LLM inference in C/C++
C++ ggml
Commits on master
scripts: Add script to verify API/ABI compatibility (#28579)
dfe4516 Christian Kastner committed 2 months ago
tests : add fusion baseline README and broaden fusion CI triggers (#28893)
bbdd9f2 Georgi Gerganov committed 2 months ago
models : fix incorrect uses of get_key_or_arr (#28868)
97e4ca7 Sigbjørn Skjæret committed 2 months ago
llama: add Maple 20B-A1B ternary MoE architecture (CPU) (#27000)
3d10bcd Alex committed 2 months ago
ggml-cpu : disable PCH and fix CACHE_LINE_SIZE ambiguity to fix heap corruption (#28882)
2f53959 Georgi Gerganov committed 2 months ago
ci : trigger self-hosted CI on changes to ci/run.sh (#28859)
89fe242 Georgi Gerganov committed 2 months ago
sycl : fix oneDNN scratchpad breaking the pool free order (#28704)
661643e Łukasz Ślusarczyk committed 2 months ago
common : move llama_n_rs_seq to before llama_decode (#28749)
093a2f8 Daniel Bevenius committed 2 months ago
ggml-cuda: fallback to F32 on device without BF16 hardware acceleration (#28846)
ad6c668 thelittlefireman committed 2 months ago
grammar : coalesce find + insert into a single insert and adjust move/copy mechanics (#26885)
7a16a6c Clint Herron committed 2 months ago
tests : exclude HY_V4 from WebGPU test-llama-archs tests (#28855)
5f436dd fairydreaming committed 2 months ago
models : guard the expert FFN size fallback in nemotron-h against a zero divisor (#28779)
e49d2c2 Yaniss Amazouz committed 2 months ago
ggml-cpu(s390x): guard VXE-only repack helpers (#28775)
6978052 Bernard Ladenthin committed 2 months ago
tests : fix typo in test-quant-type-selection for nemotron 3 nano (#28835)
243a308 Michael Taylor committed 2 months ago
ci : run editorconfig and code-style checks on ubuntu-slim (#28854)
c95f8e4 Georgi Gerganov committed 2 months ago
pi : prefer PI_MODEL_NAME env var for model disclosure (#28853)
bc52a12 Georgi Gerganov committed 2 months ago
ci : cap test-backend-ops parallel jobs at 2 and add a 3600s timeout (#28833)
002a12a Georgi Gerganov committed 2 months ago
opencl: apply the noshuffle row-alignment rule to q4_K, q5_K and q8_0, not just q6_K (#28575)
56b9eb2 Hongqiang Wang committed 2 months ago
chat : improve parsing of complex types in qwen3-coder (#28742)
790cf51 Aldehir Rojas committed 2 months ago