ngxson/llama.cpp

Pull Requests Commits

vulkan: Fix crash when FP16 mul_mat accumulation is not supported (#16796)

rillomas committed 163 days ago

Verified 2976b037

vulkan: fix shmem overrun in mmq id shader (#16873)

0cc4m committed 163 days ago

Verified d2a2673d

ggml-hexagon: respect input size when getting/setting tensor data (#16836)

l3utterfly committed 163 days ago

Verified 13002a08

ci : enable free-disk-space on cuda docker build (#16877)

CISC committed 163 days ago

Verified 6eb208d1

opencl: fix boundary handling for mul_mm (#16875)

lhez committed 163 days ago

Verified 9984cbb6

convert : update transformers requirements (#16866)

RodriMora committed 163 days ago

Verified ce18efea

server : bump request URI max length to 32768 (#16862)

chansikpark committed 164 days ago

Verified 16724b5b

server : remove n_past (#16818)

ggerganov committed 164 days ago

Verified b52edd25

cpu: introduce chunking for repack matmuls and enable matmul-id chunking on ARM64 (#16833)

max-krasnyansky committed 164 days ago

Verified 517b7170

common: fix typo in cli help text (#16864)

sbera77 committed 164 days ago

Verified 835e918d

model: add support for qwen3vl series (#16780)

JJJYmmm committed 164 days ago

Verified d261223d

cpu: introduce chunking for flash attention (#16829)

max-krasnyansky committed 164 days ago

Verified dcca0d3a

model: Add support for CogVLM model (#15002)

Tianyue-Zhao committed 164 days ago

Verified bacddc04

cuda : fix argsort with 64k+ rows (#16849)

CISC committed 164 days ago

Verified 229bf686

llama : use std::abs instead of abs (#16853)

kaetemi committed 164 days ago

Verified d7395115

vulkan: Handle argsort with a large number of rows (#16851)

jeffbolznv committed 164 days ago

Verified 052df28b

Hide latency of bias and gate-loading (#16847)

ORippler committed 164 days ago

Verified 8b11deea

vulkan: Fuse rope+set_rows (#16769)

jeffbolznv committed 164 days ago

Verified b9ce9401

llama: fix ASAN error with M-RoPE (#16848)

ngxson committed 165 days ago

Verified 3464bdac

llama: store mrope data in KV cell (#16825)

ngxson committed 165 days ago

Verified e3af5563

vulkan: Update topk_moe fusion to handle gpt's late softmax (#16656)

jeffbolznv committed 165 days ago

Verified 10fcc412

Vulkan MMQ Integer Dot Refactor and K-Quant support (#16536)

0cc4m committed 165 days ago

Verified bcf5bda6

Hexagon Op queue & dispatch optimizations (#16820)

max-krasnyansky committed 165 days ago

Verified 3eb2be1c

CUDA: use fastdiv in set-rows (#16834)

am17an committed 165 days ago

Verified e41bcce8

vendor : sync minja (#16500)

CISC committed 165 days ago

Verified 144a4ce8

vulkan: Call ggml_vk_buffer_write_2d from ggml_vk_buffer_copy (#16793)

jeffbolznv committed 165 days ago

Verified f549b000

CUDA: Fix bug in topk-moe for gpt-oss (#16821)

am17an committed 165 days ago

Verified 9a3ea685

sycl: add RMS_NORM_BACK operation support (#16808)

YaelLogic committed 165 days ago

Verified 338074c3

cuda: add SET operation support (#16804)

YaelGitAccount committed 166 days ago

Verified 851553ea

memory : remove KV cache size padding (#16812)

ggerganov committed 166 days ago

Verified 85a7d867

Newer Older