ggerganov/llama.cpp

Pull Requests Commits

opencl: add depthwise conv2d (#30900)

kurquhar committed 55 minutes ago

Verified 970b51ce

mtmd : add MiniCPM-OCR 1.0 support (#30979)

tc-mb committed 1 hour ago

Verified a08ab75a

clef : include <algorithm> for std::find and std::copy (#30981)

kleinbem committed 2 hours ago

Verified 19ab5a5e

gaugarg-nv committed 4 hours ago

Verified 36cbf415

glm5 : optimize inputs (#30932)

ggerganov committed 4 hours ago

Verified 5c19a5a7

model : fix qwen4exp layer inputs for DFlash drafters (#30927)

ggerganov committed 4 hours ago

Verified d9ee007a

Multi-column on Intel XMX matrix engines refactored (#29864)

budiony committed 5 hours ago

Verified d81d0d8f

vulkan: replace std::aligned_alloc with posix_memalign for termux compatibility (#30828)

bill88t committed 6 hours ago

Verified d7725699

nemotron : support non-standard GQA, BPE tokenizer and domyn-small pre-type (#23343)

martin-cimmino committed 7 hours ago

Verified 0693dbfb

server : support typed content input for RERANK endpoint (#29625)

timothywang21 committed 7 hours ago

Verified a5d22276

ci : disable slow WebGPU tests (#30963)

ggerganov committed 7 hours ago

Verified 5f5acfe2

ci : be able to reconfigure HTTP port for server pytest workers (#30972)

ggerganov committed 7 hours ago

Verified 1f4c2eea

sycl: fix CONV_2D wrong results for non-contiguous input (#30363)

mctylr-gh committed 10 hours ago

Verified 7170cad4

vocab : add clean_spaces GGUF key (#30940)

tkran2 committed 10 hours ago

Verified d88040c7

vocab : use Tekken pre-tokenizer for pixtral rather than the Llama 3 one (#30960)

tkran2 committed 11 hours ago

Verified e197cef4

models : Record layer inputs for all granite models (#30957)

gabe-l-hart committed 11 hours ago

Verified bf039c84

kleidiai : add Q8_0 SVE dotprod GEMV and SVE i8mm GEMM kernels (#30808)

chaxu01 committed 11 hours ago

Verified dfeea520

sycl: merge the fp16 and fp32 dequant switches (#30740)

Titaniumtown committed 13 hours ago

Verified e3797b18

webgpu : add Q2_0 support (#30914)

teddygood committed 13 hours ago

Verified 0b1281ae

dflash : load optional FC input scale (#30945)

hthadicherla committed 13 hours ago

Verified 4912ac24

ui: Add "Show tool call in progress" setting (#30681)

allozaur committed 13 hours ago

Verified 84746c5c

lora : support prism.hadamard folded weights (#30910)

vieduy committed 13 hours ago

Verified 0682a576

ui : key draft sidecar badges by model instead of repo (#30919)

allozaur committed 13 hours ago

Verified a6b83d42

opencl: add F32-to-I32 ARGMAX (#30897)

kurquhar committed 17 hours ago

Verified c7fc5094

models : add Flint model (FlintForCausalLM) (#30928)

alexey-rsi committed 21 hours ago

Verified a37cc461

jinja : fix namespace attributes shadowed by dict methods (#30941)

tdakhran committed 21 hours ago

Verified 99fb9971

model: Add support for SenseNova U1 model (#28919)

Maphist0 committed 22 hours ago

Verified 7775ffb7

llama: create new fused op for QSA mask (#30640)

am17an committed 1 day ago

Verified d454a204

sycl: q8 fast dequant (#30775)

92es committed 1 day ago

Verified 0756e83f

jinja: support hexadecimal string escapes (#30929)

alexey-rsi committed 1 day ago

Verified 9febe63d

Older