AI/ML professionals: Get 500 FREE compute hours with Dis.co. Get it now.
Run and automate deep learning experiments faster
Run experiments on hundreds of machines, on and off the cloud, manage huge data sets and gain unprecedented visibility into your experiments
DevOps for deep learning
Run and compare hundreds of experiments, version control data in the cloud or on-premise, and automate compute resources on AWS, Microsoft Azure, Google Cloud, or a local cluster.
Data management
Version control your deep learning data
Iterate faster and track the best models you’ve created, what data you train with and experiment on, what hyperparameters you used, and the associated tradeoffs.
Manage data like a pro
Easily version control and track the complete evolution of your computer vision models with datasets, hyperparameters, machines, data sources, and code.
Exploring data is one query away
Query language and functionalities that let you easily slice and dice the data in the cloud or on-premise. Compare data queries between computer vision experiments and analyze the performance of different deep learning datasets.
Accelerate training on massive datasets
Upload Gigabytes or Petabytes of training data to your framework in no time. We stream your computer vision data, cache it locally and only sync changes, saving you time and keeping your data secure.
Keep data private and secure
Protect your business and your customers' data by hosting all of your assets in your own cloud environment or on-premise data storage. Your data don’t leave your premise or cloud.
Experiment management
Run deep learning experiments 10x faster
Run, track and manage innumerable deep-learning experiments faster and with greater confidence. Visualize your experiments, and the hyperparameters used in them.
Compare experiments
Compare experiment results, hyperparameters, versions of training data and source code, so you can quickly analyze what worked and what didn’t.
Visualize and monitor experiments
Run, track, and manage all your team experiments in a single place. Visualize experiments, hyperparams, source code, data, logs, artifacts and resources for easy consumption at-a-glance in real-time.
Reproduce experiments with one click
Everything you need to revisit, examine and reproduce your experiments. Refine your models by reducing and eliminating variations when rerunning failed jobs or previous experiments.
Resource management
Deep learning for computer vision: cloud, on-premise or hybrid
Train deep learning models with ease by auto-scaling your compute resources for the best possible outcome and ROI. Manage your local, hybrid, or public cloud (AWS, Microsoft Azure, Google Cloud) compute resources as a single environment.
Get started in minutes
Install the MissingLink SDK and within minutes run and compare hundreds of computer vision experiments, version control data on cloud or on-premise, and automate compute resources on AWS, Microsoft Azure, Google Cloud, or a local cluster.
Manage and scale resources to meet the demands of your team
Whether you’re using Microsoft Azure, AWS, hybrid, or your local clusters - MissingLink is the most comprehensive deep learning platform to train your computer vision models more frequently, at lower cost and with greater confidence.
Amazon spot instances support
Take full advantage of AWS Spot Instances. Experiments that use Spot and currently submit a bid price will resume working with no changes if they get outbid reducing the cost of running experiments as compared to On-Demand pricing
Integrate with the tools you already use
Use only the MissingLink services you need and integrate seamlessly.
TensorFlow
PyCaffe
PyTorch
Keras
scikit-learn
AWS
Google cloud
Azure
Flexible, open and extensible deep learning platform
Choose only the MissingLink services you need
Using a specific data-tagging tool? Have your own custom workflow? No problem.
Our API-first deep learning platform lets you integrate just the functionality you need.
# Auto document and share your experiment results in a few lines import missinglink missinglink_callback = missinglink.KerasCallback() callbacks = [missinglink_callback] model = tf.keras.models.Sequential([ tf.keras.layers.Flatten(input_shape=(28, 28)), tf.keras.layers.Dense(128, activation='relu'), tf.keras.layers.Dense(10, activation='softmax') ]) model.compile( optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) model.fit(x_train, y_train, epochs=5, callbacks=callbacks) model.evaluate(x_test, y_test, callbacks=callbacks)
# Launch an experiment on your premises or cloud ml run xp # Fetch a slice and version of your dataset from your premises or cloud ml data clone 5685154290860032 --query "@version: aca1a37 @seed:1337 pedestrians:true" --dest-folder ./
# Train only on images with cats
query = 'cat:>0'
data_gen = missinglink_callback.bind_data_generator(
data_volume_id,
query,
deserialization_callback)
train_generator, test_generator, val_generator = data_gen.flow()
# Data is streamed. No more waiting for the entire dataset
model.fit_generator(train_generator)
