]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/shortlog
pkg/ggml/sources/whisper.cpp
2025-09-20 QeeweewCUDA: Accelerate MXFP4 table lookup using `__byte_perm...
2025-09-20 lhezopencl: fix support ops condition for `rms_norm` (llama...
2025-09-20 Ruben Ortlamvulkan: fix min subgroup 16 condition for mmid subgroup...
2025-09-20 Ihar Hrachyshkametal: fix regression when no metal devices are present...
2025-09-20 Johannes GäßlerCUDA: MoE helper in device code, better tile sizes...
2025-09-20 Georgi Gerganovmetal : add FA kernels for HS=40 (llama/15559)
2025-09-20 Chenguang LiCANN: ROPE cache sin/cos repeat (llama/15501)
2025-09-20 Ruben Ortlamvulkan: apply MUL_MAT_ID subgroup optimization to non...
2025-09-20 Jeff Bolzvulkan: Support FA with any multiple of 8 head sizes...
2025-09-20 Ruben Ortlamvulkan: enable Conv2D for Apple after MoltenVK fixed...
2025-09-20 Jeff Bolzvulkan: workaround MoltenVK compile failure in multi_ad...
2025-09-20 Johannes GäßlerCUDA: fix half2 -> half conversion for HIP (llama/15529)
2025-09-20 Jeff Bolzvulkan: optimize rms_norm, and allow the work to spread...
2025-09-20 Jeff Bolzvulkan: Rewrite synchronization to allow some overlap...
2025-09-20 Aclyvulkan : support ggml_mean (llama/15393)
2025-09-20 Jeff Bolzvulkan: optimize mul_mat_id loading row ids into shared...
2025-09-20 Reese Levineggml WebGPU: add support for quantization types (llama...
2025-09-20 rmatifggml: add `conv3d` op (llama/15182)
2025-09-20 Yavor Ivanovcuda : add Pad Reflect 1D support (llama/14659)
2025-09-20 Aaron Teoggml-cpu: Support Q5_0 and Q5_1 on s390x (llama/15486)
2025-09-20 Chenguang LiCANN: Optimize RMS_NORM using cache (llama/15419)
2025-09-20 Diego Devesasched : fix possible use of wrong ids tensor when offlo...
2025-09-20 Aclyvulkan : support conv_2d_dw with f16 weights (llama...
2025-09-20 Dong Won Kimvulkan: add exp operation (llama/15456)
2025-09-20 Jeff Bolzvulkan: Reuse conversion results in prealloc_y (llama...
2025-09-20 Xuan-Son Nguyenggml : fix condition of im2col on Metal backend (llama...
2025-09-20 R0CKSTARmusa: add GGML_UNUSED_VARS (llama/15446)
2025-09-20 Diego Devesasched : copy only the used experts when offloading...
2025-09-20 Johannes GäßlerCUDA: refactor FA support/selection code (llama/15454)
2025-09-20 Johannes GäßlerCUDA: replace GGML_CUDA_F16 with CUDA arch checks ...
2025-09-20 Jeff Bolzvulkan: shorten pipeline name strings (llama/15431)
2025-09-20 R0CKSTARmusa: fix build warnings (llama/15258)
2025-09-20 lhezopencl: mark `argsort` unsupported if cols exceed workg...
2025-09-20 SHUAI YANGCANN: optimize rope operator (llama/15335)
2025-09-20 R0CKSTARmusa: handle __hgt2_mask, available starting from MUSA...
2025-09-20 Marvin Gießingggml-cpu: add mxfp4 VSX intrinsics for Power9+ (ppc64le...
2025-09-20 Georgi Gerganovcuda : remove obsolete sources (ggml/1332)
2025-09-19 Carlos Zoidoggml : Fix MKL detection by quoting BLAS_INCLUDE_DIRS...
2025-09-08 Siva Mahadevanwhisper : prefer curl over wget in download scripts...
2025-09-05 Daniel Beveniusci : remove brew installation of cmake for macos-latest...
2025-09-04 Daniel Beveniustests : use CMake definitions for model/sample paths...
2025-08-24 TrebokoHandle negative value in padding (#3389)
2025-08-24 Thea Mukhimodels : update`./models/download-ggml-model.cmd`...
2025-08-18 Georgi Gerganovtalk-llama : sync llama.cpp
2025-08-18 Georgi Gerganovsync : ggml
2025-08-18 Reese Levineggml: Add initial WebGPU backend (llama/14521)
2025-08-18 Aaron Teoggml : initial zDNN backend (llama/14975)
2025-08-18 Georgi Gerganovcommon : handle mxfp4 enum
2025-08-18 compiladeggml-quants : fix make_qp_quants NANs and IQ1 assertion...
2025-08-18 Jeff Bolzvulkan: disable spirv-opt for bfloat16 shaders (llama...
2025-08-18 Jeff Bolzvulkan: Use larger workgroups for mul_mat_vec when...
2025-08-18 Dong Won Kimvulkan: support sqrt (llama/15370)
2025-08-18 Jeff Bolzvulkan: Optimize argsort (llama/15354)
2025-08-18 Jeff Bolzvulkan: fuse adds (llama/15252)
2025-08-18 Jeff Bolzvulkan: Support mul_mat_id with f32 accumulators (llama...
2025-08-18 Jeff Bolzvulkan: Add missing bounds checking to scalar/coopmat1...
2025-08-18 rmatifOpenCL: add initial FA support (llama/14987)
2025-08-18 lhezopencl: add initial mxfp4 support via mv (llama/15270)
2025-08-18 Georgi Gerganovvulkan : fix out-of-bounds access in argmax kernel...
2025-08-18 Georgi Gerganovvulkan : fix compile warnings on macos (llama/15340)
2025-08-18 Aaron Teoggml: initial IBM zDNN backend (llama/14975)
2025-08-18 Johannes GäßlerCUDA: fix negative KV_max values in FA (llama/15321)
2025-08-18 uvosHIP: Cleanup hipification header (llama/15285)
2025-08-18 Jeff Bolzvulkan: perf_logger improvements (llama/15246)
2025-08-18 Jason Niggml: fix ggml_conv_1d_dw bug (ggml/1323)
2025-08-18 Sigbjørn Skjæretcuda : fix GGML_CUDA_GRAPHS=OFF (llama/15300)
2025-08-18 Jonathan Graehlfinetune: SGD optimizer, more CLI args (llama/13873)
2025-08-18 uvosHIP: bump requirement to rocm 6.1 (llama/15296)
2025-08-18 Juddggml : update `ggml_rope_multi` (llama/12665)
2025-08-18 Georgi Gerganovggml : repack block_iq4_nlx8 (llama/14904)
2025-08-18 Oliver SimonsCUDA: Optimize `reduce_rows_f32` kernel, leading up...
2025-08-18 Tak-RSggml-rpc: chunk send()/recv() to avoid EINVAL for very...
2025-08-18 uvosHIP: disable sync warp shuffel operators from clr amd_w...
2025-08-18 Romain Biessysycl: Fix and disable more configurations of mul_mat...
2025-08-18 rmatifopencl: allow mixed f16/f32 `add` (llama/15140)
2025-08-18 Aman GuptaCUDA cmake: add `-lineinfo` for easier debug (llama...
2025-08-18 Chenguang LiCANN: GGML_OP_CPY optimization (llama/15070)
2025-08-18 R0CKSTARmusa: fix failures in test-backend-ops for mul_mat_id...
2025-08-18 hipuddingCANN: Add broadcast for softmax and FA (llama/15208)
2025-08-18 Charles Xukleidiai: fix unsigned overflow bug (llama/15150)
2025-08-18 David Zhaocuda: refactored ssm_scan and use CUB (llama/13291)
2025-08-18 Aman GuptaCUDA: add attention sinks for tile and wmma (llama...
2025-08-18 compiladegguf-py : add Numpy MXFP4 de/quantization support ...
2025-08-18 AN Longggml : fix field name when new ggml_backend (llama...
2025-08-18 Johannes GäßlerCUDA: attention sinks for mma FlashAttention (llama...
2025-08-18 lhezopencl: support sink in `soft_max` (attn sinks) (llama...
2025-08-18 Jeff Bolzvulkan: support fattn sinks (llama/15126)
2025-08-18 Jeff Bolzvulkan: Add env var to disable host visible vidmem...
2025-08-18 uvosHIP: add cmake option to enable compiler output of...
2025-08-18 Christian Kastnerggml: Skip backend library linking code when GGML_BACKE...
2025-08-18 Johannes GäßlerCUDA: GEMM for FP32/FP16/BF16 and ne11 <= 16 (llama...
2025-08-18 rmatiffix profiling crash (llama/15072)
2025-08-18 lhezopencl: add `swiglu_oai` and `add_id` (llama/15121)
2025-08-18 Diego Devesaggml : fix fallback to CPU for ununsupported ops (llama...
2025-08-18 Chenguang LiCANN: add support for ACL Graph (llama/15065)
2025-08-18 Georgi Gerganovllama : add gpt-oss (llama/15091)
2025-08-18 Romain Biessysycl: fix mul_mat selection (llama/15092)
2025-08-18 Christian Kastnercmake: Add GGML_BACKEND_DIR option (llama/15074)
2025-08-18 Jeff Bolzvulkan: fix build when using glslang that does not...
2025-08-18 Jeff Bolzvulkan: Use coopmat2 for conv2d (llama/14982)
next