Add computational graph core infrastructure

Pull request 'android-build' (#86) from android-build into master

Update adept config forwarder fallback in Android build script

Add BatchNorm folding into conv weights for inference in Python YOLO implementation

Fix winograd conv implementation make tile pointer vectors thread-local in parallel loop

Conv2d improvements: add depthwise and pointwise conv fast paths

Update adept config forwarder fallback in Android build script

Add BatchNorm folding into conv weights for inference in Python YOLO implementation

Fix winograd conv implementation make tile pointer vectors thread-local in parallel loop

Conv2d improvements: add depthwise and pointwise conv fast paths

Submit asynchronously on GPU devices without timeline semaphores

Add BatchNorm folding into conv weights for fused inference for YOLO

Replace scatter_strided with block memcpy in cat_tensors for cpu

Initial commit

Add Android cross-compilation support

Merge remote-tracking branch 'refs/remotes/origin/master'

Remove PyTorch references in comments

Add adaptive max pooling with argmax index tracking

Add adaptive average pooling 2d layer

Pull request 'fix/TSKDPT3-122-protobuf-includes' (#85) from fix/TSKDPT3-122-protobuf-includes into master

clang-format onnx_model.hpp and onnx_model.cpp

remove protobuf includes dependencies from ONNX public API

Pull request 'feature/TSKDPT3-118-RAII-wrapper' (#84) from feature/TSKDPT3-118-RAII-wrapper into master

implement VkCommandRecorder RAII wrapper class for commands recording