ggml-org/llama.cpp

Pull Requests Commits

spec : fuse the DFlash encoder into the KV cache injection (#27310)

wjinxu committed 15 days ago

Verified 662a0b01

vulkan: tune mat-vec rows for batched inference on Strix Halo (#27909)

SimonTeixidor committed 15 days ago

Verified 2cdae802

ggml : add MUL_MAT to the list of ops that may need additional memory (for WebGPU) (#28071)

fairydreaming committed 15 days ago

Verified 557614e0

vulkan: top_k radix select for k >= 1024 for Qwen 3.8 Flash Next (#28032)

0cc4m committed 15 days ago

Verified daef7b68

hexagon: fix CPY fence bug (#28033)

yshsharke committed 15 days ago

Verified 9723942a

metal : add remaining Q4_1/Q5_0/Q5_1 fa-vec tunings for M2 (#28017)

ring2003 committed 15 days ago

Verified bd55e6aa

rpc: avoid serializing buffers from other servers (#26500)

hmirin committed 15 days ago

Verified a7cc83bb

ggml : fix ggml_backend_buft_get_alloc_size() guard (#28038)

ggerganov committed 15 days ago

Verified 6d1479c1

kv-cells: stop the sequence scan once all sequences are seen (#28011)

ServeurpersoCom committed 16 days ago

Verified 62acc89c

ggml: add SWIGLU_CLAMP (#27930)

am17an committed 16 days ago

Verified 0190529e

llama: improve TENSOR_READ_LAZY handling (#27837)

ngxson committed 16 days ago

Verified 25781383

CUDA: use the fast mm_ids_helper path for any n_expert_used (#27978)

ServeurpersoCom committed 16 days ago

Verified f1793c1c

hip: tune rdna 3 mmq config (#26284)

itterative committed 16 days ago

Verified 0b5be7e4

hip : optimize Q2_0 dot-product path for gfx1201 (#26753)

LunalFresh committed 16 days ago

Verified e4221480

dflash: pass missing NVFP4 scales to attention operations (#28000)

JamePeng committed 16 days ago

Verified cc231cb0

common: rename --tensor-read-lazy to --lazy-mode, add -lzm shorthand (#27969)

ggerganov committed 16 days ago

Verified bebc9350

ggml : add ggml_backend_op_alloc_size_may_expand, use it in RPC (#27960)

ggerganov committed 16 days ago

Verified 73f56d10

rpc: fix apple rdma error spew on teardown (#27908)

ryan5rdx committed 16 days ago

Verified 742347b2

metal: add fa-vec tunings for M3 Ultra (#27999)

ngladitz committed 16 days ago

Verified 093adb24

metal : Add fa-vec tuning for M3 Pro (#27963)

addianto committed 16 days ago

Verified b8b743c3

vendor : update cpp-httplib to 0.54.0 (#27919)

cabelo committed 16 days ago

Verified dc7aecf7

rpc : fix pre-rdma macOS versions (#27815)

ryan5rdx committed 16 days ago

Verified 2bf04151

hexagon: support for device discovery and create sessions on demand (#27785)

max-krasnyansky committed 16 days ago

Verified 9e54e687

sycl: split long rows in TOP_K instead of one work-group per row (#27847)

Titaniumtown committed 16 days ago

Verified 370cb12e

metal : fix null-pipeline crash for F16 src1 mul_mat/mul_mat_id (#25648)

QuintinShaw committed 16 days ago

Verified d882575c

memory : copy Hadamard matrix to k_rot tensor only if it has buffer assigned to prevent crashes during context shift of unquantized K cache (#27967)

fairydreaming committed 16 days ago

Verified bdf39551

ggml: allow passing alloc dependencies in graph_optimize (#27301)

am17an committed 16 days ago

Verified 57291f26

metal : add fa-vec tunings for M2 (#27940)

ring2003 committed 16 days ago

Verified c589f0ed

opencl: use a better matmul path on two Adreno GPU generations (#27640)

wanghqc committed 16 days ago

Verified c841aeeb

metal : assert shared memory padding (#27951)

ggerganov committed 17 days ago

Verified 3173a564

Newer Older