Skip to content

Latest commit

 

History

History
149 lines (104 loc) · 4.17 KB

README.md

File metadata and controls

149 lines (104 loc) · 4.17 KB
Raw

T-1000 Advanced Prototype

ubuntu

ubuntu

OS

windows

Codacy Badge

gif

Deep reinforcement learning multi-agent algorithmic trading framework that learns to trade from experience and then evaluate with brand new data

Prerequisites

An API Key on CryptoCompare

Setup

Ubuntu

# paste your API Key on .env
cp .env.example .env
# make sure you have these installed
sudo apt-get install gcc g++ build-essential python-dev python3-dev -y
# create env
conda env create -f t-1000.yml
# activate it
conda activate t-1000

Usage

On command line

# to see all arguments available
# $ python main.py --help

# to train
python main.py -a btc eth bnb -c usd

# to test
python main.py /
    --checkpoint_path results/t-1000/model-hash/checkpoint_750/checkpoint-750

On your own file

# instatiate the environment
T_1000 = CreateEnv(assets=['OMG','BTC','ETH'],
                  currency='USDT',
                  granularity='day',
                  datapoints=600)

# define the hyperparams to train
T_1000.train(timesteps=5e4,
              checkpoint_freq=10,
              lr_schedule=[
                  [
                      [0, 7e-5],  # [timestep, lr]
                      [100, 7e-6],
                  ],
                  [
                      [0, 6e-5],
                      [100, 6e-6],
                  ]
              ],
              algo='PPO')

Once you have a sattisfatory reward_mean benchmark you can see how it performs with never seen data

# same environment
T_1000 = CreateEnv(assets=['OMG','BTC','ETH'],
                  currency='USDT',
                  granularity='day',
                  datapoints=600)

# checkpoint are saved in /results
# it will automatically use a different time period from trainnig to backtest
T_1000.backtest(checkpoint_path='path/to/checkpoint_file/checkpoint-400')

Features

  • state of the art agents
  • hyperparam grid search
  • multi agent parallelization
  • learning rate schedule
  • result analysis

"It just needs to touch something to mimic it." - Sarah Connor, about the T-1000

Monitoring

Some nice tools to keep an eye while your agent train are (of course) tensorboard, gpustat and htop

# from the project home folder
$ tensorboard --logdir=models

# show how your gpu is going
$ gpustat -i

# show how your cpu and ram are going
$ htop

Credits

To do

  • Bind the agent's output with an exchange place order API