ggml-org/llama.cpp

Pull Requests Commits

qwen4exp: keep kq_mask input the same shape

am17an committed 23 days ago

Verified f2ef14ea

llama: fix qwen4exp

am17an committed 23 days ago

Verified 56a33d69

mtmd: cap max_image to ubatch for non_causal models (#29773)

ngxson committed 23 days ago

Verified 552f18f9

meta: clear inactive AllReduce shards with FILL, not SCALE (#29793)

ggerganov committed 23 days ago

Verified 5503b04b

jinja : skip copying loop scope unless a loop filter needs it (#29776)

xarillian committed 23 days ago

Verified def4d406

llama-mmap : avoid a second full-size copy of each tensor with direct-io (#29749)

praneshgo committed 23 days ago

Verified 32dd62ee

HIP: avoid treating CDNA as dgx spark for gqa_ratio 20 in fattn_mma dqk 576 (#29572)

IMbackK committed 23 days ago

Verified f11d642a

hex-workqueue: fix race condition in seqn getting out of sync with idx_read/write (#29785)

max-krasnyansky committed 23 days ago

Verified 3aa0ce9b

BLAS : Document AOCL-BLAS build and label the device AOCL-BLAS (#29640)

pradeeptrgit committed 23 days ago

Verified b0aca3c6

common : add LLM-jp-4.1 Harmony dialect handler (#29681)

e-mon committed 23 days ago

Verified b8f96c3e

opencl: mark vec subgroup bcast as supproted for Adreno E17 compiler (#29698)

lhez committed 23 days ago

Verified 3ec4df42

vocab : honor BOS/EOS settings for PLaMo-2 and PLaMo-3 (#29734)

tokinasin committed 23 days ago

Verified db33d3cb

llama-bench : fix verbosity filter to show GGML_LOG_ERROR (#28229)

kushalgarg101 committed 23 days ago

Verified 7dad6db8

metal : use bf16 math for mxfp4 mul-mat (#29770)

ggerganov committed 23 days ago

Verified 2232bc8b

llama-bench : fix docs (#29464)

mairp committed 23 days ago

Verified 79625e05

docs : refresh CPU ops support matrix (#29666)

CaramelizedCUDA committed 23 days ago

Verified 66bcc277

model : re-enable -sm tensor for qwen4exp (#28569)

kh0pper committed 23 days ago

Verified 10f340d1

webgpu: fix SSM_SCAN binding aliasing (#29750)

yomaytk committed 23 days ago

Verified 0c1e5709

ggml-opencl : replace alloca() with std::vector (#29765)

angt committed 23 days ago

Verified f7b384c1

cuda: guard the iq4_nl dequantize row kernel against short rows (#29683)

yeahdongcn committed 23 days ago

Verified f872b591

Hexagon: optimize ALLREDUCE with support for safe scatter mode (#29757)

ebateni committed 23 days ago

Verified a4d880fd

args: fix cli download mmproj arg (#28977)

pr3pony committed 23 days ago

Verified feb9a3d6

llama : preserve original batch order for speculative decoding layer inputs (#29019)

hthadicherla committed 23 days ago

Verified 4453b535

test-llama-archs : toggle causal_attn to catch graph shape changes (#29724)

sihanyu03 committed 23 days ago

Verified 4f31296a

convert : fix LoRA conversion crash for Qwen3.5 V-head reorder (#28324)

Swigler committed 23 days ago

Verified b016f461

llama: properly handle KV on training (#28520)

ngxson committed 23 days ago

Verified 81ff93ea

batch: migrate the rest of examples to llama_batch_ext (#29601)

ngxson committed 23 days ago

Verified 60e9cf7a

glm5-next: give dead indexer slots unique scatter rows (#29745)

ServeurpersoCom committed 23 days ago

Verified 05af0d2b

ggml/gguf : fix integer overflow (#29384)

apach301 committed 23 days ago

Verified 2149c00f

codeowners : remove former ZenDNN owner (#29747)

vishalMCE committed 23 days ago

Verified 876c75b1

Older