Commit Graph

  • 6ac8e6b2ce cuda/vulkan: specify fp32-only support for some operations in supports_op (ggml/1129) cmdr2 2025-02-28 15:59:55 +05:30
  • 60d2ddebdf cuda/cpu: Increase support for fp16 unary operations (ggml/1125) cmdr2 2025-02-28 12:34:39 +05:30
  • 2e180184a8 Told cmake to install ggml-cpp.h as a public header file. (ggml/1126) petterreinholdtsen 2025-02-26 21:44:00 +01:00
  • ef40950c4a common : more general m_audio_len update logic (#2855) Ivy233 2025-03-07 16:10:03 +08:00
  • c774eec709 go : improve model download (#2756) Ryan Johnson 2025-03-07 02:03:51 -06:00
  • 5b481a27a6 common : fix audio loading by miniaudio (#2862) Dmitry Atamanov 2025-03-04 22:05:21 +05:00
  • fc7b1ee521 fix: missing include common-whisper (#2858) Lin Xiaodong 2025-03-03 02:55:11 +08:00
  • c42f67e2d2 ruby : follow audio library change (#2851) KITAITI Makoto 2025-02-28 15:09:02 +09:00
  • 339a1cba5d whisper : support GGML_BACKEND_DL (#2843) Diego Devesa 2025-02-27 13:35:07 +01:00
  • c64f3e8ada common : separate whisper sources (#2846) Georgi Gerganov 2025-02-27 12:50:32 +02:00
  • 9f83f67221 common : fix build min/max (#2845) Georgi Gerganov 2025-02-27 10:39:13 +02:00
  • 7d3da68f79 examples : use miniaudio for direct decoding flac, mp3, ogg and wav (#2759) Dmitry Atamanov 2025-02-27 12:06:54 +05:00
  • b5d21359c1 stream : stop on ^C when no audio is received (#2822) petterreinholdtsen 2025-02-27 07:59:51 +01:00
  • 17addf7104 sync : ggml Georgi Gerganov 2025-02-26 22:39:12 +02:00
  • cdaee8b4bd Support pure float16 add/sub/mul/div operations in the CUDA (and CPU) backend (ggml/1121) cmdr2 2025-02-25 18:06:34 +05:30
  • 4b60ff4f92 metal : copy kernels for quant to F32/F16 conversions (llama/12017) Gian-Carlo Pascutto 2025-02-25 10:27:58 +01:00
  • b43b9d928c opencl: fix for small models (llama/11950) lhez 2025-02-24 13:47:07 -08:00
  • e3cb412a59 Optimize mul_mat for Q4_0 on Intel GPU (llama/12035) Neo Zhang Jianyu 2025-02-24 22:33:23 +08:00
  • ac301a7d9b SYCL: Fix GGML_SYCL_DEBUG macro (llama/11995) Akarshan Biswas 2025-02-24 15:48:25 +05:30
  • 82e04e7670 ggml-cpu: Support s390x SIMD Instruction Set (llama/12019) Aaron Teo 2025-02-23 05:39:24 +08:00
  • 38ac47cd4d CUDA: app option to compile without FlashAttention (llama/12025) Johannes Gäßler 2025-02-22 20:44:34 +01:00
  • 2d70cd36d7 CUDA: optimize FA for GQA + large batches (llama/12014) Johannes Gäßler 2025-02-22 12:20:17 +01:00
  • 98dab49b9a cuda: Add Q5_1, Q5_0, Q4_1 and Q4_0 to F32 conversion support. (llama/12000) Gian-Carlo Pascutto 2025-02-22 09:43:24 +01:00
  • b1385e9aa9 CUDA: correct the lowest Maxwell supported by CUDA 12 (llama/11984) PureJourney 2025-02-21 19:21:05 +08:00
  • 48f5e893f5 MUSA: support ARM64 and enable dp4a .etc (llama/11843) Bodhi 2025-02-21 15:46:23 +08:00
  • dc21871fcb ggml-cpu: Add CPU backend support for KleidiAI library (llama/11390) Charles Xu 2025-02-20 14:06:51 +01:00
  • 64a430bc81 ggml: aarch64: implement SVE kernels for q3_K_q8_K vector dot (llama/11917) Prashant Vithule 2025-02-20 15:38:32 +05:30
  • 51a3580c79 CUDA: use async data loading for FlashAttention (llama/11894) Johannes Gäßler 2025-02-17 14:03:24 +01:00
  • 37a21dd43d vulkan: implement several ops relevant for ggml_opt (llama/11769) Rémy O 2025-02-17 07:55:57 +01:00
  • 8a22a8b17f vulkan: support multi/vision rope, and noncontiguous rope (llama/11902) Jeff Bolz 2025-02-16 01:52:23 -06:00
  • fcbcad0c90 metal : fix the crash caused by the lack of residency set support on Intel Macs. (llama/11904) Hale Chan 2025-02-16 14:50:26 +08:00
  • 4444db7360 metal : optimize dequant q6_K kernel (llama/11892) Adrian Kretz 2025-02-15 19:39:20 +01:00
  • a7fc1038ca repo : update links to new url (llama/11886) Georgi Gerganov 2025-02-15 16:40:57 +02:00
  • 1689aaf854 vulkan: initial support for IQ1_S and IQ1_M quantizations (llama/11528) Rémy O 2025-02-15 09:01:40 +01:00
  • 4b48fe449a opencl: Fix rope and softmax (llama/11833) lhez 2025-02-14 11:12:23 -08:00
  • 47cc043e69 cuda : add ampere to the list of default architectures (llama/11870) Diego Devesa 2025-02-14 15:33:52 +01:00
  • e3d9ffb98b ggml: optimize some vec dot functions for LoongArch ASX (llama/11842) Jinyang He 2025-02-14 16:54:27 +08:00
  • e22d69839d vulkan: linux builds + small subgroup size fixes (llama/11767) Eve 2025-02-14 02:59:40 +00:00
  • defe731263 llamafile: use member variable instead of constant for iq4nlt (llama/11780) Jeffrey Morgan 2025-02-13 09:05:04 -08:00
  • 4e07957bf9 musa: bump MUSA SDK version to rc3.1.1 (llama/11822) R0CKSTAR 2025-02-13 20:28:18 +08:00
  • d2c5154bb5 ggml-cpu : add chunking support to mul_mat_id (llama/11666) Diego Devesa 2025-02-13 01:02:38 +01:00
  • 4fac43fe00 ggml : x2 speed for WASM by optimizing SIMD (llama/11453) Xuan-Son Nguyen 2025-02-13 00:33:45 +01:00
  • 3be9670f17 HIP: Remove GCN from list of devices that avoid MMQ (llama/11831) uvos 2025-02-12 22:25:28 +01:00
  • 86729fcd6d HIP: Switch to std::vector in rocblas version check (llama/11820) uvos 2025-02-12 17:25:03 +01:00
  • 7fbca6304e cleanup: fix compile warnings associated with gnu_printf (llama/11811) bandoti 2025-02-12 10:06:53 -04:00
  • d597f83e1a ggml : fix multi-threaded clamp_f32 (llama/11824) Richard 2025-02-12 13:57:33 +00:00
  • e5edcc6259 ggml-cpu: Fix duplicate MATMUL_INT8 (llama/11817) Weizhao Ouyang 2025-02-12 20:22:58 +08:00
  • 556f773d53 CUDA: fix CUDART_VERSION checks (llama/11821) Johannes Gäßler 2025-02-12 13:16:39 +01:00
  • 91d02de332 Fix #11802: Compile bug - RegQueryValueExA changed to RegQueryValueEx (llama/11803) Sheldon Robinson 2025-02-11 10:55:45 -05:00
  • 1b67d72f87 CUDA: use arch list for compatibility check (llama/11775) Johannes Gäßler 2025-02-11 00:17:22 +01:00
  • 14d7c0368d fix: typos in documentation files (llama/11791) Maxim Evtush 2025-02-10 23:21:31 +01:00
  • db6e19188a vulkan: Make Vulkan optional at runtime (ggml/11493). (llama/11494) Danny Milosavljevic 2025-02-10 07:17:21 +01:00
  • b4b063a5c9 vulkan: add environment variable GGML_VK_PREFER_HOST_MEMORY to avoid VRAM allocation (llama/11592) Wagner Bruna 2025-02-10 03:08:22 -03:00
  • 930b739e7a vulkan: account for lookup tables when checking shared memory size (llama/11502) Jeff Bolz 2025-02-09 01:43:51 -06:00
  • 5981352bb5 ggml: Fix data race in ggml threadpool (llama/11736) Karol Kontny 2025-02-08 15:30:53 +01:00
  • 7561da244e CUDA: fix min. version for movmatrix (llama/11751) Johannes Gäßler 2025-02-08 10:46:07 +01:00
  • be83f342fb vulkan: print shared memory size (llama/11719) Jeff Bolz 2025-02-07 04:26:03 -06:00
  • fd369871f7 SYCL: remove XMX info from print devices (llama/11712) Akarshan Biswas 2025-02-07 14:57:53 +05:30
  • bbd8364f5e ggml : optimize and build warning fix for LoongArch (llama/11709) Jinyang He 2025-02-07 15:38:31 +08:00
  • e4102440ef SYCL: Adjust support condition for norm operators (llama/11674) Akarshan Biswas 2025-02-06 17:12:35 +05:30
  • f8242ec483 ggml : fix LoongArch compile error with 128-bit SIMD (llama/11701) junchao-zhao 2025-02-06 17:20:00 +08:00
  • ef51b4cba4 vulkan: optimize coopmat2 iq2/iq3 callbacks (llama/11521) Jeff Bolz 2025-02-06 00:15:30 -06:00
  • 6f08b24146 vulkan: initial support for IQ4_XS quantization (llama/11501) Rémy O 2025-02-06 07:09:59 +01:00
  • 7c165d7fa8 vulkan: use smaller combined allocations to avoid fragmentation (llama/11551) Jeff Bolz 2025-02-06 00:02:18 -06:00
  • 2f0cf44915 metal : avoid breaking build when metal API predates TARGET_OS_VISION (llama/11690) Charles Duffy 2025-02-05 19:52:31 -06:00
  • b9c972fd0d metal : adjust support conditions for norm operators (llama/11671) Georgi Gerganov 2025-02-05 10:57:42 +02:00
  • 01c9aafbfd CUDA: support for mat. mul. with ne03 != ne13 (llama/11656) Johannes Gäßler 2025-02-05 08:58:31 +01:00
  • bae6bbf487 CUDA: non-contiguous (RMS) norm support (llama/11659) Johannes Gäßler 2025-02-04 22:21:42 +01:00
  • c310272fa0 HIP: force max threads per block to be 1024 (llama/11621) fxzjshm 2025-02-05 02:18:38 +08:00
  • bd0b55dbe0 metal : use residency set for other platforms (llama/11648) Jhen-Jie Hong 2025-02-04 19:07:18 +08:00
  • ba4645db2c rpc: fix known RCE in rpc-server (ggml/1103) Patrick Peng 2025-02-06 09:29:13 -05:00
  • dfc6ca62f3 stream : add beam size parameter(#2836) masahji 2025-02-25 01:39:33 -08:00
  • 47e14c0529 whisper : restore big endian support (#2816) Thomas Fitzsimmons 2025-02-25 09:38:13 +00:00
  • 00ddb10fe2 select utf8 codepage on windows fix_vs_sdl2 Judd 2025-02-19 16:40:51 +08:00
  • b3a6018bbf fix building with MSVC + SDL2 Judd 2025-02-19 14:43:42 +08:00
  • d682e15090 Fixes for Windows (#2790) Judd 2025-02-06 15:37:21 +08:00
  • 46d07b9c85 cmake : fix compile assumptions for power9/etc (#2777) midnight 2025-02-05 04:41:10 -08:00
  • 33ea03f131 authors : update Georgi Gerganov 2025-02-04 13:03:40 +02:00
  • dbcc669e1a sync : ggml Georgi Gerganov 2025-02-04 13:03:09 +02:00
  • 16245b35e4 cmake: Add ability to pass in GGML_BUILD_NUMBER (ggml/1096) Christian Kastner 2025-02-04 00:17:15 +01:00
  • 898c0cb9d1 readme : add maintenance roadmap Georgi Gerganov 2025-02-04 10:50:10 +02:00
  • eb9e5032c4 ci : add stalebot Georgi Gerganov 2025-02-04 09:30:08 +02:00
  • cadfc50eab node : add max_len params in node addon (#2760) billyct 2025-02-04 04:49:06 +08:00
  • 3f91832352 talk-llama : sync llama.cpp Georgi Gerganov 2025-02-03 22:42:26 +02:00
  • cff8868b5f coreml : always convert to "neuralnetwork" (#2770) mgrachten 2025-02-03 21:36:32 +01:00
  • 90e3c5fc40 ci : more git Georgi Gerganov 2025-02-03 21:17:33 +02:00
  • e0f4cef867 ci : install git Georgi Gerganov 2025-02-03 20:12:37 +02:00
  • 234460987e ci : use ubuntu-22.04 instead of ubuntu-latest Georgi Gerganov 2025-02-03 19:50:24 +02:00
  • b8ab126343 cmake : sync cmake scripts Georgi Gerganov 2025-02-03 16:24:38 +02:00
  • edc5d9267c sync : ggml Georgi Gerganov 2025-02-03 16:05:34 +02:00
  • 344b98a44f scripts : fix sync paths Georgi Gerganov 2025-02-03 16:05:27 +02:00
  • dbeb7916b8 CUDA: fix Volta FlashAttention logic (llama/11615) Johannes Gäßler 2025-02-03 13:25:56 +01:00
  • fad2806352 HIP: fix flash_attn_stream_k_fixup warning (llama/11604) Johannes Gäßler 2025-02-02 23:48:29 +01:00
  • 9906792ec3 CUDA/HIP: add support for selectable warp size to mmv (llama/11519) uvos 2025-02-02 22:40:09 +01:00
  • c49ee07ff4 HIP: add GGML_CUDA_CC_IS_* for amd familys as increasing cc archtectures for amd gpus are not supersets of eatch other (llama/11601) uvos 2025-02-02 22:08:05 +01:00
  • f8a831779e CUDA: use mma PTX instructions for FlashAttention (llama/11583) Johannes Gäßler 2025-02-02 19:31:09 +01:00
  • 85451e3612 ci: use sccache on windows instead of ccache (llama/11545) Olivier Chafik 2025-01-31 17:12:40 +00:00
  • 43c744ce8b HIP: require at least HIP 5.5 uvos 2025-01-29 19:36:00 +01:00
  • fc2e44490d HIP: Prepare reduction operators for wave 64 uvos 2025-01-29 19:12:42 +01:00
  • f41fdad200 CUDA/HIP: add warp_size to cuda_device_info uvos 2025-01-29 17:46:23 +01:00