ngxson/llama.cpp

Pull Requests Commits

ngxson committed 103 days ago

70e73128

fix conversion script

ngxson committed 103 days ago

bd38d7fd

correct projector

ngxson committed 103 days ago

a342f52c

Merge branch 'master' into xsn/paddleocr

ngxson committed 105 days ago

03a4a49f

Merge branch 'master' into xsn/paddleocr

ngxson committed 105 days ago

030f1b20

mtmd-cli : allow using --jinja (#16718)

ngxson committed 105 days ago

Verified d0660f23

Manually link -lbsd to resolve flock symbol on AIX (#16610)

mehendarkarprajwal committed 105 days ago

Verified fe6a9882

ggml-cuda: use passed ops instead of hardcoded ops (#16712)

am17an committed 105 days ago

Verified 061f0eff

server : send partial stop string when <EOG> is reached (#15007)

matteoserva committed 105 days ago

Verified 8cf6b42d

sycl: use async memory allocation to fix crashes during graph recording (#16644)

mmichel11 committed 106 days ago

Verified 9de9672a

Add experimental ggml-hexagon backend for the Hexagon NPU (#16547)

max-krasnyansky committed 106 days ago

Verified 63d2fc46

Revert "ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_v…" (#16723)

slaren committed 106 days ago

Verified a2e0088d

webui: introduce OpenAI-compatible model selector in JSON payload (#16562)

ServeurpersoCom committed 106 days ago

Verified 9b9201f6

output text but gibberish

ngxson committed 106 days ago

ea2fbb80

ngxson committed 106 days ago

ddfaca72

rm V_RESMPL_ATTN_NORM

ngxson committed 106 days ago

be802893

fix chat template

ngxson committed 106 days ago

ac41b54a

ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_vec_set_f32 for faster fills (#16522)

sirus20x6 committed 106 days ago

Verified 19a5a3ed

tests : fix test-thread-safety when compiling with multiple backends (#16699)

Acly committed 106 days ago

Verified d8eaa26e

convert vision encoder ok

ngxson committed 106 days ago

e7a485c8

CUDA: fix bug in topk-moe softmax (#16711)

am17an committed 107 days ago

Verified 9285325c

CUDA: topk-moe: add optional parameter for gpt-oss (#16649)

am17an committed 107 days ago

Verified 03792ad9

CUDA: better error for FA kernel with 0 occupancy (#16643)

JohannesGaessler committed 107 days ago

Verified 51d1a8c9

ngxson committed 107 days ago

366abe71

ggml: add ggml_can_fuse_subgraph (#16662)

am17an committed 107 days ago

Verified 4926419c

opencl: fix warnings and clean up profiling (#16688)

lhez committed 108 days ago

Verified 6ea37f57

vulkan: Handle FA with all -inf mask values (#16447)

jeffbolznv committed 108 days ago

Verified fb349848

sycl : add PAD_REFLECT_D1 operator support (#16145)

ye-NX committed 108 days ago

Verified 6de8ed75

model : add BailingMoeV2 support (#16063)

CISC committed 108 days ago

Verified 84bf3c67

Handle legacy 'context' attachments (#16687)

allozaur committed 108 days ago

Verified c9c1972e

Older