Add support for distributed Metrics Visualization #12
Description
The README includes a sample visualization which showcases live loss metrics for a given pytorch worker on ray.
Implement a more general interface for the live publishing of pytorch metrics on raydar, which can help users understand the quality/progress of their distributed ML workflows on Ray in real time. A wishlist of supported features might include:
- Real time loss metrics as distributed training is happening
- Real time custom metrics per-model as distributed training is happening
- Support from within the pytorch api on ray