Petals

Test with a private swarm (refer to https://github.com/premAI-io/dev-portal/pull/69)

### Questions
- Where are the bottlenecks?
- There are no advantages to build_gpu? there's no way to force using both the local GPU and the swarm (without connecting the local GPU to the swarm)?
- Compare Petals with Deepspeed in a centralized scenario.