]> git.djapps.eu Git - pkg/ggml/sources/ggml/shortlog
pkg/ggml/sources/ggml
2026-03-28 Georgi Gerganovmetal : add FA instantiations for HSK=512, HSV=512...
2026-03-28 Max Krasnyanskyhexagon: general DMA and Binary Op fixes for large...
2026-03-28 lhezopencl: add q6_K gemm and gemv kernels for Adreno ...
2026-03-28 las7rpc : RCE patch (llama/20908)
2026-03-28 Rashid Ul Islammetal: add CONV_3D (llama/19927)
2026-03-28 Chenguang LiCANN: add RoPE cache preload before ACL graph capture...
2026-03-28 Dan Hoffmanfix(openvino): explicit memset in buffer_context alloca...
2026-03-28 shaofeiqiopencl: add flattened Q4_K mv and general Q4_K mm ...
2026-03-28 Johannes GäßlerCUDA: fix BF16 FA compilation (llama/20865)
2026-03-28 Neo Zhangsupport bf16 and quantized type (llama/20803)
2026-03-28 Patrick Buckleyggml-cuda: native bf16 flash attention for vec kernel...
2026-03-28 Gaurav GargIncrease number of output elements per-thread block...
2026-03-28 y198fix(rpc): prevent division by zero in deserialize_tenso...
2026-03-28 Matt CoralloAdd shader count for Intel Arc Pro B60 (llama/20818)
2026-03-28 shalinib-ibmggml-cpu: add always_inline to tinyBLAS_PPC accumulator...
2026-03-28 Jeff Bolzvulkan: change gated_delta_net to shard a column across...
2026-03-28 hipuddingCANN: add BF16 support for core operators (llama/20152)
2026-03-28 Sundaram krishnanggml: guard KleidiAI DOWNLOAD_EXTRACT_TIMESTAMP for...
2026-03-28 Rail Chabdarovhip: Avoid compiler bug in RDNA code generation during...
2026-03-28 Yiwei Shaohexagon: add Matrix Extensions (HMX) for Hexagon NPU...
2026-03-28 uvosci : add hip quality check (llama/20430)
2026-03-28 Reese Levineggml webgpu: ops support for qwen3.5 (SET, TRI_SOLVE...
2026-03-28 Evevulkan: dequantize iq4_xs 4 at a time (llama/20657)
2026-03-28 Charles Xucmake : fix build warning when kleidiai is enabled...
2026-03-28 Chenguang LiCANN: handle in-place ROPE on non-contiguous f32 tensor...
2026-03-28 Georgi Gerganovsync : llama.cpp
2026-03-28 Masashi Yoshimuraggml-webgpu: Update the `RMS_NORM` preprocessor and...
2026-03-28 Georgi Gerganovsync : llama.cpp
2026-03-28 Masashi Yoshimuraggml-webgpu: Add supports for `DIAG` and `TRI` (llama...
2026-03-28 Chenguang LiCANN: support flash attention for head dim not multiple...
2026-03-28 Reese LevineMove to no timeout for WaitAny in graph submission...
2026-03-28 Shaw Nguyenggml-cpu/x86: fix unused changemask warning in repack...
2026-03-28 uvosHIP : ignore return of hipMemAdvise [no ci] (llama...
2026-03-28 Krishna Sridharhexagon: add neg, exp, sigmoid, softplus ops, cont...
2026-03-28 Ruben Ortlamvulkan: disable mmvq on Intel Windows driver (llama...
2026-03-28 Kevin Hannonggml-blas: set mkl threads from thread context (llama...
2026-03-28 Taimur Ahmadggml-cpu: fix RVV checks in quants and repacking (llama...
2026-03-28 Ruben Ortlamvulkan: async and event fixes (llama/20518)
2026-03-28 Justin Bradfordkleidiai : fix MUL_MAT support for batched (3D) inputs...
2026-03-28 Ruben Ortlamvulkan: allow graphics queue only through env var ...
2026-03-28 Neo Zhangehance UPSCALE to support all UT cases (llama/20637)
2026-03-28 Martin Klacerkleidiai: add data type check to get_tensor_traits...
2026-03-28 Ruben Ortlamvulkan: fix flash attention dot product precision ...
2026-03-28 Aman GuptaCUDA: GDN hide memory latency (llama/20537)
2026-03-28 Sigbjørn Skjæretsycl : fix for untransposed GDA recurrent state (llama...
2026-03-16 Georgi Gerganovci : disable AMX jobs
2026-03-16 Georgi Gerganovggml : bump version to 0.9.8 (#1442) v0.9.8
2026-03-16 Georgi Gerganovggml : restore ggml_type_sizef() to aboid major version...
2026-03-16 Georgi Gerganovreadme : simplify
2026-03-16 Georgi Gerganovsync : whisper.cpp
2026-03-16 Georgi Gerganovggml : try fix arm build (whisper/0)
2026-03-15 David366AIggml : extend im2col f16 (#1434)
2026-03-15 Georgi Gerganovcommon : add nvfp4 (#0)
2026-03-15 Georgi Gerganovsync : llama.cpp
2026-03-15 Johannes GäßlerCUDA: limit number of FA stream-k CUDA blocks (llama...
2026-03-15 Pascalggml: avoid creating CUDA context during device init...
2026-03-15 MoonShadowggml/hip: fix APU compatibility - soft error handling...
2026-03-15 Bartowskiggml : guard against sumq2 being 0 in IQ4_NL (llama...
2026-03-15 PikaPikachucuda : add RDNA4-specific MMVQ parameter table for...
2026-03-15 Ruben Ortlamvulkan: use graphics queue on AMD (llama/20551)
2026-03-15 Georgi Gerganovmetal : add FA specialization for HSK = 320, HSV =...
2026-03-15 Max Krasnyanskyhexagon: Q4_0 and MXFP4 repack fixes (llama/20527)
2026-03-15 Neo Zhangadd op gated_delta_net (llama/20455)
2026-03-15 Adrien Gallouëtggml : add native AVX512-FP16 support for F16 operation...
2026-03-15 WallentriUse fp32 in cuBLAS V100 to avoid overflows, env variabl...
2026-03-15 Zijun Yuggml : add OpenVINO backend (llama/15307)
2026-03-15 Rail ChabdarovFix data race in CUDA's "cpy" kernel (influences GGML...
2026-03-15 lhezopencl: fix l2_norm (llama/20480)
2026-03-15 Georgi Gerganovgraph : remove redundant GDN state transposes (llama...
2026-03-15 rehan-10xengineerggml-cpu: add RVV vec dot kernels for quantization...
2026-03-15 Adrien Gallouëtggml : fix typo gmml (llama/20512)
2026-03-15 Georgi Gerganovmetal : fix l2 norm scale (llama/20493)
2026-03-15 Georgi Gerganovllama : disable graph reuse with pipeline parallelism...
2026-03-15 Ruben Ortlamtest-backend-ops: allow loading tests from file and...
2026-03-15 ProgenyAlphavulkan: add GATED_DELTA_NET op support (llama/20334)
2026-03-15 ProgenyAlphavulkan: fix SSM_CONV PP scaling with large ubatch sizes...
2026-03-15 Georgi Gerganovsync : llama.cpp
2026-03-15 Georgi Gerganovmetal : avoid divisions in bin kernel (llama/20426)
2026-03-15 Georgi Gerganovsync : llama.cpp
2026-03-15 Jeff Bolzvulkan: fix l2_norm epsilon handling (llama/20350)
2026-03-15 Jeff Bolzvulkan: fix OOB check in flash_attn_mask_opt (llama...
2026-03-15 Masato Nakasakavulkan: Fix ErrorOutOfHostMemory on Intel GPU when...
2026-03-15 lhezopencl: use larger workgroup size for get_rows (llama...
2026-03-15 shaofeiqiopencl: add cumsum op (llama/18981)
2026-03-15 uvoship: compile debug builds with -O2 on hip to avoid...
2026-03-15 Masashi Yoshimuraggml-webgpu: Add supports for `GGML_OP_REPEAT` (llama...
2026-03-15 Georgi Gerganovllama : enable chunked fused GDN path (llama/20340)
2026-03-15 Richard Davisonggml : add NVFP4 quantization type support (llama/19769)
2026-03-15 Daniel Beveniusllama : add support for Nemotron 3 Super (llama/20411)
2026-03-15 Georgi Gerganovmetal : fix capture_compute counter logic (llama/20410)
2026-03-15 Georgi Gerganovmetal : fix q5_k mul_mv register spill (llama/20399)
2026-03-15 Georgi Gerganovmetal : add env var to trigger graph capture (llama...
2026-03-15 uvosggml-cuda: gdn use shared mem for HIP (llama/20366)
2026-03-15 uvoscuda/hip: fix loop unrolling in ssm-conv (llama/20369)
2026-03-15 Neo Zhangfix op rope, add rope_back (llama/20293)
2026-03-15 Neo Zhangfix for failed UT case: ACC, L2_NORM, UPSCALE, fused_gl...
2026-03-15 Georgi Gerganovggml : bump RPC version (llama/20330)
2026-03-15 Reese Levineggml webgpu: faster normal quant and some k-quant matri...
2026-03-15 Charles Xukleidiai : support for concurrent sme and neon kernel...
2026-03-15 Taimur Ahmadggml-cpu: add RVV repack GEMM and GEMV for quantization...
next