ngxson/llama.cpp

Pull Requests Commits

ngxson committed 314 days ago

Verified 1f71edb1

fix Mistral3Model

ngxson committed 314 days ago

e37dec65

Apply suggestions from code review

ngxson committed 314 days ago

Verified 0822e156

correct CLIP_VISION arch name (because existing GGUF already use it)

ngxson committed 315 days ago

ad186f43

ngxson committed 315 days ago

43c2e757

rename Mode --> ModelBase

ngxson committed 315 days ago

93b5f71e

split into subclasses TextModel and VisionModel

ngxson committed 315 days ago

ddd7920e

ngxson committed 316 days ago

55651ad8

fix bad ctrl+f replace

ngxson committed 316 days ago

d59a7bb3

convert : experimental support for `--mmproj` flag

ngxson committed 316 days ago

d5e03e62

clip : refactor, add `image_manipulation` and `llava_uhd` classes (#13011)

ngxson committed 316 days ago

Verified 37b9f0d2

main : Fix Ctrl+D/newline handling (#12951)

danielzgtg committed 316 days ago

Verified 64082100

gguf-py : GGUF Editor GUI - Python + Qt6 (#12930)

christopherthompson81 committed 316 days ago

Verified aff9d107

server : use std::move whenever possible (#12936)

ngxson committed 316 days ago

Verified 35370ba9

SYCL: Refactor and enable FP16 in binary broadcast OPs (#12975)

qnixsynapse committed 317 days ago

Verified 8d660057

mtmd : add methods to access `mtmd_image_tokens` (#12906)

ngxson committed 317 days ago

Verified b9154ecf

rpc : add RPC_CMD_HELLO (#12955)

rgerganov committed 317 days ago

Verified 2db9ba14

graph : make FA compatible with MLA + add initial Metal kernels (#12953)

ggerganov committed 318 days ago

Verified 2f74c354

ggml: Re-enable CUDA graphs in presence of CONT and DUP nodes (#12970)

agray3 committed 318 days ago

Verified 207c22ec

CANN: Add support for async operator submission (#12864)

hipudding committed 318 days ago

Verified 7a395f67

llama : recognize IBM Granite 3.3 FIM tokens (#12988)

Noeda committed 318 days ago

Verified 971f245b

opencl: fix incorrect local_size index in profiling log (#12868)

kimminsu38oo committed 318 days ago

Verified 12b17501

vulkan: enable coopmat2 FA gqa and split_k optimizations more often (#12931)

jeffbolznv committed 318 days ago

Verified 015022bb

CANN: Add 310P operator support check (#12962)

noemotiovon committed 319 days ago

Verified b43d89e3

opencl: split `ggml-opencl.cl` into multiple files and cleanup (#12886)

lhez committed 319 days ago

Verified 80f19b41

metal : add FA-vec kernels for head size 96 (#12952)

ggerganov committed 320 days ago

Verified f8f820cc

CANN: Add x86 build ci (#12950)

hipudding committed 320 days ago

Verified 54a72720

CUDA/HIP: Share the same unified memory allocation logic. (#12934)

hjc4869 committed 320 days ago

Verified 84778e97

SYCL: Add ROPE vision kernel (#12887)

qnixsynapse committed 320 days ago

Verified 51067647

llama : DeepSeek V2/V3 MLA implementation (#12801)

jukofyork committed 320 days ago

Verified daa42288

Older