ggerganov/llama.cpp

Pull Requests Commits

memory : handle kv_unified for hybrid models

compilade committed 33 days ago

e549515c

opencl: fix adreno compiler detection logic (#15029)

lhez committed 33 days ago

Verified 5c0eb5ef

CUDA: use mma FA kernel for gqa > 4 on RTX 4000 (#15035)

JohannesGaessler committed 33 days ago

Verified 03d46982

cuda: make im2col a little faster (#15025)

leejet committed 33 days ago

Verified 3303c19b

kv-cache : skip alignment of n_stream in kv-cache log msg [no ci] (#15040)

danbev committed 33 days ago

Verified 4fdea540

llama : enable LLAMA_SET_ROWS=1 by default (#14959)

ggerganov committed 33 days ago

Verified a4569c41

cuda, sycl : fix batched gemm when ne02 == 1 && ne03 > 1 (#15038)

ggerganov committed 33 days ago

Verified 15e92fd3

ci : check that pre-tokenizer hashes are up-to-date (#15032)

CISC committed 33 days ago

Verified 2bf3fbf0

convert : fix Qwen3-Embedding pre-tokenizer hash (#15030)

iamlemec committed 33 days ago

Verified 711d5e6f

chat : fix multiple tool_calls on hermes-2-pro (#14962)

jhen0409 committed 33 days ago

Verified f738989d

vulkan: coopmat2 mul_mat optimizations (#14934)

jeffbolznv committed 33 days ago

Verified 4cb208c9

llama-bench: rename DB table name from test to llama_bench (#15003)

yeahdongcn committed 33 days ago

Verified 3025b621

vulkan: Support ne[3]>1 in noncontig matrix-vector multiply (#15015)

jeffbolznv committed 33 days ago

Verified ec0b1880

model : support Qwen3-Embedding (#15023)

iamlemec committed 33 days ago

Verified 339bd026

server: enable token array inputs for OAI API (#15001)

JohannesGaessler committed 33 days ago

Verified f9062755

vulkan: optimizations for direct convolution (#14933)

jeffbolznv committed 33 days ago

Verified a9f7541e

CUDA: fix MMQ nwarps for AMD with warp_size==32 (#15014)

JohannesGaessler committed 34 days ago

Verified 9c35706b

vendor : update vendored copy of google/minja (#15011)

l-austenfeld committed 34 days ago

Verified c76b420e

model : add hunyuan dense (#14878)

stevenkuang-tencent committed 34 days ago

Verified 0f5ccd6f

opencl: add f16 for `add`, `sub`, `mul`, `div` (#14984)

lhez committed 34 days ago

Verified 1c872f71

ggml : Q2k interleaving implementation - x86/x64 SIMD (#14373)

Srihari-mcw committed 34 days ago

Verified baad9488

graph : fix equal_seq() check (#14986)

ggerganov committed 35 days ago

Verified ba42794c

docker : add cann build pipline (#14591)

diannaojiang committed 35 days ago

Verified 2860d479

compare-commits.sh: support both llama-bench and test-backend-ops (#14392)

yeahdongcn committed 35 days ago

Verified 484b2091

quantize : skip tensor override when in fallback mode (#14995)

EAddario committed 35 days ago

Verified daf2dd78

llama : add simple option to enable CPU for MoE weights (--cpu-moe) (#14992)

slaren committed 35 days ago

Verified a06ed5fe

Fix params bug in diffusion example (#14993)

am17an committed 35 days ago

Verified 78452405

llama : allow other bufts when overriding to CPU, add --no-repack option (#14990)

slaren committed 35 days ago

Verified d6818d06

Vulkan: Fix minor debug mode issues (#14899)

0cc4m committed 35 days ago

Verified e08a9882

mtmd : support MiniCPM-V 4.0 (#14983)

tc-mb committed 35 days ago

Verified 952a47f4

Older