ngxson/llama.cpp

Pull Requests Commits

edit existing line in hot topics

ngxson committed 268 days ago

db85de10

correct order W, H of clip_embd_nbytes_by_img

ngxson committed 268 days ago

496f1ce8

Merge branch 'master' into xsn/mtmd_qwen2vl

ngxson committed 268 days ago

e9bff424

llama-graph : fix text position for mrope (#13159)

ngxson committed 268 days ago

Verified b6ce7430

model : Nomic Embed Text V2 with Mixture-of-Experts (MoE) architecture (#12466)

manyoso committed 268 days ago

Verified 5f5e39e1

clip : fix model size display (#13153)

ngxson committed 268 days ago

Verified eaea3253

fix(rpc): Improve input validation and error handling (#13069)

thevilledev committed 268 days ago

Verified 43ddab6e

deprecate llama-qwen2vl-cli

ngxson committed 268 days ago

d23fdc28

llama-bench: add `-d` depth arg (#13096)

thevishalagarwal committed 268 days ago

Verified 1831f538

Merge branch 'master' into xsn/mtmd_qwen2vl

ngxson committed 268 days ago

14a6ab15

mtmd : fix glm-edge redundant token count (#13139)

ngxson committed 268 days ago

Verified 4e87962e

working version

ngxson committed 268 days ago

b3035840

context : do not clear output buffer on reserve (#13152)

pockers21 committed 268 days ago

Verified fb0471d1

Merge branch 'master' into xsn/mtmd_qwen2vl

ngxson committed 269 days ago

513e9c9e

llama : (mrope) allow using normal 1D position for text token (#13138)

ngxson committed 269 days ago

Verified d2b2031e

clip : refactor set input for cgraph + fix qwen2.5vl input (#13136)

ngxson committed 269 days ago

Verified 5fa9e63b

SYCL: Add all missing unary kernels (#13074)

qnixsynapse committed 269 days ago

Verified a4c340f9

readme : update hot topics (#13150)

ggerganov committed 269 days ago

Verified d0a417f3

common : fix noreturn compile warning (#13151)

ggerganov committed 269 days ago

Verified 43f2b071

llama-chat : fix typo GML --> GLM (#13143)

ngxson committed 269 days ago

Verified e5d6c255

musa: fix typo in cc control (#13144)

yeahdongcn committed 269 days ago

Verified f0dd6a19

CUDA: fix q_nope_absorbed prec for DS 2 Lite f16 (#13137)

JohannesGaessler committed 269 days ago

Verified 69699be4

arg : fix unused variable (#13142)

ngxson committed 269 days ago

Verified 85f36e5e

llama-bench : Add `--override-tensors` arg (#12922)

4onen committed 269 days ago

Verified c0a97b76

llama-chat : fix wrong template in GLM4-0414 (#13140)

matteoserva committed 269 days ago

Verified ced44be3

decode_embd_batch::set_position_...

ngxson committed 269 days ago

8646e364

mtmd : add qwen2vl and qwen2.5vl

ngxson committed 269 days ago

8742f8a2

llava : add clip_n_output_tokens, deprecate clip_n_patches

ngxson committed 269 days ago

6f7a55c4

musa: fix build warning (#13129)

yeahdongcn committed 270 days ago

Verified e291450b

Fixes Qwen2.5VL segfault during inference with https://github.com/ggml-org/llama.cpp/pull/12402 as has_qwen2vl_merger migration was incomplete (#13133)

LostRuins committed 270 days ago

Verified 59e991c2

Older