Where to find the WandB path for a trained model ?

When you instantiate an Evaluator to compute metrics, you can pass a wandb_path to indicate where you want the metrics to be logged in WandB.

If you want to create a new WandB run for these metrics, you can do:

>>> import wandb
>>> run = wandb.init(entity = your_entity, project=your_project)
>>> wandb_path = run.path

>>> from multivae.metrics import CoherenceEvaluator, CoherenceEvaluatorConfig
>>> coherence_config = CoherenceEvaluatorConfig(
...     batch_size = 128,
...     wandb_path = wandb_path,
...     num_classes = 10
... )
>>> CoherenceEvaluator(
...     model = model,
...     classifiers = classifiers,
...     test_set = test_data,
...     eval_config = coherence_config
... )

However you might want to log the metrics to the same WandB run that you used for training. To do so, you need to retrieve the wandb_path used for training:

• Either from the training folder:

  >>> from multivae.trainers.base.callbacks import load_wandb_path_from_folder
  >>> wandb_path = load_wandb_path_from_folder('dir_path')
  

• Or from the WandB online page for your run