decision transformer gym walker2d expert

Introduction

The Decision Transformer model is a reinforcement learning model trained on expert trajectories sampled from the Gym Walker2D environment. It leverages the architecture detailed in the paper [https://arxiv.org/abs/2106.01345].

Architecture

The model uses a Transformer architecture tailored for decision-making tasks in reinforcement learning settings. This approach allows the model to effectively handle sequential decision processes by predicting future actions based on past experiences and current states.

Training

This model was trained using expert trajectories, implying it learned from high-quality data from the Gym Walker2D environment. During inference, it requires specific normalization coefficients for input features:

• Mean: [1.2384834e+00, 1.9578537e-01, ..., 2.6285544e-03]
• Standard Deviation: [0.06664903, 0.16980624, ..., 6.1784487]

These coefficients are crucial for preprocessing the input data to match the training distribution.

Guide: Running Locally

To run the Decision Transformer model locally, follow these steps:

1. Clone the Repository: Obtain the model files from the Hugging Face repository.
2. Install Dependencies: Ensure you have Python and PyTorch installed, along with necessary libraries such as transformers.
3. Normalize Input: Use the provided mean and standard deviation values to normalize your input data.
4. Inference: Execute the model using the normalized inputs to predict actions.

For optimal performance, especially during training or extensive inference, consider using cloud GPUs from providers like AWS or Google Cloud.

License

The model files and associated documentation are subject to the licensing terms specified in the original repository. Make sure to review and comply with these terms when using the model.