ncnn is a high-performance neural network inference framework optimized for the mobile platform
android
ios
caffe
deep-learning
neural-network
mxnet
tensorflow
vulkan
keras
inference
pytorch
artificial-intelligence
simd
darknet
arm-neon
high-preformance
ncnn
onnx
mlir
-
Updated
Apr 20, 2021 - C++
The npcomp project has a really nice feature where they get a hash of the actual github workflow content, and use that in the cache key for the LLVM/MLIR build. This lets you tweak build settings in that file without munging up the cache key every time, so we should probably do something similar.
https://github.com/llvm/mlir-npcomp/blob/28a0f02746f4a1315d8e17d9560f7ae1c255236c/.github/workflows