ggerganov/llama.cpp

Pull Requests Commits

ui : move catalog url to models-discover constants

allozaur committed 2 hours ago

9b5a2ba4

ui : move HF constants to constants and enums modules

allozaur committed 2 hours ago

83de1280

ui : follow MODEL_ID regex rename in huggingface service

allozaur committed 2 hours ago

308a9590

ui : add huggingface hub data layer

allozaur committed 2 hours ago

593f93e7

ui : lowercase reasoning capability and use sidecar enums in tests

allozaur committed 2 hours ago

f286815f

ui : address review on model id parsing

allozaur committed 2 hours ago

67a429e7

ui : parse sidecar types in model ids

allozaur committed 2 hours ago

434927bc

ui : add format:files script for formatting changed files only

allozaur committed 2 hours ago

c1a7930d

ui : rename model API types to mirror endpoints

allozaur committed 2 hours ago

115a0d1c

ui : remove dead ChatFormActionAddMcpSubmenu component

allozaur committed 2 hours ago

4e7e76a2

metal : add fa-vec tunings for M1 (#28078)

nikwen committed 3 hours ago

Verified 5d4a3be2

CUDA: extend MOE fusion to specdec, earlier MOE glu fusion and topk-router fusion were restricted to 1 token (#27621)

ynankani committed 3 hours ago

Verified 41ef91f7

sycl : Enhance to get the free memory of Intel GPU (#27968)

arthw committed 4 hours ago

Verified a32af33d

ci : add check for unzip (#28082)

CISC committed 4 hours ago

Verified 580e88d8

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

wjinxu committed 5 hours ago

Verified 662a0b01

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

SimonTeixidor committed 5 hours ago

Verified 2cdae802

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

fairydreaming committed 6 hours ago

Verified 557614e0

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

0cc4m committed 9 hours ago

Verified daef7b68

hexagon: fix CPY fence bug (#28033)

yshsharke committed 20 hours ago

Verified 9723942a

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

ring2003 committed 20 hours ago

Verified bd55e6aa

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

hmirin committed 21 hours ago

Verified a7cc83bb

ggml : fix ggml_backend_buft_get_alloc_size() guard (#28038)

ggerganov committed 21 hours ago

Verified 6d1479c1

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

ServeurpersoCom committed 23 hours ago

Verified 62acc89c

ggml: add SWIGLU_CLAMP (#27930)

am17an committed 23 hours ago

Verified 0190529e

llama: improve TENSOR_READ_LAZY handling (#27837)

ngxson committed 23 hours ago

Verified 25781383

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

ServeurpersoCom committed 1 day ago

Verified f1793c1c

hip: tune rdna 3 mmq config (#26284)

itterative committed 1 day ago

Verified 0b5be7e4

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

LunalFresh committed 1 day ago

Verified e4221480

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

JamePeng committed 1 day ago

Verified cc231cb0

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

ggerganov committed 1 day ago

Verified bebc9350

Older