ngxson/llama.cpp

Pull Requests Commits
vulkan: fix shmem overrun in mmq id shader (#16873)
0cc4m committed 319 days ago
Verified d2a2673d
ggml-hexagon: respect input size when getting/setting tensor data (#16836)
l3utterfly committed 319 days ago
Verified 13002a08
ci : enable free-disk-space on cuda docker build (#16877)
CISC committed 320 days ago
Verified 6eb208d1
opencl: fix boundary handling for mul_mm (#16875)
lhez committed 320 days ago
Verified 9984cbb6
convert : update transformers requirements (#16866)
RodriMora committed 320 days ago
Verified ce18efea
server : bump request URI max length to 32768 (#16862)
chansikpark committed 320 days ago
Verified 16724b5b
server : remove n_past (#16818)
ggerganov committed 320 days ago
Verified b52edd25
cpu: introduce chunking for repack matmuls and enable matmul-id chunking on ARM64 (#16833)
max-krasnyansky committed 320 days ago
Verified 517b7170
common: fix typo in cli help text (#16864)
sbera77 committed 320 days ago
Verified 835e918d
model: add support for qwen3vl series (#16780)
JJJYmmm committed 320 days ago
Verified d261223d
cpu: introduce chunking for flash attention (#16829)
max-krasnyansky committed 320 days ago
Verified dcca0d3a
model: Add support for CogVLM model (#15002)
Tianyue-Zhao committed 320 days ago
Verified bacddc04
cuda : fix argsort with 64k+ rows (#16849)
CISC committed 320 days ago
Verified 229bf686
llama : use std::abs instead of abs (#16853)
kaetemi committed 320 days ago
Verified d7395115
vulkan: Handle argsort with a large number of rows (#16851)
jeffbolznv committed 320 days ago
Verified 052df28b
Hide latency of bias and gate-loading (#16847)
ORippler committed 321 days ago
Verified 8b11deea
vulkan: Fuse rope+set_rows (#16769)
jeffbolznv committed 321 days ago
Verified b9ce9401
llama: fix ASAN error with M-RoPE (#16848)
ngxson committed 321 days ago
Verified 3464bdac
llama: store mrope data in KV cell (#16825)
ngxson committed 321 days ago
Verified e3af5563
vulkan: Update topk_moe fusion to handle gpt's late softmax (#16656)
jeffbolznv committed 321 days ago
Verified 10fcc412
Vulkan MMQ Integer Dot Refactor and K-Quant support (#16536)
0cc4m committed 321 days ago
Verified bcf5bda6
Hexagon Op queue & dispatch optimizations (#16820)
max-krasnyansky committed 321 days ago
Verified 3eb2be1c
CUDA: use fastdiv in set-rows (#16834)
am17an committed 321 days ago
Verified e41bcce8
vendor : sync minja (#16500)
CISC committed 321 days ago
Verified 144a4ce8
vulkan: Call ggml_vk_buffer_write_2d from ggml_vk_buffer_copy (#16793)
jeffbolznv committed 321 days ago
Verified f549b000
CUDA: Fix bug in topk-moe for gpt-oss (#16821)
am17an committed 321 days ago
Verified 9a3ea685
sycl: add RMS_NORM_BACK operation support (#16808)
YaelLogic committed 321 days ago
Verified 338074c3
cuda: add SET operation support (#16804)
YaelGitAccount committed 322 days ago
Verified 851553ea
memory : remove KV cache size padding (#16812)
ggerganov committed 322 days ago
Verified 85a7d867
llama-bench : clarify benchmarked parts of the computation (#16823)
ggerganov committed 322 days ago
Verified a8ca18b4
Terms Of Use Privacy Policy Refund Policy Impressum