LLM inference in C/C++
C++ ggml
Commits on master
vendor : update cpp-httplib to 0.54.0 (#27919)
dc7aecf Alessandro de Oliveira Faria (A.K.A.CABELO) committed 2 months ago
hexagon: support for device discovery and create sessions on demand (#27785)
9e54e68 Max Krasnyansky committed 2 months ago
sycl: split long rows in TOP_K instead of one work-group per row (#27847)
370cb12 Titaniumtown committed 2 months ago
metal : fix null-pipeline crash for F16 src1 mul_mat/mul_mat_id (#25648)
d882575 QuintinShaw committed 2 months ago
memory : copy Hadamard matrix to k_rot tensor only if it has buffer assigned to prevent crashes during context shift of unquantized K cache (#27967)
bdf3955 fairydreaming committed 2 months ago
ggml: allow passing alloc dependencies in graph_optimize (#27301)
57291f2 Aman Gupta committed 2 months ago
opencl: use a better matmul path on two Adreno GPU generations (#27640)
c841aee Hongqiang Wang committed 2 months ago
metal : add remaining fa-vec tunings for M4 Pro (#27915)
17252c7 Niklas Wenzel committed 2 months ago
vulkan: combine duplicated fastdiv functions, rename the one optimizing small divs (#27526)
c9ca51c Jeff Bolz committed 2 months ago
model: qwen4exp: reduce number of graph splits (#27880)
6fe7498 Xuan-Son Nguyen committed 2 months ago
vulkan: fix missing view-alias dependencies in ggml_vk_graph_optimize (#27812)
b387ddf Eric A Stalee committed 2 months ago
ggml : fix conv_transpose_2d for multiple batches (#26132)
a43c398 Tekin Ertekin committed 2 months ago
Vulkan: add hoisting support for row IDs and expert count in shaders (#26686)
90c26fc ravel7524 committed 2 months ago
OpenVINO: Update OV to 2026.3.1, whisper.cpp support, Qwen3.5 on NPU, and new ops (#27843)
511f9c1 Zijun Yu committed 2 months ago
convert: prevent ndarray conversion in LazyChunkedTensor (#27869)
b19cbe9 Xuan-Son Nguyen committed 2 months ago
sycl: use TILE for quantized KV decode on BMG (#26689)
d077b4c Ozymandias_EBON committed 2 months ago
sycl: bind the f16 KV cache in place for the oneDNN SDPA path (#27468)
be87620 Titaniumtown committed 2 months ago
metal : add fa-vec tunings for M3 Max, M5 and M5 Pro (#27863)
8963a9b Georgi Gerganov committed 2 months ago
tests : run test-save-load-state across all architectures (#27755)
4e97ac8 Georgi Gerganov committed 2 months ago