ggml-org/llama.cpp

Pull Requests Commits

skill: note about model-specific CLI arguments + testings (#29808)

ngxson committed 26 days ago

Verified 68e79bd8

ci: fix Fusion / metal by updating the qwen4exp baseline (#29812)

ServeurpersoCom committed 26 days ago

Verified e358d591

llama : clamp kpool re-pool bound to existing pools (#29805)

ggerganov committed 26 days ago

Verified 81e39ad3

hexagon: shared strided DMA copy for CPY and CONCAT, any-dim CONCAT via DMA (#29685)

njsyw1997 committed 26 days ago

Verified dcd387a4

server: return HTTP 400 for invalid embedding requests (#29060)

SamMalayek committed 26 days ago

Verified d775ebf3

convert : write Gemma embedding scale for DFlash drafts (#29802)

kabu1204 committed 26 days ago

Verified 2b36825c

cuda : route sm70 to the Turing MMVQ nwarps table (#29753)

tkittich committed 26 days ago

Verified 42d95816

metal : release temporary private transfer buffers (#29777)

mvanlamz committed 26 days ago

Verified 13b4d713

webgpu: add bfloat16 support for MUL_MAT/MUL_MAT_ID/GET_ROWS- #29358 (#29358)

yomaytk committed 26 days ago

Verified 4b1622af

llama : fix invalid assert in recurrent memory (#29799)

ggerganov committed 26 days ago

Verified 869034b4

CUDA: Handle compute type for NVFP4 on cublass path (#29173)

ynankani committed 26 days ago

Verified b56f34ab

Qwen4Exp: add MTP (#29761)

am17an committed 26 days ago

Verified c061df19

llama: fix qwen4exp (#29751)

am17an committed 26 days ago

Verified 66e0c17e

CUDA: Make CCCL configurable + pin it to 3.4.3 for CI jobs (#29792)

ORippler committed 27 days ago

Verified 76776785

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

ngxson committed 27 days ago

Verified 552f18f9

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

ggerganov committed 27 days ago

Verified 5503b04b

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

xarillian committed 27 days ago

Verified def4d406

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

praneshgo committed 27 days ago

Verified 32dd62ee

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

IMbackK committed 27 days ago

Verified f11d642a

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

max-krasnyansky committed 27 days ago

Verified 3aa0ce9b

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

pradeeptrgit committed 27 days ago

Verified b0aca3c6

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

e-mon committed 27 days ago

Verified b8f96c3e

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

lhez committed 27 days ago

Verified 3ec4df42

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

tokinasin committed 27 days ago

Verified db33d3cb

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

kushalgarg101 committed 27 days ago

Verified 7dad6db8

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

ggerganov committed 27 days ago

Verified 2232bc8b

llama-bench : fix docs (#29464)

mairp committed 27 days ago

Verified 79625e05

docs : refresh CPU ops support matrix (#29666)

CaramelizedCUDA committed 27 days ago

Verified 66bcc277

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

kh0pper committed 27 days ago

Verified 10f340d1

webgpu: fix SSM_SCAN binding aliasing (#29750)

yomaytk committed 27 days ago

Verified 0c1e5709

Newer Older