Investigate availability of Flower metadata

[elichad]

Some example metadata that might be useful:

• input dataset ids
• configuration parameter values (though many of these come from the pyproject.toml so we might already have this part covered)
• aggregation method
• performance metrics
• resource usage metrics
• start/end time of the run

Some of these are exposed in the stdout when Flower is running (see sample below), but I don't know if these are directly accessible within the client and server scripts.

This will feed the decision of whether to integrate RO-Crate support directly into Flower, or whether to make a plugin/tool that sits alongside it/can be incorporated into the Flower config scripts

(.venv) eli@AP-5Y3TQ73:~/feda/quickstart-pytorch$ flwr run .
INFO :      Starting FedAvg strategy:
INFO :          ├── Number of rounds: 3
INFO :          ├── ArrayRecord (0.24 MB)
INFO :          ├── ConfigRecord (train): {'lr': 0.1}
INFO :          ├── ConfigRecord (evaluate): (empty!)
INFO :          ├──> Sampling:
INFO :          │       ├──Fraction: train (1.00) | evaluate ( 0.50)
INFO :          │       ├──Minimum nodes: train (2) | evaluate (2)
INFO :          │       └──Minimum available nodes: 2
INFO :          └──> Keys in records:
INFO :                  ├── Weighted by: 'num-examples'
INFO :                  ├── ArrayRecord key: 'arrays'
INFO :                  └── ConfigRecord key: 'config'

...

INFO :      Strategy execution finished in 49.36s
INFO :      
INFO :      Final results:
INFO :      
INFO :          Global Arrays:
INFO :                  ArrayRecord (0.238 MB)
INFO :      
INFO :          Aggregated ClientApp-side Train Metrics:
INFO :          { 1: {'train_loss': '2.2276e+00'},
INFO :            2: {'train_loss': '2.1386e+00'},
INFO :            3: {'train_loss': '2.0864e+00'}}
INFO :      
INFO :          Aggregated ClientApp-side Evaluate Metrics:
INFO :          { 1: {'eval_acc': '1.0100e-01', 'eval_loss': '2.3375e+00'},
INFO :            2: {'eval_acc': '1.4480e-01', 'eval_loss': '2.2189e+00'},
INFO :            3: {'eval_acc': '1.7760e-01', 'eval_loss': '2.1385e+00'}}
INFO :      
INFO :          ServerApp-side Evaluate Metrics:
INFO :          { 0: {'accuracy': '1.0000e-01', 'loss': '2.3035e+00'},
INFO :            1: {'accuracy': '1.1100e-01', 'loss': '2.3362e+00'},
INFO :            2: {'accuracy': '1.4640e-01', 'loss': '2.2177e+00'},
INFO :            3: {'accuracy': '1.8500e-01', 'loss': '2.1380e+00'}}
INFO :      

Saving final model to disk...