How to beginner · 3 min read

How to create wandb reports

Q: How to create wandb reports

Use the wandb Python package to log metrics and artifacts during training, then create reports by organizing runs and visualizations in the wandb web app. Programmatically, you can log data with wandb.log() and generate reports via the dashboard or export data for custom reports.

Quick answer

Use the wandb Python package to log metrics and artifacts during training, then create reports by organizing runs and visualizations in the wandb web app. Programmatically, you can log data with wandb.log() and generate reports via the dashboard or export data for custom reports.

PREREQUISITES

Python 3.8+
pip install wandb
Weights & Biases account (free tier available)
wandb API key configured via <code>wandb login</code> or environment variable

Setup

Install the wandb package and authenticate your account to enable logging and report creation.

bash

pip install wandb
wandb login

output

wandb: Currently logged in as: your_username (use `wandb login --relogin` to force relogin)

Step by step

Log metrics and artifacts during your ML experiment, then create a report in the wandb web interface by grouping runs and adding visualizations.

python

import wandb

# Initialize a new run
wandb.init(project="my-project", entity="my-entity")

# Log metrics in a loop or training step
for epoch in range(3):
    loss = 0.1 * (3 - epoch)  # example loss
    accuracy = 0.8 + 0.05 * epoch
    wandb.log({"epoch": epoch, "loss": loss, "accuracy": accuracy})

# Finish the run
wandb.finish()

output

wandb: Tracking run with wandb version 0.15.0
wandb: Run URL: https://wandb.ai/your_entity/my-project/runs/run_id
Logging metrics for epochs 0 to 2...

Common variations

You can create reports directly in the wandb web app by selecting runs and adding panels like line charts, tables, and media. For automation, use the wandb.Api() to fetch run data and generate custom reports programmatically.

python

import wandb

api = wandb.Api()
runs = api.runs("your_entity/my-project")

for run in runs:
    print(f"Run: {run.name}, Accuracy: {run.summary.get('accuracy')}")

output

Run: run-20260401_1234, Accuracy: 0.9
Run: run-20260402_5678, Accuracy: 0.85

Troubleshooting

If you see wandb: ERROR API key not found, run wandb login or set WANDB_API_KEY environment variable.
If runs do not appear in the dashboard, check your internet connection and project name spelling.
For slow logging, reduce logging frequency or batch metrics.

✅

Key Takeaways

Use wandb.init() and wandb.log() to track experiment metrics.
Create and customize reports in the wandb web app by grouping runs and adding visualizations.
Automate report generation by accessing run data with wandb.Api().
Always authenticate with wandb login or environment variables before logging.
Troubleshoot common issues by verifying API keys and project configuration.

Verified 2026-04

Verify ↗