HM 2

by Yufeng Guo

install

by installation, follow the steps:


mkdir build && cd build
cmake ../
make
cp ./gra* ../hw2/
cp ./pytorch_apis.py ./hw2/

3 layer fully connected network:

with linear layer was implemented through our own CUDA kernel funtion

NVCC:11.8, Nvidia A100

conda create -n dgl python=3.7
conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.2 -c pytorch
pip install dgl-cu102==0.6.1

Name		Name	Last commit message	Last commit date
Latest commit History 42 Commits
__pycache__		__pycache__
build		build
data/MNIST/raw		data/MNIST/raw
hw2		hw2
hw3		hw3
pybind11 @ cca4c51		pybind11 @ cca4c51
.gitignore		.gitignore
.gitmodules		.gitmodules
CMakeLists.txt		CMakeLists.txt
LICENSE		LICENSE
Makefile		Makefile
README.md		README.md
csr.h		csr.h
dlpack.h		dlpack.h
generate_class_code.py		generate_class_code.py
generate_gp_apis_code.py		generate_gp_apis_code.py
generate_header_code.py		generate_header_code.py
generate_pybind_code.py		generate_pybind_code.py
generate_sparse_code.py		generate_sparse_code.py
generate_tf_code.py		generate_tf_code.py
generated_pybind.h		generated_pybind.h
gp_apis.py		gp_apis.py
gp_apis_tf.py		gp_apis_tf.py
kernel.cu		kernel.cu
kernel.dsl		kernel.dsl
kernel.h		kernel.h
kernel_pybind.cpp		kernel_pybind.cpp
main.py		main.py
op.h		op.h
op1.h		op1.h
pytorch_apis.py		pytorch_apis.py
test.cu		test.cu
test.py		test.py
tf_apis.py		tf_apis.py