Logging

REAX supports logging metrics to various backends.

Supported Loggers

• CsvLogger: Logs metrics to a CSV file.

• TensorBoardLogger: Logs metrics to TensorBoard.

• WandbLogger: Logs metrics to Weights & Biases.

Configuration

To use a logger, pass it to the Trainer:

from reax.loggers import TensorBoardLogger

logger = TensorBoardLogger("tb_logs", name="my_model")
trainer = reax.Trainer(logger=logger)

Logging Metrics

Inside your Module, you can log metrics using self.log():

def training_step(self, batch, batch_idx):
    loss = ...
    self.log("train_loss", loss)
    return loss

Console Logging

By default, the ProgressBar will display logged metrics in the console progress bar. You can control this with the prog_bar argument:

self.log("acc", acc, prog_bar=True)