ggerganov/llama.cpp

Pull Requests Commits

opencl: add `backend_synchronize` (#13939)

lhez committed 129 days ago

Verified 71e74a3a

OpenCL: Add concat, tsembd, upscale, tanh, pad and repeat (#13840)

rmatif committed 129 days ago

Verified bfb1e012

server : disable speculative decoding for SWA models (#13970)

ggerganov committed 129 days ago

Verified 36375762

metal : use F32 accumulators in FA kernels (#13975)

ggerganov committed 129 days ago

Verified ea394d7a

gemma : more consistent attention scaling for v2 and v3 (#13951)

ggerganov committed 129 days ago

Verified 5582c49c

`server`: update deepseek reasoning format (pass reasoning_content as diffs) (#13933)

ochafik committed 129 days ago

Verified c9bbc779

mtmd : fix memory leak in mtmd_helper_eval_chunk_single (#13961)

ngxson committed 129 days ago

Verified bfd32279

cmake : Handle mixed-case 'Power' strings in POWER CPU detection (#13966)

shalinib-ibm committed 129 days ago

Verified 093e3f1f

sycl: quantize and reorder the input to q8_1 when reorder is enabled (#13826)

AD2605 committed 130 days ago

Verified 663445b0

gguf: fix failure on version == 0 (#13956)

JohannesGaessler committed 130 days ago

Verified 7675c555

convert : fix nomic-bert-moe mask token (#13757)

CISC committed 130 days ago

Verified 5e1c3aed

convert : fix vocab padding code for bert models (#13954)

CISC committed 130 days ago

Verified c496fe0b

ggml: check if non-native endian model is being loaded (#13943)

taronaeo committed 130 days ago

Verified e57bb87c

ggerganov committed 130 days ago

f3a4b165

vulkan : Remove unexpected ; (ggml/1253)

dg0yt committed 130 days ago

108009f5

cmake : Fix broken CMake error messages (ggml/1252)

dg0yt committed 130 days ago

d337252a

ggml : remove ggml_graph_import and ggml_graph_export declarations (ggml/1247)

rgerganov committed 130 days ago

af6f91db

sync : whisper.cpp (ggml/1250)

ggerganov committed 130 days ago

a7b8d35f

ggml : install dynamic backends (ggml/1240)

rgerganov committed 130 days ago

6eba72b7

ggml : Print backtrace on uncaught C++ exceptions (ggml/1232)

danielzgtg committed 130 days ago

fedf034a

readme : update bindings (#13950)

ddh0 committed 131 days ago

Verified 8726392d

parallel : fix n_junk == 0 (#13952)

ggerganov committed 131 days ago

Verified c0462171

kv-cache : split implementation in separate sources (#13920)

ggerganov committed 131 days ago

Verified 0fc16b42

threading: support for GGML_SCHED_PRIO_LOW, update thread info on Windows to avoid throttling (#12995)

max-krasnyansky committed 131 days ago

Verified 053b1539

docs : Note about necessity of having libcurl installed for standard build. (#13945)

jpodivin committed 131 days ago

Verified b3a89c3d

server: allow unclosed thinking tags (#13931)

ochafik committed 131 days ago

Verified e15898d1

llama : deprecate explicit kv_self defrag/update calls (#13921)

ggerganov committed 131 days ago

Verified 803f8baf

llama : use n_swa + n_ubatch cells for SWA cache (#13833)

ggerganov committed 131 days ago

Verified 3600cc28

webui : Replace alert and confirm with custom modals. (#13711)

igardev committed 132 days ago

Verified c7e0a205

llama : auto-batch preparation (#13845)

ggerganov committed 132 days ago

Verified 3f55f781

Newer Older