ggerganov/llama.cpp

Pull Requests Commits

Fixes Qwen2.5VL segfault during inference with https://github.com/ggml-org/llama.cpp/pull/12402 as has_qwen2vl_merger migration was incomplete

LostRuins committed 28 days ago

37ae6a28

clip : Add Qwen2.5VL support (#12402)

HimariO committed 28 days ago

Verified ca2bb89e

common : add common_remote_get_content (#13123)

ngxson committed 29 days ago

Verified 2d451c80

clip : improve projector naming (#13118)

ngxson committed 29 days ago

Verified 4753791e

ggml: move fp16/bf16 conversion optimizations to CPU backend + export conversion APIs (#13107)

SongXiaoXi committed 29 days ago

Verified 77d5e9a7

grammar : handle maxItems == 0 in JSON schema (#13117)

rick-github committed 29 days ago

Verified d5fe4e81

llama : fix K-shift with quantized K and BLAS backend (#13113)

slaren committed 30 days ago

Verified 295354ea

Force FP32 compute in GLM4 FFN Down (#13101)

city96 committed 30 days ago

Verified 558a7647

clip : fix pixtral on some GPU backends (#13097)

ngxson committed 30 days ago

Verified edb18b6e

change the reorder tensor from init to execute OP (#13003)

NeoZhangJianyu committed 30 days ago

Verified 514c4560

rpc : do not wait for response when sending RPC_CMD_SET_TENSOR (#12943)

rgerganov committed 30 days ago

Verified 553a5c3a

clip : remove boi/eoi embeddings for GLM-edge model (#13081)

ngxson committed 31 days ago

Verified 13be08da

embeddings : fix batch sizes (#13076)

ggerganov committed 31 days ago

Verified 226251ed

ggml : fix trailing whitespaces (#0)

ggerganov committed 31 days ago

87616f06

ggerganov committed 31 days ago

63b49114

ggml : Depthwise 2D convolution (ggml/1152)

Acly committed 31 days ago

c6e8cc28

CUDA: use switch statements in constexpr functions (#13095)

JohannesGaessler committed 31 days ago

Verified b10d8bfd

cmake : do not include ./src as public for libllama (#13062)

ggerganov committed 31 days ago

Verified 13b45488

clang-tidy : disable warning about missing math parenthesis (#13091)

ggerganov committed 31 days ago

Verified 572b3141

arg : add --no-mmproj-offload (#13093)

ngxson committed 31 days ago

Verified 7c727fbe

arg : clean up handling --mmproj with -hf (#13082)

ngxson committed 31 days ago

Verified 80982e81

metal : fix floating-point range of attention scores in FA kernels (#13090)

ggerganov committed 31 days ago

Verified 7604a7d6

vulkan: matmul gcn tuning (#13016)

netrunnereve committed 31 days ago

Verified b3b6d862

llama-mtmd-cli: Sigint rework in mtmd vision example (#13080)

pl752 committed 32 days ago

Verified 56304069

mtmd : Support Pixtral 12B (#13065)

ngxson committed 32 days ago

Verified ecda2ec4

convert : Append mult-eos,half-rope,bos to GLM4-0414 and Z (#13021)

piDack committed 32 days ago

Verified eb1776b1

rpc : add command line option for number of threads for the CPU backend (#13060)

rgerganov committed 32 days ago

Verified 2cca6c01

CUDA: noncont MMVQ + batched bs1 MUL_MAT_ID (#13014)

JohannesGaessler committed 33 days ago

Verified 658987cf

mtmd : support SmolVLM (version 1 and 2) (#13050)

ngxson committed 33 days ago

Verified dc39a5e7

security : add note about RPC and server functionality (#13061)

ggerganov committed 33 days ago

Verified ab47dec3

Older