ggml-org/llama.cpp

Pull Requests Commits

mtmd: mtmd_audio_streaming_istft (#18645)

tdakhran committed 2 hours ago

Verified ccbc84a5

llama-params-fit: fix last devices with low VRAM (#18494)

JohannesGaessler committed 3 hours ago

Verified 68b4d516

ggml : optimize cuda ssm_scan using warp-level reduction (#18505)

Aadeshveer committed 3 hours ago

Verified 24af22fc

arg: use CSV escape style for multiple-value args (#18643)

ngxson committed 5 hours ago

Verified 07fbe19f

vulkan: support buffer_from_host_ptr (#18467)

jeffbolznv committed 5 hours ago

Verified ea13cba8

ggml-cuda: refactor cuda graph usage (#18637)

am17an committed 6 hours ago

Verified 090b137e

mmq.cu: tune mmq/rocblas switching for RDNA (#18537)

Beinsezii committed 6 hours ago

Verified 96892952

server : add thinking content blocks to Anthropic Messages API (#18551)

686f6c61 committed 6 hours ago

Verified 3d26a09d

gguf-py : add requests to dependencies (#18629)

ckastner committed 14 hours ago

Verified bd2a93d4

ggml : fix avx512bf16 build (#18623)

angt committed 15 hours ago

Verified e75ee110

CANN: Make `valid_values` variable `static const` (#18627)

rauletorresc committed 18 hours ago

Verified da9b8d33

ggml webgpu: add CEIL operation support (#18605)

tnguyen21 committed 1 day ago

Verified e443fbcf

model : add LFM2-ColBert-350M (#18607)

tdakhran committed 1 day ago

Verified 73d284a2

CUDA: fix FA FP16 accumulator overflow for Granite (#18614)

JohannesGaessler committed 1 day ago

Verified df17a4c9

add YoutuVLForConditionalGeneration architectures (#18620)

f291400 committed 1 day ago

Verified 1871f0ba

ggml-cuda: check for srcs outside the cgraph (#18583)

am17an committed 1 day ago

Verified f47edb8c

server : fix router child env in containerized environments (#18562)

v-sayapin committed 1 day ago

Verified da143b99

vulkan: fix topk_moe_sigmoid_norm_bias failures in GLM-4.6 (#18582)

jeffbolznv committed 1 day ago

Verified f1768d8f

models : fix backend assignment for Granite/Nemotron graphs (#18599)

ggerganov committed 1 day ago

Verified 2da64a2f

vulkan: handle quantize_q8_1 overflowing the max workgroup count (#18515)

jeffbolznv committed 1 day ago

Verified b37124d2

llama : refactor rope_freq_base/scale_swa conversion and init (#18553)

CISC committed 1 day ago

Verified eadc4184

CANN: add operator fusion support for ADD + RMS_NORM (#17512)

noemotiovon committed 1 day ago

Verified 67e3f6f6

doc: clarify that steps also apply to linux for opencl (#18002)

alosslessdev committed 1 day ago

Verified 92ac1e01

ci : init git lfs in every build for RISC-V (#18590)

alitariq4589 committed 1 day ago

Verified 8e3a7611

sampling : add support for backend sampling (#17004)

danbev committed 2 days ago

Verified d3dce4e0

model : mtmd : make input norm optional in LFM2-VL (#18594)

tdakhran committed 2 days ago

Verified 4974bf53

CUDA: disable cuda graph when using n-cpu-moe (#18593)

am17an committed 2 days ago

Verified 908a9e5a

ggml-cuda: remove unused params in ggml_cuda_graph (#18579)

am17an committed 2 days ago

Verified 5126c41c

common/grammar : replace problematic backtracking regex `[\s\S]*` (#18342)

aldehir committed 3 days ago

Verified cef1d23c

graph : fix graph reuse logic when `n_pos_per_embd > 1` (#18566)

ggerganov committed 3 days ago

Verified c69c7ebc

Older