE_PUBLIC_SERVER variable reading (#24845) Ilya 2026-08-13 20:31:51 +03:00
  • aee56b3abf OpenVINO: Qwen3.5, memory optimization, and test-recurrent-state-rollback (#26952) Zijun Yu 2026-08-14 01:30:48 +08:00
  • a97123e497 [SYCL] Support host pinned mem to improve SYCL Host-to-Device Memory Access (#26789) Neo Zhang 2026-08-14 01:05:33 +08:00
  • 2606220d9f chat : fix LFM2 tool call arg name prefix ambiguity (#26960) Aldehir Rojas 2026-08-13 11:18:44 -05:00
  • 981184e49a server : serve index.html with no-cache (#27006) Emanuil Rusev 2026-08-13 17:59:45 +03:00
  • 1d2869c6e5 spec : auto-detect mtp draft model type (#27005) Sigbjørn Skjæret 2026-08-13 13:39:16 +02:00
  • 4a84b0ad10 metal : add TQ2_0 support (#26980) Georgi Gerganov 2026-08-13 14:33:53 +03:00
  • f65e568fd8 common : auto-detect spec type from draft GGUF metadata (#26814) aic0d3r 2026-08-13 12:34:27 +02:00
  • 0d0bfcd4fd spec: enable backend sampling for both dflash & dspark (#26958) Ruixiang Wang 2026-08-13 11:07:28 +02:00
  • eeae28b67e ggml-cpu/ops: vectorize flash-attention V-cache F16 to F32 conversion (#26947) jinzihao 2026-08-13 17:04:46 +08:00
  • 154d57af3e sycl: remove separate fp32 type promotion in gemm non-oneDNN path (#26372) Ian Faust 2026-08-13 11:02:18 +02:00
  • 1ee1cd9bc6 sycl: fuse UNARY(silu|sigmoid|softplus) + MUL (#26411) Titaniumtown 2026-08-13 01:41:38 -07:00
  • 8efbf65dbd sycl : Add DMMV ESIMD Q3_K kernel (#26251) Todd Malsbary 2026-08-13 01:32:36 -07:00
  • d415e65a57 sycl : enhance concat to support Q4_0, Q4_1, Q5_0, Q5_1, Q8_0 (#26800) Neo Zhang 2026-08-13 16:24:28 +08:00
  • decaf508bb server: refactor + correctness fixes for metrics (#26920) Xuan-Son Nguyen 2026-08-13 10:02:01 +02:00
  • e79e4bf660 ggml-hip : remove -funsafe-math-optimizations (#26696) Jim Wu 2026-08-12 23:38:02 -07:00
  • d86c7d62df ui: Clean up contexts, remove prop drilling from Chat Form Actions (#26951) Aleksander Grygier 2026-08-13 08:21:15 +02:00
  • f2efd64141 ui: Move styles/ to $lib scope (#26950) Aleksander Grygier 2026-08-13 08:13:52 +02:00
  • 094e53db1c ui: Stores architecture improvements (#26910) Aleksander Grygier 2026-08-13 08:11:30 +02:00
  • a6040c925c refactor: Clean up UI types (#26909) Aleksander Grygier 2026-08-13 08:07:34 +02:00
  • 1f368f354d ggml : fix arm builds, unused var (#26991) Georgi Gerganov 2026-08-13 07:57:24 +03:00
  • e21152dc96 ui: Constants refactor (#26908) Aleksander Grygier 2026-08-13 06:53:56 +02:00
  • 8e7f22b67e common: add system-level config file (#26118) Johnathan Craig Maudlin 2026-08-12 18:02:27 -04:00
  • 84e908c625 ci: fix thread sanitizer + remove ccache (#26927) Eve 2026-08-12 16:01:12 +00:00
  • 9558fa44c9 ci : disable ubuntu-rocm (#26969) Sigbjørn Skjæret 2026-08-12 15:41:44 +02:00
  • 7a9ff95979 disable rocm cache (#26962) Sigbjørn Skjæret 2026-08-12 15:41:43 +02:00
  • 7ec78d48e4 sd: sync with master-816-487de75 (#2395) Wagner Bruna 2026-08-12 10:35:09 -03:00
  • dfab7c1bf0 stop seq fix Concedo 2026-08-12 21:34:31 +08:00
  • f556b13a4e Merge branch 'upstream' into concedo_experimental Concedo 2026-08-12 21:34:23 +08:00
  • 680a9ae63d cmake : introduce semantic versioning (#26839) Daniel Bevenius 2026-08-12 14:15:03 +02:00
  • d8a8beac22 gguf : harden loader against malformed tensor dims and metadata types (#25596) HarrisonSec 2026-08-12 05:07:48 -07:00
  • 132753bf4e kleidiai: Add runtime feature detection mechanism for aarch64/kleidiai (#26076) Jonathan Clohessy 2026-08-12 12:49:11 +01:00
  • ece98b87f7 model : disallow integer dflash sliding_window_pattern (#26900) Sigbjørn Skjæret 2026-08-12 13:24:10 +02:00
  • af05a42a7c sync : ggml Georgi Gerganov 2026-08-12 13:59:44 +03:00
  • 13fd0bb55e cmake : add config version support (ggml/1582) Daniel Bevenius 2026-08-12 12:46:06 +02:00
  • 5d9e5ac30e server : support slot save/restore with media inputs (#26640) Chipmunk 2026-08-12 19:20:28 +09:00
  • ddd2691551 fix muse glimmer assistant tag Concedo 2026-08-12 18:13:57 +08:00
  • 75c8184210 hold the mutex for the whole step_callback function Concedo 2026-08-12 18:05:05 +08:00
  • 4dd127584b ui: add read_media tool (#25877) parabelboi 2026-08-12 12:03:32 +02:00
  • 533aa18897 make preview sticky for sdcpp Concedo 2026-08-12 17:13:20 +08:00
  • 89e0aa6fd3 opencl: default FA c8 cluster width to 16 on X1E (#26433) Hongqiang Wang 2026-08-11 23:10:27 -07:00
  • a4a4c51f3d tests : update speculative params (#26925) Georgi Gerganov 2026-08-12 08:08:19 +03:00
  • a7cd2f0e98 vulkan: add TQ2_0 (ternary) support (#25850) michaeltrabalka-tech 2026-08-12 01:07:23 -04:00
  • 55f453b924 wavtokenizer-dec : bound posnet/convnext block_count against n_layer_all (#26892) Oğuzhan Akkaya 2026-08-12 01:06:16 -04:00
  • 6eff593262 convert : handle per_layer_config in Gemma4 (transformers 5.15) (#26882) Wang Zhiyu 2026-08-12 13:05:13 +08:00
  • cb27fe9c35 opencl: use flat mv q5_k when weight exceeds image1d_buffer_t limit (#26880) lhez 2026-08-11 22:02:28 -07:00
  • 0b1bad14ff chat : fix muse-glimmer detection of tool calls after EOM (#26879) ruanslv 2026-08-11 16:15:20 -04:00
  • 7b13a8404d ci : add missing release check (#26923) Sigbjørn Skjæret 2026-08-11 20:20:40 +02:00
  • ebb546b7e9 CUDA: only disable CUDA graphs when mul_mat_id actually needs a stream sync (#26802) Rafail Giavrimis 2026-08-11 18:50:03 +01:00
  • 5988633170 cuda : add warp-per-row wkv7 kernel for single-token decode (#26111) 0 2026-08-12 01:46:23 +08:00
  • f785fc9ea4 spec : update speculative-simple (#26904) Georgi Gerganov 2026-08-11 19:52:12 +03:00
  • ba360efe1f chat : tighten bare function parsing for Qwen models (#26793) Aldehir Rojas 2026-08-11 10:58:54 -05:00
  • 70dfba5aee ci : add windows-rocm to check-release (#26897) Sigbjørn Skjæret 2026-08-11 17:48:27 +02:00
  • 38406d597f imatrix.cpp: Move finite check and only check touched experts (#26861) Bartowski 2026-08-11 11:18:19 -04:00
  • 2468576f24 requirements: use stable torch packages on s390x (#26864) Niklas Wenzel 2026-08-11 15:58:53 +02:00
  • 5925082d19 fix noisy callback no longer encodes a preview, and preview generation triggers only one-shot Concedo 2026-08-11 21:21:50 +08:00
  • 5d16e81dd9 convert : keep quantization scales for nemotron --mtp export (#26903) ynankani 2026-08-11 13:19:05 +00:00
  • cc078b45b6 Dflash support for nemotron-3.5 (#26905) lnigam 2026-08-11 18:46:26 +05:30
  • c98c00f9ba sd: generation progress fixes (#2391) Wagner Bruna 2026-08-11 10:03:11 -03:00
  • 6e62ba5384 mtmd: support pocket-tts (#26871) Xuan-Son Nguyen 2026-08-11 14:18:30 +02:00
  • 8d274dd7c6 ui: fix context gauge for single-model usage (#25738) Tom Tan 2026-08-11 04:42:48 -07:00
  • d135a116da dflash works with muse glimmer now Concedo 2026-08-11 18:25:25 +08:00
  • db8b62dc9b fix more debug printing Concedo 2026-08-11 18:18:09 +08:00
  • 6473847b60 wip dflash support Concedo 2026-08-11 18:16:33 +08:00
  • 7eb3a3405a fixed some stdout prints Concedo 2026-08-11 18:11:51 +08:00
  • 92c526d0e8 fix mtp Concedo 2026-08-11 17:59:16 +08:00
  • 704485942a ci: hip-quality-check: update vgpr spill ignore list (#26859) uvos 2026-08-11 11:47:57 +02:00
  • 1138b851fa model-conversion : use save_output_data for causual embeddings [no ci] (#26890) Daniel Bevenius 2026-08-11 11:41:38 +02:00
  • 9afff1b748 tests : fix running server tests on windows (#26889) Georgi Gerganov 2026-08-11 12:07:15 +03:00
  • 7fd4acc35c reasoning budget for muse glimmer Concedo 2026-08-11 15:43:23 +08:00
  • dcfe29317e Merge branch 'upstream' into concedo_experimental Concedo 2026-08-11 15:13:15 +08:00
  • b39ff27d6f muse glimmer jinja and tool calls working Concedo 2026-08-11 15:06:30 +08:00
  • 153d324bcf llama: add default load-mode auto, which avoids mmap on iGPUs (#26081) Ruben Ortlam 2026-08-11 08:20:46 +02:00
  • b3df57286c tests : clean-up server test, use tests.sh in ci (#26886) Georgi Gerganov 2026-08-11 09:07:13 +03:00
  • 4801e3c567 tests : disable backend sampler hip multi output (#26878) Jim Wu 2026-08-10 21:21:32 -07:00