Welcome to elsciRL
This web application helps you run reinforcement learning experiments with various agents and adapters. Follow these steps to get started:
1. Application Setup
In the Application tab:
- Select your desired application/environment
- Choose a local configuration for the problem
- Select observed state data if needed
- Pick an analysis file for results visualization
2. Configuration
In the Config tab:
- Optionally import a preset configuration
- Set training parameters (episodes, repeats, seeds)
- Set testing parameters
- Select your agents and their adapters
- Configure agent-specific parameters
3. Instruction Input
In the Instruction Input tab:
- Optionally provide language instructions for the agent
- Use the LLM Instruction Planner to validate instructions
- Review and confirm the agent's understanding
4. Training
In the Train Model tab:
- Review your configuration
- Click "Run Experiment" to start training
- Monitor training progress in real-time
5. Results
In the Results tab:
- View validated instructions used in training
- Examine problem-specific results
- Analyze variance in performance
Tips
- You can save your configuration for future use
- Use the LLM Instruction Planner to ensure clear instructions
- Monitor the training log for progress and potential issues
- Results are automatically saved and can be compared across runs
App Interface Guide
Provide language instructions to help the agent or leave blank if you want to run an experiment without
User Input
What the agent see to match with your input instructions
Was the result correct?