If you're using fastai to train your models, W&B has an easy integration using the WandbCallback. Explore the details in interactive docs with examples →​

Start Logging with W&B

First install Weights & Biases and log in:

!pip install wandb
​
import wandb
wandb.login()

pip install wandb
wandb login

Then add the callback to the learner or fit method:

import wandb
from fastai.callback.wandb import *
​
# start logging a wandb run
wandb.init(project='my_project')
​
# To log only during one training phase
learn.fit(..., cbs=WandbCallback())
​
# To log continuously for all training phases
learn = learner(..., cbs=WandbCallback())

WandbCallback Arguments

WandbCallback accepts the following arguments:

For custom workflows, you can manually log your datasets and models:

• log_dataset(path, name=None, medata={})

• log_model(path, name=None, metadata={})

Note: any subfolder "models" will be ignored.

Data Visualization with W&B Tables

Use W&B Tables to log, query, and analyze your data. You can think of a W&B Table as a DataFrame that you can interact with inside W&B. Tables support rich media types, primitive and numeric types, as well as nested lists and dictionaries.

This pseudo-code shows you how to log images, along with their ground truth and predicted class, to W&B Tables:

# Create a new W&B Run
wandb.init(project="mnist")
​
# Create a W&B Table
my_table = wandb.Table(columns=["id", "image", "labels", "prediction"])
​
# Get your image data and make predictions
image_tensors, labels = get_mnist_data()
predictions = model(image_tensors)
​
# Add your image data and predictions to the W&B Table
for idx, im in enumerate(image_tensors): 
  my_table.add_data(idx, wandb.Image(im), labels[idx], predictions[id])
​
# Log your Table to W&B
wandb.log({"mnist_predictions": my_table})

This is will produce a Table like this:

For more examples of data visualization with W&B Tables, please see the documentation.

Examples

• ​Visualize, track, and compare Fastai models: A thoroughly documented walkthrough

• ​Image Segmentation on CamVid: A sample use case of the integration