Console logs

When you run an experiment, you may notice various messages printed to your console. W&B captures console logs and displays them in the W&B App. Use these messages to debug and monitor the behavior of your experiment.

View console logs

Access console logs for a run in the W&B App:

1. Navigate to your project in the W&B App.
2. Select a run within the Runs table.
3. Click the Logs tab in the project sidebar.

Only 10,000 lines of your logs are shown due to storage limitations

Types of console logs

W&B captures several types of console logs: informational messages, warnings, and errors, with a prefix to indicate the log’s severity.

Informational messages

Informational messages provide updates about the run’s progress and status. They are typically prefixed with wandb:.

wandb: Starting Run: abc123
wandb: Run data is saved locally in ./wandb/run-20240125_120000-abc123

Warning messages

Warnings about potential issues that don’t stop execution are prefixed with WARNING:

WARNING Found .wandb file, not streaming tensorboard metrics.
WARNING These runs were logged with a previous version of wandb.

Error messages

Error messages for serious issues are prefixed with ERROR:. These indicate problems that may prevent the run from completing successfully.

ERROR Unable to save notebook session history.
ERROR Failed to save notebook.

Console log settings

Within your code, pass the wandb.Settings object to wandb.init() to configure how W&B handles console logs. Within wandb.Settings, you can set the following parameters to control console log behavior:

• show_errors: If set to True, error messages are displayed in the W&B App. If set to False, error messages are not shown.
• silent: If set to True, all W&B console output will be suppressed. This is useful for production environments where you want to minimize console noise.
• show_warnings: If set to True, warning messages are displayed in the W&B App. If set to False, warning messages are not shown.
• show_info: If set to True, informational messages are displayed in the W&B App. If set to False, informational messages are not shown.

The following example shows how to configure these settings:

import wandb

settings = wandb.Settings(
    show_errors=True,  # Show error messages in the W&B App
    silent=False,      # Disable all W&B console output
    show_warnings=True # Show warning messages in the W&B App
)

with wandb.init(settings=settings) as run:
    # Your training code here
    run.log({"accuracy": 0.95})

Custom logging

W&B captures console logs from your application, but it does not interfere with your own logging setup. You can use Python’s built-in print() function or the logging module to log messages.

import wandb

with wandb.init(project="my-project") as run:
    for i in range(100, 1000, 100):
        # This will log to W&B and print to console
        run.log({"epoch": i, "loss": 0.1 * i})
        print(f"epoch: {i} loss: {0.1 * i}")

The console logs will look similar to the following:

1 epoch:  100 loss: 1.3191105127334595
2 epoch:  200 loss: 0.8664389848709106
3 epoch:  300 loss: 0.6157898902893066
4 epoch:  400 loss: 0.4961796700954437
5 epoch:  500 loss: 0.42592573165893555
6 epoch:  600 loss: 0.3771176040172577
7 epoch:  700 loss: 0.3393910825252533
8 epoch:  800 loss: 0.3082585036754608
9 epoch:  900 loss: 0.28154927492141724

Time stamps

Time stamps are automatically added to each console log entry. This allows you to track when each log message was generated.

You can toggle the time stamps in the console logs on or off. Within the console page select the Timestamp visible dropdown in the top left corner. You can choose to show or hide the time stamps.

Search console logs

Use the search bar at the top of the console logs page to filter logs by keywords. You can search for specific terms, labels, or error messages.

Filter with custom labels

Parameters prefixed by x_ (such as x_label) are in public preview. Create a GitHub issue in the W&B repository to provide feedback.

You can filter console logs based on the labels you pass as arguments for x_label in wandb.Settings in the UI search bar located at the top of the console log page.

import wandb

# Initialize a run in the primary node
with wandb.init(
    entity="entity",
    project="project",
    settings=wandb.Settings(
        x_label="custom_label"  # (Optional) Custom label for filtering logs
    )
) as run:
    # Your code here

Download console logs

Download console logs for a run in the W&B App:

1. Navigate to your project in the W&B App.
2. Select a run within the Runs table.
3. Click the Logs tab in the project sidebar.
4. Click the download button in the top right corner of the console logs page.

Copy console logs

Copy console logs for a run in the W&B App:

1. Navigate to your project in the W&B App.
2. Select a run within the Runs table.
3. Click the Logs tab in the project sidebar.
4. Click the copy button in the top right corner of the console logs page.