Brevn

Brev is a strong AI/ML development platform designed to help developers create, train, and use machine learning models easily on the cloud. It makes running code on GPU instances simple, with the option to start small and grow as needed.

Key Features

Easy to Use

Brev is made by developers for developers, solving the problems of building and scaling ML projects. It gives a light yet strong platform good for many tasks.

Cost-Effective

Brev brings together many cloud providers to offer the best GPU options at great prices, ensuring you get good value for your money.

Powerful and Flexible

Brev lets users run any code on their GPU instances, making it very versatile for training and deploying ML models.

Brev CLI

The Brev Command Line Interface (CLI) simplifies managing instances with quick access and various tasks done right from the command line. Key commands include:

brev refresh Syncs instances created through the UI for CLI access.
brev list Lists all workspaces created or within an organization.
brev start Starts a stopped instance.
brev stop Temporarily pauses a Brev workspace.
brev port forward Forwards a port from a Brev workspace to a local machine.
brev delete Deletes a workspace from the account.
brev set Sets the organization context for commands.
brev login Logs into the Brev account and sets up necessary configurations.
brev logout Removes keys and logs out.
brev ssh key Retrieves SSH keys to add to the git provider.

Notebook Templates

Brev offers a collection of helpful AI/ML notebook templates for various tasks. Each notebook comes with minimum GPU specs and a 1-click deploy badge for easy setup on a GPU.

Categories

LLM Fine-tuning/Training Includes notebooks for fine-tuning models like Llama3, Llama2, Mistral, and Mixtral.
Multi-Modal Models Notebooks for models combining different types of data.
Computer Vision/Image Segmentation Templates for tasks like inference on Llama3 using TensorRT-LLM and running Segment Anything.
Miscellaneous Other applications and tools, such as deploying models to Replicate and speeding up data science using NVIDIA RAPIDS.

Community and Support

Brev encourages community involvement and provides multiple channels for support:

Discord Join the #help channel and create a support ticket with the /help command.
Email Contact [email protected] for assistance.
Shared Slack Channel Paid users get prioritized support through a shared Slack channel.

Additional Resources

Brevitas A PyTorch library for neural network quantization, supporting post-training quantization (PTQ) and quantization-aware training (QAT).
Debugging Guide Addresses common issues and provides troubleshooting steps for Brev users. To get started with Brev, install the CLI using the following command:

brew install brevdev/homebrew-brev/brev && brev login

For more information, visit Brev.dev.