EPFL | Deep Learning (EE-559) (Spring 2020) | Mini-project 2
- This is our implementaion for the mini-project 2 in the Deep leaning course at EPFL.
- Team member: Pengkang Guo, Xiaoyu Lin
- [report]
The objective of this project is to design a mini "deep learning framework" using only pytorch's tensor operations and the standard math library, hence in particular without using autograd or the neural-network modules.
Pytorch
From the root of the project: python test.py
- module.py: the implementation of the modules
- Includes
Linear
,Relu
,Tanh
,LossMSE
andSequential
.
- Includes
- test.py: the required basic Python script using our framework
module.py
.
- Generates the dataset, initializes and trains the required model with three hidden layers of 25 units.
- Generates an output file,
logs.out
, logging the loss and error rate of each epoch. Calculates and prints the average test error rate, average time and their standard deviations.
- test_plot.py: an upgraded version of
test.py
.
- Has all the functions of
test.py
. - Generates the images of the dataset, training error rate and test error rate.
- Has all the functions of
- test_torch.py: an Pytorch version of
test_plot.py
.
- has all the functions of
test_plot.py
but is implemented using Pytorch
- has all the functions of