Skip to content

Logging and Metrics Examples

Quick examples for logging and evaluation. For comprehensive documentation, see the Logging Guide and Metrics Guide.

Multiple Loggers

import autotimm as at  # recommended alias
from autotimm import AutoTrainer, LoggerConfig

trainer = AutoTrainer(
    max_epochs=10,
    logger=[
        LoggerConfig(backend="tensorboard", params={"save_dir": "logs/tb"}),
        LoggerConfig(backend="csv", params={"save_dir": "logs/csv"}),
    ],
)
trainer.fit(model, datamodule=data)

Full Logging Documentation - All logger backends, configurations, and features.

MLflow Tracking

from autotimm import AutoTrainer, LoggerConfig

trainer = AutoTrainer(
    max_epochs=10,
    logger=[
        LoggerConfig(
            backend="mlflow",
            params={
                "experiment_name": "cifar10-experiments",
                "tracking_uri": "http://localhost:5000",
            },
        ),
    ],
)
trainer.fit(model, datamodule=data)

Weights & Biases Integration

from autotimm import AutoTrainer, LoggerConfig

trainer = AutoTrainer(
    max_epochs=20,
    logger=[
        LoggerConfig(
            backend="wandb",
            params={
                "project": "my-project",
                "name": "resnet50-cifar10",
                "tags": ["baseline", "resnet"],
            },
        ),
    ],
)

Detailed Evaluation

Configure metrics with MetricManager:

from autotimm import ImageClassifier, LoggingConfig, MetricConfig, MetricManager

metric_configs = [
    MetricConfig(
        name="accuracy",
        backend="torchmetrics",
        metric_class="Accuracy",
        params={"task": "multiclass"},
        stages=["train", "val", "test"],
        prog_bar=True,
    ),
    MetricConfig(
        name="f1",
        backend="torchmetrics",
        metric_class="F1Score",
        params={"task": "multiclass", "average": "macro"},
        stages=["val", "test"],
    ),
]

metric_manager = MetricManager(configs=metric_configs, num_classes=10)

model = ImageClassifier(
    backbone="resnet50",
    num_classes=10,
    metrics=metric_manager,
    logging_config=LoggingConfig(
        log_learning_rate=True,
        log_gradient_norm=True,
        log_confusion_matrix=True,
    ),
)

Running Examples

python examples/logging_inference/multiple_loggers.py
python examples/logging_inference/mlflow_tracking.py
python examples/logging_inference/detailed_evaluation.py

See Also: