LLM inference in C/C++
C++ ggml
Commits on master
graph : fix unused input tensors in minimax m3 graph (#26519)
c50b34a Georgi Gerganov committed 3 months ago
model: M3: Move MSA into a new memory implementation (#26338)
67d5978 timkhronos committed 3 months ago
llama : allocate indexer cache only in "full" indexer layers (#26474)
563dec8 fairydreaming committed 3 months ago
CUDA: Add backend sampler for penalties sampler (#25262)
96278e3 Konrad Moren committed 3 months ago
CUDA: Fix data-races when reusing SMEM in block_reduce (#26385)
9bd4c09 Oliver Simons committed 3 months ago
server: add notice for upcoming default port change 8080 --> 9931 (#26508)
0b14b87 Xuan-Son Nguyen committed 3 months ago
opencl: limit local workgroup size for GLU operation (#26383)
0ab9d6f mgroeber9110 committed 3 months ago
metal: implement DeepSeek V4 hyper-connections (#26459)
fffbcbd Georgi Gerganov committed 3 months ago
convert: add option to create separate dspark GGUF (#26452)
3581ba0 Aman Gupta committed 3 months ago
opencl: bugfix increment ref_count in ggml_backend_opencl_init() (#26162)
c745be2 akleine committed 3 months ago
model : load MiMo V2 MTP tensors only if used (#26412)
75587a0 Sigbjørn Skjæret committed 3 months ago
vulkan: extend topk_moe fusion to support sqrt(softplus) (#26124)
a7a6d0d Jeff Bolz committed 3 months ago
vendor : update BoringSSL to 0.20260730.0 (#26353)
815a2a5 Alessandro de Oliveira Faria (A.K.A.CABELO) committed 3 months ago
agents: clarify comment style and jinja knowledge (#26405)
89482bd Xuan-Son Nguyen committed 3 months ago
chat : enable tool call in thinking for DS4 (#26269)
ddd4ec1 Piotr Wilkin (ilintar) committed 3 months ago