Commit Graph

  • b7db9e7aac whisper : remove empty .gitmodules file [no ci] (#3085) Daniel Bevenius 2025-04-28 15:52:05 +02:00
  • f3c42399a3 talk-llama : sync llama.cpp (#3084) Georgi Gerganov 2025-04-28 16:40:23 +03:00
  • 28dcdff4c5 ci : disable publishing of java binding [no ci] (#3086) Daniel Bevenius 2025-04-28 15:38:52 +02:00
  • 50218b935d build : Add Moore Threads GPU support and update GitHub workflow for MUSA build (#3069) R0CKSTAR 2025-04-28 16:06:41 +08:00
  • f9b2dfdd8c examples : fix deprecated FFmpeg functions (#3073) Pedro 2025-04-28 01:16:50 -03:00
  • 50fda73f4c ruby : add encoder begin callback related methods (#3076) KITAITI Makoto 2025-04-26 04:33:11 +09:00
  • 1c20f46887 ci : enable bindings java job (#3070) Daniel Bevenius 2025-04-25 14:56:06 +02:00
  • adaea088bc ruby : add cmake option (#0) Georgi Gerganov 2025-04-24 20:38:43 +03:00
  • 6c0d843f9d cuda : fix unused variable compile warning (#0) Georgi Gerganov 2025-04-24 18:59:06 +03:00
  • efb800557f sync : ggml Georgi Gerganov 2025-04-24 18:41:48 +03:00
  • 337becefb9 opencl : remove obsolete files (skip) (ggml/1200) Georgi Gerganov 2025-04-24 18:41:17 +03:00
  • 11ae30c19e sync : ggml Georgi Gerganov 2025-04-24 18:41:36 +03:00
  • 88c3cecd43 opencl: split ggml-opencl.cl into multiple files and cleanup (llama/12886) lhez 2025-04-24 17:46:49 +03:00
  • fe4acb33e3 ggml : fix trailing whitespaces (llama/0) Georgi Gerganov 2025-04-24 17:22:27 +03:00
  • fd5a3e1bc6 CUDA: use switch statements in constexpr functions (llama/13095) Johannes Gäßler 2025-04-24 15:57:10 +02:00
  • 01e1600edd metal : fix floating-point range of attention scores in FA kernels (llama/13090) Georgi Gerganov 2025-04-24 10:38:30 +03:00
  • cf3eb291ab vulkan: matmul gcn tuning (llama/13016) Eve 2025-04-24 07:18:33 +00:00
  • 3d54b68ea7 CUDA: noncont MMVQ + batched bs1 MUL_MAT_ID (llama/13014) Johannes Gäßler 2025-04-22 21:27:40 +02:00
  • 11218294db ggml : add SSE 4.2 and x64 base variant for CPUs without AVX (llama/12871) Diego Devesa 2025-04-21 18:13:51 +02:00
  • 33c89ade7d SYCL: Add non-contiguous support in ROPE (llama/12993) Akarshan Biswas 2025-04-21 19:13:30 +05:30
  • 27a56e7243 vulkan: support noncontiguous rms_norm (llama/13031) Jeff Bolz 2025-04-20 03:50:02 -05:00
  • f4ca3e2f9c metal: add neg operator (llama/13029) Jeffrey Morgan 2025-04-19 22:28:40 -07:00
  • 0287a5c51b SYCL: Refactor and enable FP16 in binary broadcast OPs (llama/12975) Akarshan Biswas 2025-04-18 19:27:56 +05:30
  • 24d29c55df rpc : add RPC_CMD_HELLO (llama/12955) Radoslav Gerganov 2025-04-18 10:13:42 +03:00
  • 36019c35a3 graph : make FA compatible with MLA + add initial Metal kernels (llama/12953) Georgi Gerganov 2025-04-17 18:16:36 +03:00
  • 4e936e2afa ggml: Re-enable CUDA graphs in presence of CONT and DUP nodes (llama/12970) Alan Gray 2025-04-17 14:19:42 +01:00
  • 314ce5981e CANN: Add support for async operator submission (llama/12864) hipudding 2025-04-17 20:34:16 +08:00
  • cb7642b0f5 opencl: fix incorrect local_size index in profiling log (llama/12868) kimminsu 2025-04-17 06:25:57 +09:00
  • 7db8f278f0 vulkan: enable coopmat2 FA gqa and split_k optimizations more often (llama/12931) Jeff Bolz 2025-04-16 13:37:25 -05:00
  • be42a19eab CANN: Add 310P operator support check (llama/12962) Chenguang Li 2025-04-16 16:21:05 +08:00
  • b8755670ca metal : add FA-vec kernels for head size 96 (llama/12952) Georgi Gerganov 2025-04-15 14:45:05 +03:00
  • 483eecae62 CANN: Add x86 build ci (llama/12950) hipudding 2025-04-15 19:08:55 +08:00
  • 43e3d25d93 CUDA/HIP: Share the same unified memory allocation logic. (llama/12934) David Huang 2025-04-15 17:20:38 +08:00
  • e1dbf9a42e SYCL: Add ROPE vision kernel (llama/12887) Akarshan Biswas 2025-04-15 14:07:42 +05:30
  • ee0013865d ggml : Add AVX512 implementation of GEMM - Q4_Kx8 (llama/12829) Srihari-mcw 2025-04-15 11:52:36 +05:30
  • 32a407166b CANN: Opt ROPE optimization (llama/12865) Chenguang Li 2025-04-15 10:09:35 +08:00
  • 622f981853 CANN: Optimize CANN buffer pool memory management (llama/12875) Xinpeng Dou 2025-04-15 10:04:24 +08:00
  • d049d67065 SYCL: Fix im2col (llama/12910) Akarshan Biswas 2025-04-14 17:53:53 +05:30
  • 877308838e rpc : use ggml_context_ptr (llama/12938) Radoslav Gerganov 2025-04-14 13:59:34 +03:00
  • d87dfcf7c0 ggml : Depthwise 2D convolution (ggml/1152) Acly 2025-04-17 14:16:45 +02:00
  • 915c14ef10 ggml: use _mm[512/256]_dpbusd[_avx]_epi32 to directly accumulate into the result register (llama/12773) SXX 2025-04-14 13:47:55 +08:00
  • 5d33d3c929 ggml: disable CUDA graphs for unsupported DUP and CONT node types (llama/12891) Alan Gray 2025-04-13 22:12:21 +01:00
  • 751e42b21e vulkan: use aligned loads for flash attention mask (llama/12853) Jeff Bolz 2025-04-12 03:44:48 -05:00
  • e8ee32d12d sycl: Support sycl_ext_oneapi_limited_graph (llama/12873) Ewan Crawford 2025-04-11 15:32:14 +02:00
  • e9ce285135 SYCL: Add fp16 type support to unary op kernels (llama/12788) Akarshan Biswas 2025-04-11 13:33:50 +05:30
  • b942f451b6 ggml: fix compilation error s390x (llama/12848) Aaron Teo 2025-04-11 13:20:07 +08:00
  • e6410faf99 cpu: fix cpu backend's supports-op for GET_ROWS_BACK. fixes a fatal when running test-backend-ops with only the CPU backend (ggml/1190) cmdr2 2025-04-11 12:14:19 +05:30
  • 182df69384 CANN: Support more ops (llama/12841) Chenguang Li 2025-04-10 08:51:52 +08:00
  • 3bf9691dfd Fixes #12823 (llama/12830) Prajwal B Mehendarkar 2025-04-10 04:48:01 +05:30
  • ba444e9c23 ggml-cpu-impl.h: do not redefine bool on POWER9 (llama/12856) Piotr Kubaj 2025-04-09 23:00:34 +00:00
  • c6caf8eef2 ggml-impl.h: fix build on POWER9 (llama/12855) Piotr Kubaj 2025-04-09 23:00:25 +00:00
  • 6cae79a1d7 CANN: Support Opt CONV_TRANSPOSE_1D and ELU (llama/12786) Chenguang Li 2025-04-09 14:04:14 +08:00
  • b9bfe0c693 vulkan: In coopmat2 mmq, load q4_k/q5_k scales through shared memory (llama/12833) Jeff Bolz 2025-04-09 00:25:08 -05:00
  • 1d50c6ac22 vulkan: Use fp16 for the flash attention P*V multiplication (llama/12783) Jeff Bolz 2025-04-09 00:12:57 -05:00
  • 79f23d9132 cuda : add f32 to bf16 copy op (llama/12806) Sigbjørn Skjæret 2025-04-08 23:21:31 +02:00
  • ee2cbeeb74 llama : fix FA when KV cache is not used (i.e. embeddings) (llama/12825) Georgi Gerganov 2025-04-08 19:54:51 +03:00
  • 868a5ce310 ggml: don't include arm_neon.h when using CUDA 12 with ARM Neon (ggml/1187) cmdr2 2025-04-10 17:53:08 +05:30
  • b9c71fae5a ggml : add bilinear upscale support (ggml/1185) Diego Devesa 2025-04-09 12:32:13 +02:00
  • 6d67c6d93d ggml : add more generic custom op, remove deprecated custom ops (ggml/1183) Diego Devesa 2025-04-09 12:31:34 +02:00
  • 12cade118e Revert "sycl:remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor" (llama/12812) Neo Zhang Jianyu 2025-04-08 15:03:21 +08:00
  • fd1c725e65 opencl: better identify Adreno GPU (llama/12760) lhez 2025-04-07 13:22:54 -07:00
  • d33fd00cfe cuda : fix HIP and MUSA BF16 (llama/0) Georgi Gerganov 2025-04-07 13:18:07 +03:00
  • 3e0d89782a sycl: remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor (llama/12734) zhouwg 2025-04-07 23:22:57 +08:00
  • 7074b622eb CANN: fix typo in ggml-cann (llama/12733) zhouwg 2025-04-07 19:34:14 +08:00
  • b8d3e45342 CANN: Refactor to reduce duplicate code (llama/12731) hipudding 2025-04-07 17:10:36 +08:00
  • 1901505138 musa: fix compilation warnings in mp_22/31 (llama/12780) R0CKSTAR 2025-04-06 21:23:54 +08:00
  • 3c26dd3353 vulkan: fix NaN issue in flash attention shader (llama/12776) Jeff Bolz 2025-04-06 04:03:47 -05:00
  • d792d2a2dc vulkan: Use unclamped loads for flash attention mask (llama/12720) Jeff Bolz 2025-04-06 03:47:13 -05:00
  • 8add58aa5e Vulkan: Tune Vulkan mmq int dot shader for performance (llama/12767) 0cc4m 2025-04-05 18:04:03 +02:00
  • 8f8ede1b12 sycl: allow ggml-sycl configuration and compilation using Visual Studio project/solution (llama/12625) Nicolò Scipione 2025-04-04 16:00:46 +02:00
  • 3a6fe8d767 cmake: fix ggml-shaders-gen compiler paths containing spaces (llama/12747) Ronny Brendel 2025-04-04 15:12:40 +02:00
  • 76231bda56 vulkan: Hybrid waitForFences/getFenceStatus to reduce fence latency (llama/12630) Jeff Bolz 2025-04-04 00:54:35 -05:00
  • 785437c253 vulkan: set cmake minimum and project name in vulkan-shaders (llama/12744) Jeff Bolz 2025-04-04 00:53:20 -05:00
  • 2f0612cb1c CUDA: Prefer vector flash decoding kernel for Gemma models (llama/12738) Gaurav Garg 2025-04-03 21:50:29 +05:30
  • e944065d5b vulkan: Fix missing cmake logic for dot product extension (llama/12721) Jeff Bolz 2025-04-03 10:08:26 -05:00
  • ccc7b5df0b fix MUSA compiler warning (llama/12704) a3sh 2025-04-03 15:32:55 +08:00
  • fbed36851e CANN: Support operator SIN COS ARGMAX (llama/12709) Chenguang Li 2025-04-03 15:18:08 +08:00
  • d1d847f184 Simplify and improve CUDA graphs through use of indirect copy pointers (llama/9017) Alan Gray 2025-04-03 02:31:15 +01:00
  • 337f91d4a6 CANN: Fix failed test cases (llama/12708) hipudding 2025-04-03 08:49:51 +08:00
  • 317a0031f9 opencl: use max_alloc_size in backend ctx instead of querying again (llama/12705) lhez 2025-04-02 17:01:42 -07:00
  • b243416918 vulkan: Implement split_k for coopmat2 flash attention. (llama/12627) Jeff Bolz 2025-04-02 14:25:08 -05:00
  • 6e532c7187 cmake: remove caching from vulkan coopmat checks (llama/12719) bandoti 2025-04-02 14:56:26 -03:00
  • 2105b110d3 vulkan: Implement grouped query attention in the coopmat2 FA shader (llama/12559) Jeff Bolz 2025-04-02 12:40:32 -05:00
  • f82622180f Vulkan: Fix mmq int dot float cache size (llama/12722) 0cc4m 2025-04-02 19:12:30 +02:00
  • a71c64512a llama : add option to override model tensor buffers (llama/11397) Diego Devesa 2025-04-02 14:52:01 +02:00
  • 1e9c2f87f1 ggml : simplify Arm fp16 CPU logic (ggml/1177) Georgi Gerganov 2025-04-07 12:25:15 +03:00
  • 06ce8f83e6 CUDA: don't convert BF16 weights to FP32 (ggml/1174) Sigbjørn Skjæret 2025-04-04 21:05:12 +02:00
  • 8b92060a10 coreml : set convert_to="mlprogram" in convert Daniel Bevenius 2025-04-23 08:24:38 +02:00
  • 7858eddd10 ci : disable freeBSD job in build.yml (#3064) Daniel Bevenius 2025-04-22 11:07:54 +02:00
  • 3a88f1e504 examples : add HEAPU8 to exported runtime methods (#3062) Daniel Bevenius 2025-04-20 19:40:25 +02:00
  • f0d2bfbfb7 ruby : make Ruby bindings installed with build options (#3056) KITAITI Makoto 2025-04-17 18:49:58 +09:00
  • 170b2faf75 whisper : add no_context parameter to whisper_params (#3045) Sacha Arbonel 2025-04-16 06:24:38 +02:00
  • f8a3509b6d examples : add FFmpeg v7.0 support to ffmpeg-transcode.cpp (#3038) Fujimoto Seiji 2025-04-15 13:09:00 +09:00
  • 2a2d21c75d ruby: use CMake in build process (#3043) KITAITI Makoto 2025-04-14 18:18:27 +09:00
  • 9cfcd6cc45 docs : update README.md to note newer nvidia gpus (#3031) Jeff Klassen 2025-04-11 00:54:51 -06:00
  • e853620270 addon.node : support max_context api for addon.node (#3025) Lin Xiaodong 2025-04-11 12:36:38 +08:00
  • 549db9376f whisper : reduce delta_min from 1000ms to 100ms (#3028) Georgi Gerganov 2025-04-11 07:23:02 +03:00
  • 33a25e4dda docs : document how to use 'WHISPER_FFMPEG' build option (#3029) Fujimoto Seiji 2025-04-11 01:21:38 +09:00
  • becd0c888e whisper : reduce delta_min from 1000ms to 100ms gg/whisper-short-audio-check Georgi Gerganov 2025-04-10 12:25:29 +03:00
  • 43f5030aeb docs : fix README.md (#3024) Ekaitz Zárraga 2025-04-09 19:49:37 +02:00