ggml-org/llama.cpp

Pull Requests Commits

cont : disable tensor path with kv_rows

ggerganov committed 15 days ago

Verified 5982b75d

cont : add non-vec support

ggerganov committed 15 days ago

Verified b0f0c211

metal : support flash_attn_ext_rows (kv_rows) in vec kernels

ggerganov committed 15 days ago

Verified 5cc97ed4

llama + CUDA: flash_attn_ext_rows (fix unified kv for multi-seq)

am17an committed 15 days ago

56d73a89

llama: fix k-pool scatter data race on shared sequences (#29994)

ServeurpersoCom committed 15 days ago

Verified 43fe9c64

hexagon: ssm-conv updates (#29971)

tboinovski1 committed 15 days ago

Verified 5e03bdd8

hexagon: add pool op support (#29995)

aparmp-quic committed 15 days ago

Verified 50569eb8

ci : add 1accel label [no ci] (#30016)

CISC committed 15 days ago

Verified 7049ff0c

ci : skip container re-tagging when Require Docker is disabled (#30008)

ggerganov committed 15 days ago

Verified c2503049

hexagon: matmul and flash-atten scalability updates (#29974)

max-krasnyansky committed 15 days ago

Verified 8345f333

llama.cpp : bump version to 0.6.0 (#29997)

ggerganov committed 15 days ago

Verified d8123504

common, server : report model input/output modalities in GET /models (#29987)

angt committed 15 days ago

Verified 4d60b4d0

ggerganov committed 15 days ago

c06f8416

ggml : bump version to 0.26.0 (ggml/1652)

ggerganov committed 15 days ago

f05c8b27

CUDA: make the alloc_deps check batch independent (#29986)

am17an committed 15 days ago

Verified e117148a

vulkan: fix Flash Attention shmem write out of bounds (#29988)

0cc4m committed 15 days ago

Verified 6c59c400

vulkan: revert mul_mat_id tile selection PR #29182 (#29936)

virajwad committed 15 days ago

Verified 3c9e747f

cuda: use the vector lightning indexer kernel on MUSA (#29990)

ServeurpersoCom committed 15 days ago

Verified b809b886

ci : add "Require Docker" flag to make-release workflow (#29989)

ggerganov committed 15 days ago

Verified 994e8f22

webui: Use toLocaleString() format consistently across chat message statistics (#27990)

grivera64 committed 15 days ago

Verified 9d853bb3

ci : disable failing test on virtual Metal device (#29993)

ggerganov committed 15 days ago

Verified 8f9ae20c

server: support vision input for Clef (#29969)

ngxson committed 15 days ago

Verified 9871df59

CUDA: Optimize accumulation in mmq for NVFP4 type (#29857)

kmorennv committed 15 days ago

Verified 8b2fbaf3

ci : disable unused qemu in docker build (#29984)

CISC committed 16 days ago

Verified 2ed93db4

server: reject partial media truncation (#24076)

he-yufeng committed 16 days ago

Verified 8e164219

vulkan: fix stale prealloc_y reuse across flash attention and soft_max (#29591)

fxgsell committed 16 days ago

Verified 806eee98

vulkan: sparse flash attention for quantized K/V (#29639)

fxgsell committed 16 days ago

Verified b3daa077

llama : fix unexpected graph reallocation in the k-pool models (#29958)

ggerganov committed 16 days ago

Verified c173a53b

kv-cache: fix restoring mismatched KV cache rotation by saving exact rotation metadata (#28498)

eapache committed 16 days ago

Verified 21079106

ci : winget urls must be separate strings (#29978)

CISC committed 16 days ago

Verified e5983d67

Older