wylab/llama.cpp
1
0
Fork 1
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-07-16 17:35:58 +02:00
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • e1ab084803 mtmd : fix idefics3 preprocessing (#16806) b6858 Xuan-Son Nguyen 2025-10-27 23:12:16 +01:00
  • 5a4ff43e7d llama : disable pipeline parallelism if compute buffer allocation fails (#16748) b6857 Diego Devesa 2025-10-27 13:51:28 -07:00
  • 10640e31aa ggml : fix interpolate with align-corners and ne=1 (#16700) b6856 Acly 2025-10-27 21:50:22 +01:00
  • 80d28f104c HIP: fix AMDGPU_TARGETS, update documentation (#16803) b6855 Johannes Gäßler 2025-10-27 21:39:49 +01:00
  • c55d53acec model : add LightOnOCR-1B model (#16764) b6854 Xuan-Son Nguyen 2025-10-27 16:02:58 +01:00
  • 945501f5ea llama: fix leaked buffers for mmap + split files (#16765) b6853 Johannes Gäßler 2025-10-27 09:17:31 +01:00
  • 75cbdd3fce test-backend-ops: print failed tests at the end (#16785) b6852 Aman Gupta 2025-10-27 09:25:10 +08:00
  • 2b9bd9bf4e sycl: add ROLL operation support (#16665) b6851 tamarPal 2025-10-27 03:20:24 +02:00
  • 59fc1ec8e8 sycl: add REPEAT_BACK operation support (#16734) b6850 shani-f 2025-10-27 03:19:50 +02:00
  • 75d33b9302 CUDA: support for weight clamp in top-k norm (#16702) b6849 Aman Gupta 2025-10-27 09:06:16 +08:00
  • 3470a5c891 ggml-alloc : make gallocr prefer chunks that allow memory reuse (#16788) b6848 Acly 2025-10-26 23:19:03 +01:00
  • bd562fe4f7 cuda : use fast copy when src and dst are of different type and contiguous (#16789) b6847 Sigbjørn Skjæret 2025-10-26 21:31:41 +01:00
  • bbac6a26b2 ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs to support large batch (#16744) b6846 leejet 2025-10-27 02:13:31 +08:00
  • 73a48c9790 convert : enable expert group selection for all models with it (#16691) b6845 Sigbjørn Skjæret 2025-10-26 17:21:23 +01:00
  • f696428ce8 graph : add clamping to ffn_moe_weights_sum to avoid div-by-zero (#16655) b6844 Sigbjørn Skjæret 2025-10-26 17:20:32 +01:00
  • 7cce4f8158 model : set res->t_embd in SmallThinker models (#16782) b6843 Sigbjørn Skjæret 2025-10-26 16:08:52 +01:00
  • 8d8862829c docs : add Jamba to Text-only models list (#16778) amirai21 2025-10-26 14:01:20 +02:00
  • f77c13b91f CUDA: General GEMV fusion (#16715) b6841 Aman Gupta 2025-10-26 19:28:04 +08:00
  • 3cfa9c3f12 vulkan: deduplicate Microsoft Direct3D12 devices (#16689) b6840 Gilad S. 2025-10-26 06:37:38 +02:00
  • 5d195f17bc convert : handle mmproj filename/path properly (#16760) b6839 Galunid 2025-10-25 20:41:36 +02:00
  • 226f295f4d model : set res->t_embd in PLaMo2 models (#16766) b6838 Shunta Saito 2025-10-25 19:26:27 +09:00
  • f90b4a8efe vulkan: delete dead code (#16732) b6837 Giuseppe Scrivano 2025-10-25 10:59:54 +02:00
  • 8423d01931 vulkan: Optimize SSM_SCAN (#16645) b6836 Jeff Bolz 2025-10-25 00:04:12 -05:00
  • 5cca2542ac convert : avoid dequantizing mxfp4 for GPT-OSS (#16756) b6835 compilade 2025-10-24 20:52:00 -04:00
  • 55945d2ef5 ggml: fix CUDA grid launch condition for large block_nums.y in binbcast (#16742) b6834 leejet 2025-10-25 03:39:37 +08:00
  • 0bcb40b48c CUDA: use CUB for arbitary size argsort (#16754) b6833 Aman Gupta 2025-10-24 20:46:19 +08:00
  • 69e9ff0103 webui: support q URL parameter (#16728) Florian Badie 2025-10-24 14:10:29 +02:00
  • d7f794eadb convert : avoid dequantizing mxfp4 for GPT-OSS compilade/fix-prequant-mxfp4-gpt-oss Francis Couture-Harpin 2025-10-24 07:46:34 -04:00
  • 5a91109a5d model-conversion : add trust_remote_code for orig model run [no ci] (#16751) Daniel Bevenius 2025-10-24 12:02:02 +02:00
  • f8f071fadd convert : handle pre-quantized models (#14810) b6830 compilade 2025-10-23 16:31:41 -04:00
  • 0bf47a1dbb server: add memory breakdown print (#16740) b6829 Johannes Gäßler 2025-10-23 21:30:17 +02:00
  • 93fbd407f3 Merge branch 'master' into compilade/convert-prequant compilade/convert-prequant Francis Couture-Harpin 2025-10-23 14:23:12 -04:00
  • dd62dcfab9 convert : Make mistral-common dependency optional (#16738) Julien Denize 2025-10-23 15:54:46 +02:00
  • d0660f237a mtmd-cli : allow using --jinja (#16718) b6827 Xuan-Son Nguyen 2025-10-23 15:00:49 +02:00
  • fe6a9882ac Manually link -lbsd to resolve flock symbol on AIX (#16610) b6826 Prajwal B Mehendarkar 2025-10-23 17:07:31 +05:30
  • 061f0eff02 ggml-cuda: use passed ops instead of hardcoded ops (#16712) b6825 Aman Gupta 2025-10-23 19:14:06 +08:00
  • 8cf6b42d46 server : send partial stop string when <EOG> is reached (#15007) b6824 matteo 2025-10-23 11:32:24 +02:00
  • 9de9672adb sycl: use async memory allocation to fix crashes during graph recording (#16644) b6823 Matthew Michel 2025-10-22 20:05:15 -05:00
  • 63d2fc46e1 Add experimental ggml-hexagon backend for the Hexagon NPU (#16547) b6822 Max Krasnyansky 2025-10-22 13:47:09 -07:00
  • a2e0088d92 Revert "ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_v…" (#16723) b6821 Diego Devesa 2025-10-22 11:20:55 -07:00
  • 9b9201f65a webui: introduce OpenAI-compatible model selector in JSON payload (#16562) Pascal 2025-10-22 16:58:23 +02:00
  • 19a5a3edfd ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_vec_set_f32 for faster fills (#16522) sirus20x6 2025-10-22 05:14:14 -05:00
  • d8eaa26e4d tests : fix test-thread-safety when compiling with multiple backends (#16699) b6818 Acly 2025-10-22 12:01:22 +02:00
  • 9285325ce0 CUDA: fix bug in topk-moe softmax (#16711) b6817 Aman Gupta 2025-10-22 12:33:08 +08:00
  • 03792ad936 CUDA: topk-moe: add optional parameter for gpt-oss (#16649) b6816 Aman Gupta 2025-10-21 22:40:38 +08:00
  • 51d1a8c997 CUDA: better error for FA kernel with 0 occupancy (#16643) b6815 Johannes Gäßler 2025-10-21 15:27:53 +02:00
  • 4926419c4d ggml: add ggml_can_fuse_subgraph (#16662) b6814 Aman Gupta 2025-10-21 16:43:14 +08:00
  • 6ea37f5739 opencl: fix warnings and clean up profiling (#16688) b6813 lhez 2025-10-20 22:26:17 -07:00
  • fb349848f3 vulkan: Handle FA with all -inf mask values (#16447) b6812 Jeff Bolz 2025-10-20 22:16:08 -05:00
  • 6de8ed7519 sycl : add PAD_REFLECT_D1 operator support (#16145) b6811 YehuditE 2025-10-21 01:21:12 +03:00
  • 84bf3c6778 model : add BailingMoeV2 support (#16063) b6810 Sigbjørn Skjæret 2025-10-20 21:38:20 +02:00
  • c9c1972e2c Handle legacy 'context' attachments (#16687) Aleksander Grygier 2025-10-20 19:49:02 +02:00
  • b617cfd289 ggml-alloc : fix leak when reusing a tensor with a larger size (#16679) b6808 Diego Devesa 2025-10-20 05:53:50 -07:00
  • 79068501fa Prevent premature submission on IME input (#16673) Aleksander Grygier 2025-10-20 14:21:12 +02:00
  • 0e4a0cf2fa Import/Export UX improvements (#16619) Aleksander Grygier 2025-10-20 13:29:14 +02:00
  • 13f2cfad41 Enable per-conversation loading states to allow having parallel conversations (#16327) Aleksander Grygier 2025-10-20 12:41:13 +02:00
  • 06332e2867 llama-batch: fix build fails with -Werror=missing-braces (#16614) b6804 takuya kodama 2025-10-20 16:27:09 +08:00
  • 72d53e6c4d readme: update bindings (#16651) Ron Evans 2025-10-20 10:20:04 +02:00
  • 2330de7b84 SYCL: Add support for FLOOR,CEIL,ROUND and TRUNC unary operators (#16613) b6802 safranowith 2025-10-20 11:08:32 +03:00
  • 7062dd8460 llama-context: only warn on pooling_type when user specified (#16674) b6801 takuya kodama 2025-10-20 15:44:21 +08:00
  • 0398752dd4 model : add Granite Hybrid types (#16635) b6800 Giuseppe Scrivano 2025-10-19 23:54:31 +02:00
  • 4f73d0a951 ci : fix binaries release failure for s390x (binaries may not work yet) (#16664) b6799 Aaron Teo 2025-10-20 05:06:39 +08:00
  • f0076dc5a0 metal : adjust .get_alloc_size to be alloc friendly gg/metal-alloc-size Georgi Gerganov 2025-10-19 17:20:54 +03:00
  • cec5edbcae ci : avoid manual updates of docs/ops.md (#16663) Sigbjørn Skjæret 2025-10-19 14:03:25 +02:00
  • fcb235b466 ci: include s390x release binaries (#16648) Aaron Teo 2025-10-19 18:37:47 +08:00
  • 55754bebd5 CODEOWNERS: update for ggml-cuda/mmf (#16660) Aman Gupta 2025-10-19 15:37:12 +08:00
  • ee09828cb0 HIP: fix GPU_TARGETS (#16642) b6795 Johannes Gäßler 2025-10-18 14:47:32 +02:00
  • e56abd2098 vulkan: Implement topk_moe fused shader, ported from CUDA (#16641) b6794 Jeff Bolz 2025-10-18 05:22:57 -05:00
  • 38355c6c8e CUDA: use registers instead of smem in topk-moe (#16647) b6793 Aman Gupta 2025-10-18 17:52:53 +08:00
  • 81387858f1 opencl: transposed gemm/gemv moe kernel with mxfp4,f32 (#16602) b6792 Shawn Gu 2025-10-17 17:55:32 -07:00
  • 66b0dbcb2d llama-model: fix insonsistent ctxs <-> bufs order (#16581) b6791 Johannes Gäßler 2025-10-17 17:41:09 +02:00
  • 41386cf365 rpc : report actual free memory (#16616) b6790 Radoslav Gerganov 2025-10-17 18:02:52 +03:00
  • 3d4e86bbeb vulkan: Add State Space Model (SSM) Operations Support (#16463) b6789 Giuseppe Scrivano 2025-10-17 14:23:47 +02:00
  • 342c728d03 ggml : fix SpaceMit IME array out-of-bounds in task assignment (#16629) b6788 muggle-stack 2025-10-17 18:01:23 +08:00
  • ababae7e1e webui: reorganize settings layout (#16607) Pascal 2025-10-17 10:35:03 +02:00
  • b19491599d vulkan: fix debug build (add_rms_len/data not found) (#16624) b6786 Jeff Bolz 2025-10-17 02:31:04 -05:00
  • 9ad4f1931e metal : add CONV_TRANSPOSE_2D (#16542) b6785 Ilia Ilmer 2025-10-17 02:33:58 -04:00
  • 79967ec596 grammar : use int64_t to avoid int overflows in int schema to grammar conversion logic (#16626) b6784 Olivier Chafik 2025-10-17 06:59:31 +01:00
  • ceff6bb253 SYCL SET operator optimized for F32 tensors (#16350) b6783 GittyBurstein 2025-10-17 05:36:40 +03:00
  • 1bb4f43380 mtmd : support home-cooked Mistral Small Omni (#14928) b6782 Xuan-Son Nguyen 2025-10-16 19:00:31 +02:00
  • 683fa6ba4e fix: added a normalization step for MathJax-style \[\] and \(\) delimiters (#16599) Pascal 2025-10-16 16:28:41 +02:00
  • b22572e97d sycl : add ARANGE operator (#16362) b6780 GittyBurstein 2025-10-16 16:26:21 +03:00
  • 7a50cf388a CANN: format code using .clang-format (#15863) b6779 Chenguang Li 2025-10-16 16:41:11 +08:00
  • 6f5d924637 common : Update the docs on -t --threads (#16236) b6778 takasurazeem 2025-10-16 01:11:33 -04:00
  • adc9b60f19 ggml-cpu: replace putenv with setenv for const-correctness (#16573) b6777 takuya kodama 2025-10-16 13:10:32 +08:00
  • ee50ee1ead SYCL: Add GGML_OP_MEAN operator support (#16009) b6776 yael-works 2025-10-16 07:21:28 +03:00
  • 7adc79c032 gguf-py : add support for endian conversion of BF16 data (#16594) b6775 Aleksei Nikiforov 2025-10-15 22:43:08 +02:00
  • 466c1911ab cpu : add FLOOR, CEIL, ROUND and TRUNC unary operators (#16083) b6774 safranowith 2025-10-15 22:24:51 +03:00
  • 0cb7a0683b opencl: add q8_0 mm support (#16469) b6773 lhez 2025-10-15 10:51:04 -07:00
  • d93f8439b0 opencl: fix FA for f32 (#16584) lhez 2025-10-15 10:48:28 -07:00
  • f9fb33f263 Add server-driven parameter defaults and syncing (#16515) Aleksander Grygier 2025-10-15 16:22:20 +02:00
  • f4ce81c45e metal: optimise GGML_OP_SUM (#16559) b6770 Sam/Samuel 2025-10-15 23:05:56 +09:00
  • 17304cbcc1 server : fix img token logs (#16595) b6769 Georgi Gerganov 2025-10-15 16:53:12 +03:00
  • 3e3cb19f64 llama-quant: add support for mmproj (#16592) b6768 Xuan-Son Nguyen 2025-10-15 14:48:08 +02:00
  • 5acd455460 CUDA: Changing the CUDA scheduling strategy to spin (#16585) b6767 Julius Tischbein 2025-10-15 13:54:15 +02:00
  • 554fd578a5 server : fix mtmd checkpoints (#16591) b6766 Georgi Gerganov 2025-10-15 12:51:27 +03:00
  • fa882fd2b1 metal : avoid using Metal's gpuAddress property (#16576) b6765 Georgi Gerganov 2025-10-14 20:33:05 +03:00
  • ffa059034c vulkan: Add ACC_TYPE_VEC2 implementation (#16203) b6764 SavicStefan 2025-10-14 19:18:05 +02:00
  • 120bf7046d CUDA + openCL: fix bug in accessing rms_norm->src while doing fusion (#16577) b6763 Aman Gupta 2025-10-14 22:48:08 +08:00
  • 4258e0cfe7 vulkan: Support FA with K/V in F32 (#16543) b6762 Jeff Bolz 2025-10-14 08:53:37 -05:00