Skip to main content
Log metrics, hyperparameters, tables, and custom data to W&B.

​
Log a CSV file to as a table

"""
Log a CSV file to W&B as a table.

Replace:
- values enclosed in angle brackets with your own
- <name>.csv in pd.read_csv() with the name of your CSV file
"""

import wandb
import pandas as pd

# Read CSV as a DataFrame object (pandas)
dataframe = pd.read_csv("<name>.csv")

# Convert the DataFrame into a W&B Table
table = wandb.Table(dataframe=dataframe)

# Start a W&B run to log data
with wandb.init(project="<project>") as run:

    # Log the table to visualize it in the W&B UI
    run.log({"<table_name>": table})

​
Log a custom summary metric

"""
Log a custom summary metric to W&B.
"""
import wandb
import random 

with wandb.init(project="<project>") as run:
    # Log a custom summary metric with a random integer value between 1 and 10
    run.summary["<metric_name>"] = random.randint(1, 10)

​
Log a custom summary metric

"""
Log a custom summary metric to W&B.

Replace:
- values enclosed in angle brackets with your own
- <summary_function_a> and <summary_function_b> in run.define_metric() with
    one of "max", "min", "last", "mean", "best", or "none".
"""
import wandb
import random

with wandb.init() as run:
    # summary_function_a and summary_function_b summary values for metric_name_a
    run.define_metric(name="<metric_name_a>", summary="<summary_function_a>")
    run.define_metric(name="<metric_name_a>", summary="<summary_function_b>")

    for i in range(10):
        log_dict = {
            "metric_name_a": random.uniform(0, 1 / (i + 1)),
        }
        run.log(log_dict)

​
Download and log an existing artifact from a registry collection

"""
Download and log an existing artifact from a W&B registry collection.
Replace the placeholders with actual registry, collection names, entity,
project, and version.
"""
import wandb

# Construct the full artifact name with version
registry_name = "<registry_name>"  # Specify the registry name
collection_name = "<collection_name>" # Specify the collection name
version = 0 # Specify the version of the artifact to use
artifact_name_registry = f"wandb-registry-{registry_name}/{collection_name}:v{version}"

# Initialize a W&B run in the different team and project
with wandb.init(entity="<entity>", project="<project>") as run:
    # Use the model artifact from the registry
    registry_model = run.use_artifact(artifact_or_name=artifact_name_registry)

    # Download the model to a local directory
    local_model_path = registry_model.download()

​
Log a histogram plot

"""
Log a histogram plot to W&B. 

Replace:
- values enclosed in angle brackets with your own
- [[1, 2], [2, 3]] in wandb.Table(data=) with your own 2D row-oriented array of values to plot

Pass a 2D row-oriented array of values to data (wandb.Table(data=)), with
column names specified in the `columns` parameter.
"""
import wandb

# Start a new run
with wandb.init(entity="<entity>", project="<project>") as run:
        
    # Create a table with the columns to plot
    table = wandb.Table(data=[[1, 2], [2, 3]] , columns=["<a_column>", "<b_column>"])

    # Use the table to populate various custom charts
    histogram = wandb.plot.histogram(table, value='<b_column>', title='Histogram')

    # Log custom tables, which will show up in customizable charts in the UI
    run.log({'histogram_1': histogram})

​
Initialize a run and log hyperparameters

"""Initializes a W&B run and logs hyperparameters."""
import wandb

config = {
    "learning_rate": 0.01,
    "batch_size": 32,
    "optimizer": "adam",
}

with wandb.init(project="<project>", config=config) as run:
    # Training and logging code goes here
    pass

​
Log a line plot

"""
Log a line plot to W&B. 

Replace:
- values enclosed in angle brackets with your own
- [[1, 2], [2, 3]] in wandb.Table(data=) with your own 2D row-oriented array of values to plot

Pass a 2D row-oriented array of values to data (wandb.Table(data=)), with
column names specified in the `columns` parameter.

Column names must match the x and y parameters in the `wandb.plot.line()` plotting function.
"""
import wandb

# Start a new run
with wandb.init(entity="<entity>", project="<project>") as run:
        
    # Create a table with the columns to plot
    table = wandb.Table(data=[[1, 2], [2, 3]], columns=["<column_a>", "<column_b>"])

    # Use the table to populate various custom charts
    line_plot = wandb.plot.line(table, x='<column_a>', y='<column_b>', title='<title>')
    
    # Log custom tables, which will show up in customizable charts in the UI
    run.log({'<line_plot>': line_plot})

​
Initialize a run and log a metric

"""Initializes a W&B run and logs a metric."""
import wandb

with wandb.init(project="<project>") as run:
    # Training and logging code goes here

    # Example of logging a metric
    run.log({"accuracy": 0.95})

​
Log a scatter plot

"""
Log a scatter plot to W&B. 

Replace:
- values enclosed in angle brackets with your own
- [[1, 2], [2, 3]] in wandb.Table(data=) with your own 2D row-oriented array of values to plot

Pass a 2D row-oriented array of values to data (wandb.Table(data=)), with
column names specified in the `columns` parameter.

Column names must match the x and y parameters in the `wandb.plot.scatter()` plotting function.
"""
import wandb

# Start a new run
with wandb.init(entity="<entity>", project="<project>") as run:
        
    # Create a table with the columns to plot
    table = wandb.Table(data=[[1, 2], [2, 3]], columns=["<a_column>", "<b_column>"])

    # Use the table to populate various custom charts
    scatter = wandb.plot.scatter(table, x='<a_column>', y='<b_column>', title='<title>')

    # Log custom tables, which will show up in customizable charts in the UI
    run.log({'scatter_1': scatter})

​
Explicitly log a summary metric

"""
Explicitly log a summary metric to W&B.
"""
import wandb
import random 

with wandb.init(project="<project>") as run:
    # Log a custom summary metric with a random integer value between 1 and 10
    run.summary["<metric_name>"] = random.randint(1, 10)

​
Log a table

"""
Log a table to W&B.

Replace: 
- values enclosed in angle brackets with your own
- [[1, 2], [2, 3]] in wandb.Table(data=) with your own 2D row-oriented array of values to log
"""
import wandb

# Create a table object with two columns and two rows of data
my_table = wandb.Table(
    columns=["<a_column>", "<b_column>"],
    data=[[1, 2], [2, 3]],
    log_mode="<log_mode>"
    )

# Start a new run
with wandb.init(entity="<entity>", project="<project>") as run:
    # Log the table to W&B
    run.log({"<table_name>": my_table})