# GAT Pytorch Run
##### from: Jessica

## What did I run?
I downloaded the [pytorch version](https://github.com/ML4ITS/mtad-gat-pytorch), made new environment in anaconda, installed all the package in requirements.txt using pip. I was able to run the SMD file after performing the preprocessing.

# clone repo
git clone https://github.com/ML4ITS/mtad-gat-pytorch.git
cd mtad-gat-pytorch

# install req
pip install -r requirements.txt 

# get dataset
cd datasets && wget https://s3-us-west-2.amazonaws.com/telemanom/data.zip && unzip data.zip && rm data.zip &&
cd data && wget https://raw.githubusercontent.com/khundman/telemanom/master/labeled_anomalies.csv &&
rm -rf 2018-05-19_15.00.10 && cd .. && cd ..

# preproc
python preprocess.py --dataset smd

# train
python train.py --dataset smd --group 1-1 --lookback 150 --epochs 10 

## How was the running process?
The running results are as follow:

## What I think could go wrong?

Preprocessing file：
	a. the training reads pkl file so the preprocess needs to change the csv to pkl
                pay attention to the paths (datasets/.../processed/name.pkl
	b. make sure to change the path during training
	c. need to specify a new preprocessing function (preprocess.py)
		i. Need to split train/test? -> maybe change the path to be the same
		ii. How to deal with the label function? We don't have that
			- redefine the loss function?

## Recommendation
1. Just ask for RD's version... They've already done it anyways
2. Work on the preprocess.py because I think it's different for every dataset. They even separate the preprocessing method for SMD and MSL. our data should use a different one as well.