Skip to content

🐝 Weights & Biases Integration

What you'll learn

How to integrate W&B for rich experiment visualization — interactive dashboards for your FlowyML pipelines.

Visualize training runs, track artifacts, and collaborate with your team using Weights & Biases.

Why Weights & Biases?

Feature Benefit
Visualization Interactive charts for loss, accuracy, and system metrics
Collaboration Share results with your team instantly
Artifacts Track dataset and model versioning
Sweeps Hyperparameter optimization with built-in sweep support

🐝 Setup

1
2
pip install wandb
wandb login  # Authenticate with your API key

Configuration

Enable W&B tracking for your pipeline run:

1
2
3
4
5
6
7
8
9
from flowyml.integrations.wandb import WandBTracker

pipeline.run(
    tracker=WandBTracker(
        project="flowyml-demo",
        entity="my-team",
        tags=["training", "v2"],
    )
)

WandBTracker Parameters

Parameter Type Default Description
project str required W&B project name
entity str None Team or user entity
tags list[str] [] Run tags for filtering
group str None Group related runs
name str None Custom run name

📊 Logging Custom Metrics

FlowyML auto-captures step inputs/outputs, but you can add custom logs:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
import wandb
from flowyml import step

@step
def train(data):
    for epoch in range(10):
        loss = train_epoch(data)
        wandb.log({"epoch": epoch, "loss": loss, "accuracy": 1 - loss})

    # Log rich media
    wandb.log({"confusion_matrix": wandb.plot.confusion_matrix(...)})
    wandb.log({"roc": wandb.plot.roc_curve(...)})

    return model

📦 W&B Artifacts

Use W&B Artifacts for dataset and model versioning:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
@step
def load_data():
    run = wandb.init()
    artifact = run.use_artifact("mnist:v1")
    data_dir = artifact.download()
    return load_dataset(data_dir)

@step
def save_model(model):
    artifact = wandb.Artifact("my-model", type="model")
    artifact.add_file("model.pt")
    wandb.log_artifact(artifact)

Best Practices

Tag your runs

Use tags like ["training", "v2", "gpu"] to filter runs in the W&B dashboard.

Use groups for experiments

Group related runs with group="hyperparameter_search" to compare them in a single view.