ggerganov/llama.cpp

Pull Requests Commits

ggml-cuda: gdn use shared mem for HIP (#20366)

IMbackK committed 21 minutes ago

Verified 5f91b1d5

cuda/hip: fix loop unrolling in ssm-conv (#20369)

IMbackK committed 23 minutes ago

Verified 9ef7523e

Fix agentic mcp image single model (#20339)

ServeurpersoCom committed 56 minutes ago

Verified 00de6153

vendor : update cpp-httplib to 0.37.0 (#20207)

cabelo committed 2 hours ago

Verified e1a39999

vendor : update miniaudio to 0.11.25 (#20209)

cabelo committed 2 hours ago

Verified 4f2f0a16

fix op rope, add rope_back (#20293)

arthw committed 3 hours ago

Verified 0cec84f9

fix for failed UT case: ACC, L2_NORM, UPSCALE, fused_glu, unary (#20283)

arthw committed 3 hours ago

Verified b2e1427c

model : qwen3vl reranker text support (#20332)

ViniciosLugli committed 6 hours ago

Verified 4d99d450

llama-quant : correct `n_attention_wv` usage (#20357)

ddh0 committed 9 hours ago

Verified 10e5b148

ggml : bump RPC version (#20330)

ggerganov committed 9 hours ago

Verified 90b27318

ggml webgpu: faster normal quant and some k-quant matrix operations, better shader parameter handling (#20173)

reeselevine committed 13 hours ago

Verified aa2d278a

Reduce level of content parser warning message to avoid log spam on non-debug verbosity (#20347)

pwilkin committed 15 hours ago

Verified 6c770d16

examples : fix empty items in json_schema_to_grammar.py [no ci] (#19968)

RayXu14 committed 15 hours ago

Verified 8d880ac0

docs: update CPU backend ops to mark POOL_1D as supported (#20304)

a3894281 committed 15 hours ago

Verified 0f1e9d14

models : fix assert in mamba2 (cont) (#20335)

ggerganov committed 16 hours ago

Verified 1274fbee

server : make 2 checkpoints near the end of the prompt (#20288)

ggerganov committed 16 hours ago

Verified a7b3dee7

common : fix incorrect uses of stoul (#20313)

CISC committed 18 hours ago

Verified ec947d2b

kleidiai : support for concurrent sme and neon kernel execution (#20070)

chaxu01 committed 22 hours ago

Verified 0cd4f472

ggml-cpu: add RVV repack GEMM and GEMV for quantization types (#19121)

taimur-10x committed 22 hours ago

Verified af237f30

metal: handle command buffer failures gracefully in synchronize (#20306)

JulianPscheid committed 22 hours ago

Verified 1a5631be

llama-quant : fail early on missing imatrix, refactor type selection, code cleanup (#19770)

ddh0 committed 23 hours ago

Verified 1dab5f5a

common: consolidate PEG string parsers (#20263)

aldehir committed 1 day ago

Verified c96f608d

model: fix step3.5 n_rot (#20318)

ngxson committed 1 day ago

Verified 0842b9b4

llama: dynamic head_dim and n_rot for SWA (#20301)

ngxson committed 1 day ago

Verified 59db9a35

server: Parse port numbers from MCP server URLs in CORS proxy (#20208)

eapache committed 1 day ago

Verified 23fbfcb1

metal : extend mul_mv_ext to BF16, Q2_K, Q3_K (#20250)

arkavo-com committed 1 day ago

Verified e22cd0aa

server : fix checkpoints n_tokens calculation (#20287)

ggerganov committed 1 day ago

Verified 96cfc499

metal : add upscale (#20284)

ggerganov committed 1 day ago

Verified ed0007aa

server : warn swa-full is not supported for non-SWA models (#20291)

ggerganov committed 1 day ago

Verified 344ee2a3

server : fix off-by-1 in server_tokens::size_up_to_pos() (#20279)

ggerganov committed 1 day ago

Verified d6e15564

Older