Transfomer is a smart reinforcment learning agent that can profitably trade cryptocurrencies for $USD. The agent is called transformer because of its transfomer based model is used to determine the most optimally action.
cd transformer/transformer
python main.py
- For testing set the train argument to false.
python main.py --train False
.
├── common
│ ├── dataset.py
│ ├── env.py
│ ├── memory.py
│ ├── run.py
│ └── transformer.py
├── etc
│ ├── avg_scores_ltc_2.png
│ ├── btc_scores.png
│ ├── btc_trend.png
│ └── logo.png
├── examples
│ └── torchscript
│ └── main.cpp
├── README.md
├── requirements.txt
└── transformer
├── agent.py
├── main.py
└── test.py
- Generate a candle / grammian image dataset.
- Improve execution time
- Fix the dims on the concatenation of the models from NxSxE --> SxNxE with permutation
This project uses Reinforcement Learning methods on cryptocurrencies.
- Turn OLHC [open, low, high close] into an Image ➱ Click Here
- Using a transformer model on images ➱ Click Here
- Alterations to the transformer model (GTrXL) ➱ Click Here
- Deep Reinforcement Learning with Double Q-learning ➱ Click Here
- For more about the transformer model click here