ggml-org/llama.cpp

Pull Requests Commits

imatrix : use GGUF regardless of the output filename

compilade committed 48 days ago

1ef3cc1a

imatrix : use GGUF by default

compilade committed 48 days ago

53f65c35

convert : text-only support for GLM-4.1V-9B-Thinking (#14823)

jacekpoplawski committed 48 days ago

Verified a12363bb

CUDA: fix overflow in FA, tune performance (#14840)

JohannesGaessler committed 48 days ago

Verified a86f52b2

CUDA: fix compilation with GGML_CUDA_F16 (#14837)

JohannesGaessler committed 49 days ago

Verified b284197d

ci : correct label refactor->refactoring (#14832)

CISC committed 49 days ago

Verified 221c0e0c

CUDA: fix quantized KV cache + multiple sequences (#14822)

JohannesGaessler committed 49 days ago

07a19e27

tests : add non-cont K,V FA tests

ggerganov committed 49 days ago

18f3b5ff

memory : handle saving/loading null layers in recurrent memory (#14675)

l3utterfly committed 49 days ago

Verified 7233358d

ggml: fix loongarch quantize_row_q8_1 error (#14827)

lixing-star committed 49 days ago

Verified 6c88b3bb

CANN: weight format to NZ for Ascend310P3 (#14407)

tqgy6 committed 49 days ago

Verified 14c28dfc

CUDA: add fused rms norm (#14800)

am17an committed 49 days ago

Verified 8c988fa4

ggml : model card yaml tab->2xspace (#14819)

csabakecskemeti committed 50 days ago

Verified acd6cb1c

vulkan: fix rms_norm_mul to handle broadcasting dim0 (#14817)

jeffbolznv committed 50 days ago

Verified 84712b60

llama : add model type detection for rwkv7 7B&14B (#14816)

MollySophia committed 50 days ago

Verified d4d1522b

imatrix: add option to display importance score statistics for a given imatrix file (#12718)

EAddario committed 50 days ago

Verified d1aa0cc5

Mtmd: add a way to select device for vision encoder (#14236)

stduhpf committed 50 days ago

Verified c8ade300

cuda : implement bf16 cpy ops and enable bf16 cont (#14763)

CISC committed 50 days ago

Verified e28c0b80

opencl: remove unreachable `return` (#14806)

lhez committed 50 days ago

Verified 8e6f8bc8

server : allow setting `--reverse-prompt` arg (#14799)

MollySophia committed 50 days ago

Verified adef8178

cuda: remove linking to cublasLt (#14790)

yeahdongcn committed 50 days ago

Verified 48b86c4f

opencl: fix `im2col` when `KW!=KH` (#14803)

CISC committed 50 days ago

Verified 38d3af1b

opencl: add conv2d kernel (#14403)

rmatif committed 51 days ago

Verified 6c9ee3b1

sycl: Fix im2col (#14797)

Rbiessy committed 51 days ago

Verified cd465d82

kleidiai: add support for get_rows (#14676)

chaxu01 committed 51 days ago

Verified 92204260

docs : fix backends table in README.md (#14796)

rgerganov committed 51 days ago

Verified 2ba1333b

vulkan/cuda: Fix im2col when KW!=KH (#14789)

jeffbolznv committed 51 days ago

Verified c2e058f1

llama : fix `--reverse-prompt` crashing issue (#14794)

MollySophia committed 51 days ago

Verified c82d48ec

server : add parse_special option to /tokenize endpoint (#14783)

IsaacDynamo committed 51 days ago

Verified b4efd77f

docs : fix link for tools/perplexity in README.md (#14780)

am17an committed 52 days ago

Verified 2be60cbc

Older