Configuration & Hyperparameters

Logging config

t = Tracker(project="demo", config={"lr": 0.01, "batch": 32})

The config is stored with the run and shown in the run’s Config tab.

Automatic capture

Config capture aggregates several sources automatically:

• Explicit dicts passed to Tracker(config=...)
• argparse command-line arguments
• OmegaConf / Hydra configurations

import argparse

parser = argparse.ArgumentParser()
parser.add_argument("--lr", type=float, default=0.01)
args = parser.parse_args()

t = Tracker(project="demo")   # --lr is captured automatically

Code snapshots

Trailer records a git snapshot (commit, branch, diff) so each run is reproducible.

Metric directions

Declare whether higher or lower is better — best is computed semantically:

t = Tracker(
    project="demo",
    metric_directions={
        "accuracy": "max",
        "loss": "min",
        "val/f1": "max",
    },
)

The summary (best, best_step) respects these directions.