TimeSeries
We provide own pytorch wrap
base:
- network # BaseNet : provide get_trainable and get_restorable methods
- layers # Layer : provide get_trainable and get_restorable methods
- reader # Reader : wrap for pytorch dataset, need to use file configuration
- transform # provide a numpy and putorch version of non parametric, non trainable transformation
- writer # adapter to tensorboard
- installer # use it when you want download and parse your specific dataset
contrib:
- utils
- shape_utils
train # lr save_num num epoch optimizer config
test # num config
config # all configuration
You have to build your model inheriting from a BaseNet, after that you must describe your model in init method,
also you must create your own reader inheriting Reader and create your own Dataset class inheriting from Dataset
Then you can launch training and testing by
- python3.6 train.py
- python3.6 test.py
HOW TO PLOT DATA FROM MOEX
We provide communication with http://www.moex.com with the help of open-api. To plot graph of the security you want, you need firstly install requirements.txt, to do this:
virtualenv -p python3 envname
pip install --upgrade virtualenv
pip install -r requirements.txt
And then you if you make this command:
(timenv) MacBook-Air-Roman:TimeSeries romandegtyarev$ python plotter.py ABRD
you'll see something like that:
The list of all securities:
(timenv) MacBook-Air-Roman:TimeSeries romandegtyarev$ python plotter.py list