ngxson/llama.cpp

Pull Requests Commits

Merge branch 'master' into xsn/move_llava_to_mtmd

ngxson committed 235 days ago

37154044

clip : fix confused naming ffn_up and ffn_down (#13290)

ngxson committed 235 days ago

Verified 5215b91e

convert : bailingmoe : set yarn metadata if present (#13312)

CISC committed 235 days ago

Verified ae803bfc

change ref everywhere

ngxson committed 235 days ago

86e4ab13

mv llava to mtmd

ngxson committed 235 days ago

a87e1cf2

SYCL: Disable mul_mat kernels for noncontiguous tensor b (#13308)

qnixsynapse committed 235 days ago

Verified 66645a52

mtmd : add C public API (#13184)

ngxson committed 235 days ago

Verified 27aa2595

rpc : use backend registry, support dl backends (#13304)

slaren committed 235 days ago

Verified 9fdfcdae

ggml : activate s390x simd for Q3_K (#13301)

taronaeo committed 235 days ago

Verified 6eb7d25c

llava/mtmd : fixes to fully support dl backends (#13303)

slaren committed 235 days ago

Verified 86bd60d3

llama : build windows releases with dl backends (#13220)

slaren committed 236 days ago

Verified 9f2da587

CUDA: fix race condition in MMQ stream-k fixup (#13299)

JohannesGaessler committed 236 days ago

Verified 93c4e239

CUDA: fix race condition in MMQ ids_dst (#13294)

JohannesGaessler committed 236 days ago

Verified 8afbd968

vulkan: Additional type support for unary, binary, and copy (#13266)

jeffbolznv committed 236 days ago

Verified 8ae5ebcf

imatrix: fix oob writes if src1 is not contiguous (#13286)

JohannesGaessler committed 236 days ago

Verified 3e959f09

clip : revert the change of BOI/EOI token for GLM-edge (⚠️ breaking change) (#13259)

ngxson committed 236 days ago

Verified 36667c8e

llama : Llama-3_1-Nemotron-Ultra-253B-v1 support (#12843)

ymcki committed 236 days ago

Verified 3bf785f3

llama : move end-user examples to tools directory (#13249)

slaren committed 237 days ago

Verified 1d36b367

sync : ggml (#13268)

ggerganov committed 237 days ago

Verified b3444392

context : fix reorder logic (#13267)

ggerganov committed 237 days ago

Verified a75cb30d

ggml : Enable MMA for BF16 in llamafile_sgemm (#13148)

shalinib-ibm committed 237 days ago

Verified 3f3769ba

llama-model : support Qwen2 embedding models and pooling_mode_lasttoken (#13245)

cebtenzzre committed 237 days ago

Verified 2f567611

convert : use correct context length for nomic-embed-text-v2 (#13216)

cebtenzzre committed 237 days ago

Verified 7d212348

convert : converting mmproj for Qwen2/2.5VL from convert_hf_to_gguf (#13209)

ngxson committed 237 days ago

Verified 074e42ab

kv-cache : separate recurrent vs non-recurrent impl (#12799)

ggerganov committed 237 days ago

Verified c642bc01

llama : orion rope type is neox (#13261)

CISC committed 238 days ago

Verified cb06a3c3

llama : plamo rope type is neox (#13260)

CISC committed 238 days ago

Verified 626083fa

llama-chat : reset glmedge chat template (#13253)

piDack committed 238 days ago

Verified 2af68801

mtmd-cli : fix out_of_range when input image path is empty (#13244)

ahmedshakill committed 238 days ago

Verified e84773ab

server : add cache reuse card link to help (#13230)

ggerganov committed 238 days ago

Verified fab647e8

Older