Training RL models for financial trading using TensorTrade on Amazon SageMaker Studio

The Medium blog post that provides additional context about this work is available at medium.com/@martin.paradesi/training-rl-models-for-financial-trading-using-tensortrade-on-aws-sagemaker-studio-9ebc7996c277 .

[@FromIVtoV]

Thank you for posting this !!:D

[@SafetyPartner]

This is exactly what I've been looking for! Thank you for sharing! :)

[@martinparadesi]

You're welcome.

[@jonclement]

nice. I'm surprised it takes so long to provision sagemaker instances. Think I'll stick with colab + local. But would consider aws when I'm looking to scale (since I've got some good spot price bidding algs.

[@martinparadesi]

Thanks for the comment, Jon. Yes, this was just a proof-of-concept and a local execution of this notebook should work fine to determine the "best" RL strategy for a single ETF/stock. However, if one wants to execute this notebook for several stocks, test multiple RL reward schemes, deploy "best" models as endpoints and generate inferences on real data, I'd highly recommend leveraging Amazon SageMaker services.

[@jonclement]

yup. lots of options out there. for scaling I'd leverage the fact that sagemaker is just a Docker container notebook and look for the most flexible/cheapest place to scale it. so yes, an aws gpu ec2 spot price is 70% discounted over the norm and is probably the best option.

[@vincevangoat1]

hey and how can you use this on the actual market?

[@martinparadesi]

Hi, it is possible to use this on the market but it is important to understand the risks involved in trading.