]> git.djapps.eu Git - pkg/ggml/sources/ggml/shortlog
pkg/ggml/sources/ggml
2025-03-27 Guus Waalsggml-vulkan: remove unused find_program(glslc) (llama...
2025-03-27 Jeff Bolzvulkan: Add N/2 and N/4 optimized paths in coopmat2...
2025-03-27 Danielevulkan: subgroup size tuning (llama/12087)
2025-03-27 Jeff Bolzvulkan: use fp32 in coopmat2 q4_k dequant function...
2025-03-27 Jeff Bolzvulkan: Pad N dimension of B matrix for coopmat2 perf...
2025-03-27 Jeff Bolzvulkan: Adjust coopmat2 tile sizes and selection heuris...
2025-03-27 Christian Kastnercmake : enable building llama.cpp using system libggml...
2025-03-27 Akarshan BiswasSYCL: set extras only on GGML_TYPE_Q4_0 (llama/12366)
2025-03-27 aubreyliSYCL: Delete redundant plus sign and space (llama/12391)
2025-03-27 fairydreamingSYCL : support non-contiguous tensors in binary ops...
2025-03-27 Chenguang LiMUL_MAT optimization (llama/12382)
2025-03-27 Alberto Cabrera... sycl : variable sg_size support for mmvq kernels (llama...
2025-03-27 uvosCUDA/HIP: Fix fattn-vec-* when device warp size is...
2025-03-27 Jeff Bolzvulkan: fix bug in coopmat1 mul_mat_id (llama/12316)
2025-03-27 uvosCUDA/HIP: refractor mmqv to unify the calculation of...
2025-03-27 jklincnggml-backend : fix backend search path (llama/12330)
2025-03-27 BB-fatmetal : Cache the Metal library at the device context...
2025-03-27 Evemat vec double buffer (llama/12188)
2025-03-27 R0CKSTARmusa: support new arch mp_31 and update doc (llama...
2025-03-27 Henry Linjamäkiopencl: use OpenCL C standard supported by the device...
2025-03-27 Georgi Gerganovtests : fix test-quantize-fns to init the CPU backend...
2025-03-27 Jason C.Hggml-backend : make path_str compatible with C++20...
2025-03-27 Daniel Beveniusggml : skip intermediate .air file when compiling ...
2025-03-26 Akarshan Biswasci: disable test-opt for now (#1158)
2025-03-25 Akarshan Biswasci: Initial SYCL setup (#1157)
2025-03-13 cmdr2Create CONTRIBUTING.md (#1146)
2025-03-13 bssrdfgpt-2 : add comment about KV cache type (#1142)
2025-03-10 Christian Kastnercmake: Enable specifying exact PowerPC CPU architecture...
2025-03-10 Christian Kastnercmake: Comment out GGML_BIN_DIR for now (#1139)
2025-03-08 Georgi Gerganovspm : remove (#1135)
2025-03-08 Georgi Gerganovsync : whisper.cpp upstream/0.0.1802
2025-03-08 Dmitry Atamanovcommon : fix audio loading by miniaudio (whisper/2862)
2025-03-07 Georgi Gerganovsync : llama.cpp
2025-03-07 Rémy Oggml-cpu: faster AVX2 variant for IQ1_M (llama/12216)
2025-03-07 BB-fatmetal : simplify kernel arguments using a struct (...
2025-03-07 Daniel Beveniusmetal : fix default.metallib build (llama/12224)
2025-03-07 lhezopencl: Noncontiguous `norm`, `rms_norm`, disable ...
2025-03-07 xiaofeicmake : fix undefined reference errors for std::filesys...
2025-03-07 Johannes GäßlerCUDA: fix FA logic for PTX 7.0 and CC >= 7.5 (llama...
2025-03-07 uvosHIP/CUDA: set the paramerter value in maintain_cuda_gra...
2025-03-07 Henry Linjamäkiopencl : fix buffer alignment (llama/12197)
2025-03-07 Henry Linjamäkiopencl : fix `ulong` kernel args were set from `int...
2025-03-07 simon886212opencl : fix profile-related errors (llama/12095)
2025-03-07 Rémy Oggml-cpu: Faster IQ1 mul_mat_vec on AVX2 using BMI2...
2025-03-07 Akarshan BiswasSYCL: Disable f16 Unary OPs as not supported by the...
2025-03-07 Plamen Minevggml : fix GGMLMetalClass ODR (llama/12200)
2025-03-07 vmobilisggml : ggml_compute_forward_concat() for arbitrary...
2025-03-06 Christian Kastnerpkg-config: Use CMake install paths for lib, include...
2025-03-04 Georgi Gerganovvulkan : sync (llama/0)
2025-03-04 Georgi Gerganovsync : llama.cpp
2025-03-04 mgroeber9110ggml : portability fixes for VS 2017 (llama/12150)
2025-03-04 David HuangHIP: implement FlashAttention via rocWMMA for CDNA...
2025-03-04 Diego Devesatest-backend-ops : add option -p to filter by op params...
2025-03-04 ag2s20150909ggml : fix kleidiai build (llama/12159)
2025-03-04 Akarshan BiswasSYCL: Move CPY kernels to a separate file and add few...
2025-03-04 Diego Devesaggml-backend : keep paths in native string type when...
2025-03-04 Erik ScholzCUDA: compress mode option and default to size (llama...
2025-03-04 William Tambelliniggml : upgrade init_tensor API to return a ggml_status...
2025-03-04 Rémy Ovulkan: add specific MMV kernels for IQ2 and IQ3 quants...
2025-03-04 Johannes GäßlerCUDA: fix logic for V100 + GGML_CUDA_FORCE_MMQ (llama...
2025-03-04 Prashant Vithuleggml: aarch64: implement SVE kernels for q2_k_q8_k...
2025-03-04 hipuddingCANN: Fix build error with GCC 13 (llama/11990)
2025-03-04 Evevulkan: matmul dequantization improvements (llama/12015)
2025-03-04 Danielevulkan: improve im2col (llama/11826)
2025-03-04 Vladimir Vuksanoviccmake: Fix ggml backend dependencies and installation...
2025-03-04 Jeff Bolzvulkan: fix assertion when qy_needs_dequant (llama...
2025-03-04 Molly Sophiaggml-cpu: Fix build with sve (llama/12059)
2025-03-03 cmdr2cuda: unary ops as float + de-duplicate (#1130)
2025-02-28 cmdr2cuda/vulkan: specify fp32-only support for some operati...
2025-02-28 cmdr2cuda/cpu: Increase support for fp16 unary operations...
2025-02-27 Georgi Gerganovsync : whisper.cpp
2025-02-27 Diego Devesawhisper : support GGML_BACKEND_DL (whisper/2843)
2025-02-27 Georgi Gerganovci : fix workflow name
2025-02-27 Georgi Gerganovexamples : remove dr_wab.h (#1127)
2025-02-27 Georgi Gerganovsync : whisper.cpp
2025-02-27 Georgi Gerganovcommon : separate whisper sources (whisper/2846)
2025-02-27 Georgi Gerganovcommon : fix build min/max (whisper/2845)
2025-02-27 Dmitry Atamanovexamples : use miniaudio for direct decoding flac,...
2025-02-27 midnightcmake : fix compile assumptions for power9/etc (whisper...
2025-02-26 petterreinholdtsenTold cmake to install ggml-cpp.h as a public header...
2025-02-25 cmdr2Support pure float16 add/sub/mul/div operations in...
2025-02-25 Georgi Gerganovsync : llama.cpp
2025-02-25 Gian-Carlo... metal : copy kernels for quant to F32/F16 conversions...
2025-02-25 lhezopencl: fix for small models (llama/11950)
2025-02-25 Neo Zhang JianyuOptimize mul_mat for Q4_0 on Intel GPU (llama/12035)
2025-02-25 Akarshan BiswasSYCL: Fix GGML_SYCL_DEBUG macro (llama/11995)
2025-02-25 Aaron Teoggml-cpu: Support s390x SIMD Instruction Set (llama...
2025-02-25 Johannes GäßlerCUDA: app option to compile without FlashAttention...
2025-02-25 Johannes GäßlerCUDA: optimize FA for GQA + large batches (llama/12014)
2025-02-25 Gian-Carlo... cuda: Add Q5_1, Q5_0, Q4_1 and Q4_0 to F32 conversion...
2025-02-25 PureJourneyCUDA: correct the lowest Maxwell supported by CUDA...
2025-02-25 BodhiMUSA: support ARM64 and enable dp4a .etc (llama/11843)
2025-02-25 Charles Xuggml-cpu: Add CPU backend support for KleidiAI library...
2025-02-25 Prashant Vithuleggml: aarch64: implement SVE kernels for q3_K_q8_K...
2025-02-25 Johannes GäßlerCUDA: use async data loading for FlashAttention (llama...
2025-02-25 Rémy Ovulkan: implement several ops relevant for ggml_opt...
2025-02-25 Jeff Bolzvulkan: support multi/vision rope, and noncontiguous...
2025-02-25 Hale Chanmetal : fix the crash caused by the lack of residency...
2025-02-25 Adrian Kretzmetal : optimize dequant q6_K kernel (llama/11892)
2025-02-25 Georgi Gerganovrepo : update links to new url (llama/11886)
next