Experiment Tracking

Track, compare, and visualize with 5 lines of code

Add a few lines to your script to start logging results. Our lightweight integration works with any Python script.

try a live notebook



# Flexible integration for any Python script

import wandb

# 1. Start a W&B run

wandb.init(project='gpt3')

# 2. Save model inputs and hyperparameters

config = wandb.config

config.learning_rate = 0.01

# Model training here

# 3. Log metrics over time to visualize performance

wandb.log({"loss": loss})

import wandb

# 1. Start a W&B run

wandb.init(project='gpt3')

# 2. Save model inputs and hyperparameters

config = wandb.config

config.learning_rate = 0.01

# Model training here

# 3. Log metrics over time to visualize performance

with tf.Session() as sess:

# ...

wandb.tensorflow.log(tf.summary.merge_all())

import wandb

# 1. Start a new run

wandb.init(project="gpt-3")

# 2. Save model inputs and hyperparameters

config = wandb.config

config.learning_rate = 0.01

# 3. Log gradients and model parameters

wandb.watch(model)

for batch_idx, (data, target) in

enumerate(train_loader):

if batch_idx % args.log_interval == 0:

# 4. Log metrics to visualize performance

wandb.log({"loss": loss})

import wandb

from wandb.keras import WandbCallback

# 1. Start a new run

wandb.init(project="gpt-3")

# 2. Save model inputs and hyperparameters

config = wandb.config

config.learning_rate = 0.01

... Define a model

# 3. Log layer dimensions and metrics over time

model.fit(X_train, y_train, validation_data=(X_test, y_test),

callbacks=[WandbCallback()])

import wandb

wandb.init(project="visualize-sklearn")

# Model training here

# Log classifier visualizations

wandb.sklearn.plot_classifier(clf, X_train, X_test, y_train, y_test, y_pred, y_probas, labels, model_name='SVC', feature_names=None)

# Log regression visualizations

wandb.sklearn.plot_regressor(reg, X_train, X_test, y_train, y_test, model_name='Ridge')

# Log clustering visualizations

wandb.sklearn.plot_clusterer(kmeans, X_train, cluster_labels, labels=None, model_name='KMeans')

# 1. Import wandb and login

import wandb
wandb.login()

# 2. Define which wandb project to log to and name your run

wandb.init(project="gpt-3", run_name='gpt-3-base-high-lr')

# 3. Add wandb in your Hugging Face `TrainingArguments`

args = TrainingArguments(... , report_to='wandb')

# 4. W&B logging will begin automatically when your start training your Trainer

trainer = Trainer(... , args=args)
trainer.train()

import wandb

# 1. Start a new run

wandb.init(project="visualize-models", name="xgboost")

# 2. Add the callback

bst = xgboost.train(param, xg_train, num_round, watchlist, callbacks=[wandb.xgboost.wandb_callback()])

# Get predictions

pred = bst.predict(xg_test)

Visualize and compare every experiment

See model metrics stream live into interactive graphs and tables. It is easy to see how your latest model is performing compared to previous experiments, no matter where you are training your models.

Quickly find and re-run previous model checkpoints

Save everything you need to reproduce models later— the latest git commit, hyperparameters, model weights, and even sample test predictions. You can save experiment files and datasets directly to W&B or store pointers to your own storage.

import wandb
from transformers import DebertaV2ForQuestionAnswering

# 1. Create a wandb run

project = wandb.init(project='turkish-qa')

# 2. Connect to the model checkpoint you want on W&B

wandb_model = run.use_artifact('sally/turkish-qa/deberta-v2:v5')

# 3. Download the model files to a directory

model_dir = wandb_model.download()

# 4. Load your model

model = DebertaV2ForQuestionAnswering.from_pretrained(model_dir)

From “When Inception-ResNet-V2 is too slow” by Stacey Svetlichnaya

Monitor your CPU and GPU usage

Visualize live metrics like GPU utilization to identify training bottlenecks and avoid wasting expensive resources.

Debug performance in real time

See how your model is performing and identify problem areas during training. We support rich media including images, video, audio, and 3D objects.

COVID-19 main protease in complex N3 on left and COVID-19 main protease in complex with Z31792168 on right

COVID-19 main protease in complex N3 (left) and COVID-19 main protease in complex with Z31792168 (right) from “Visualizing Molecular Structure with Weights & Biases” by Nicholas Bardy

Dataset versioning with deduplication 100GB free storage

Automatically version logged datasets, with diffing and deduplication handled by Weights & Biases, behind the scenes.

simple three part illustration of artifacts workflow

Accessible anywhere

Check the latest training model and results on desktop and mobile. Use collaborative hosted projects to coordinate across your team.

The Science of Debugging with W&B Reports

By Sarah Jane of Latent Space

We use Weights & Biases as a way to share results and learnings such that we can build on top of each other's work. The W&B Reports feature has been one of the most critical...

View Report



Seamlessly share progress across projects

Manage team projects with a lightweight system of record. It's easy to hand off projects when every experiment is automatically well documented and saved centrally.

EXPERIMENTS

A system of record for your model training

Track, compare, and visualize with 5 lines of code

Visualize and compare every experiment

Quickly find and re-run previous model checkpoints

Monitor your CPU and GPU usage

Debug performance in real time

Dataset versioning with deduplication 100GB free storage

Accessible anywhere

The Science of Debugging with W&B Reports

Seamlessly share progress across projects

Never lose your progress again. Try Weights & Biases today.