ggml-org/llama.cpp

Pull Requests Commits

update offload_op logic

0cc4m committed 12 hours ago

ca3481f3

use transfer_queue for async transfers on AMD, synchronize with timeline semaphore

0cc4m committed 13 hours ago

5abb7d55

vulkan: fix and enable cpy_tensor_async function

0cc4m committed 17 hours ago

6943f830

ggml : fix AMX and add batched support (#19925)

angt committed 1 day ago

Verified 4e76d24f

vulkan: fix fp16 Flash Attention on Windows AMD RDNA2 and below (#19921)

0cc4m committed 1 day ago

Verified 723c7106

mtmd : fix padding of n_tokens (#19930)

ggerganov committed 1 day ago

Verified 37964f44

server : fix ctx checkpoint restore logic (#19924)

ggerganov committed 1 day ago

Verified 01cd448b

kv-cache : fix can_shift() check to take into account M-RoPE (#19928)

ggerganov committed 1 day ago

Verified 99bd67c9

llama: Add option to merge gate and exp weights (#19139)

am17an committed 1 day ago

Verified b68d7516

ggml-virtgpu: improve the reliability of the code (#19846)

kpouget committed 1 day ago

Verified ffaafde1

server: fix load-on-startup not respected in ini file (#19897)

drrros committed 1 day ago

Verified efba35a8

jinja : correct default size for string slices (#19913)

EZForever committed 1 day ago

Verified 9b62913b

model : add Jina Embeddings v5 Nano (partial EuroBERT) support (#19826)

maximilianwerk committed 1 day ago

Verified 66287bda

gguf : avoid too many file size calls (#19919)

ggerganov committed 1 day ago

Verified 1ca3d1de

server : fix typo in server README.md (#19900)

yggdrasil75 committed 1 day ago

Verified bd723005

support permuted, remove check s0/s10 (#19889)

arthw committed 2 days ago

Verified 2943210c

vulkan: check for memory overlap before doing fusion (#19768)

jeffbolznv committed 2 days ago

Verified 3769fe6e

common : add more aliases for sampler CLI params (#19797)

ddh0 committed 2 days ago

Verified 832aa947

ci : update the ROCm/HIP toolchain versions [no ci] (#19891)

slojosic-amd committed 2 days ago

Verified 3af34b9f

server : enable multi-modal prompt caching (#19877)

ggerganov committed 2 days ago

Verified f20469d9

server : support multi-modal context checkpoints (#19849)

ggerganov committed 2 days ago

Verified d7d826b3

scripts: update corpus of compare-logprobs (#19326)

ngxson committed 2 days ago

Verified c747294b

ci : update Windows ROCm build to 26.Q1 [no ci] (#19810)

superm1 committed 2 days ago

Verified 8fdf269d

gguf : fix ftell/fseek for Windows (#19870)

aldehir committed 2 days ago

Verified a96a1120

models : fix graph splits (#19866)

ggerganov committed 3 days ago

Verified 24464195

server: fix query params lost when proxying requests in multi-model router mode (#19854)

ServeurpersoCom committed 3 days ago

Verified 47eb12b9

ggml/gguf : prevent integer overflows (#19856)

ggerganov committed 3 days ago

Verified 418dea39

model : update label for LFM2-24B-A2B (#19848)

tdakhran committed 3 days ago

Verified da426cb2

server : support max_completion_tokens request property (#19831)

rgerganov committed 3 days ago

Verified c830f99c

Vulkan Scalar Flash Attention Refactor (#19625)

0cc4m committed 3 days ago

Verified aa6f918c

Older