Experiment-Specific Issues

SklearnCvExperiment Not Finding Best Parameters

Cause: Search space doesn’t include good values or not enough iterations.

Solutions:

1. Verify search space includes reasonable values:

   # Make sure these are sensible for your model
   search_space = {
       "n_estimators": [10, 50, 100, 200, 500],
       "max_depth": [None, 3, 5, 10, 20],
   }
   

2. Increase iterations or use smarter optimizer:

   optimizer = BayesianOptimizer(
       search_space=space,
       n_iter=200,  # More iterations
       experiment=experiment,
   )
   

PyTorch Lightning Metric Not Found

Cause: The metric name doesn’t match what’s logged during training.

Solution: Check your Lightning module logs the correct metric:

class MyModel(L.LightningModule):
    def validation_step(self, batch, batch_idx):
        loss = self.compute_loss(batch)
        self.log("val_loss", loss)  # Must match objective_metric

experiment = TorchLightningExperiment(
    lightning_module=MyModel,
    objective_metric="val_loss",  # Must match self.log name
    ...
)