W&B Humanoid-v5 Benchmark (via Internet): Weights & Biases Dashboard
AgentGPT is a one-click, cloud-based platform for distributed reinforcement learning. It lets you easily host your environment simulators—either locally or in the cloud—and connect them to a central training job on AWS SageMaker. This enables efficient data collection and scalable multi-agent training using a GPT-based RL policy.
pip install agent-gpt-aws --upgrade- Config hyperparams & SageMaker:
agent-gpt config --batch_size 256 agent-gpt config --role_arn arn:aws:iam::123456789012:role/AgentGPTSageMakerRole
- List & Clear current configuration:
agent-gpt list agent-gpt clear
- Run your environment (gym/unity/unreal, etc.) before training starts:
agent-gpt simulate local agent-gpt simulate cloud
-
Train a gpt model on AWS:
agent-gpt train
-
Run agent gpt on AWS:
agent-gpt infer
- Cloud & Local Hosting: Quickly deploy environments (Gym/Unity) with a single command.
- Parallel Training: Connect multiple simulators to one AWS SageMaker trainer.
- Real-Time Inference: Serve a GPT-based RL policy for instant decision-making.
- Cost-Optimized: Minimize expenses by centralizing training while keeping simulations local if needed.
- Scalable GPT Support: Train Actor (policy) and Critic (value) GPT models together using reverse transitions.