​

0.2.65

​

Features

• Added new RTX4090 GPUs
• Added beam example CLI to download example apps
• Added asgi abstraction in v2
• Added real-time compute utilization metrics to dashboard

​

Bug Fixes

• Support beam configure in headless mode for CI/CD integration
• Fixed out of space error with image build worker
• Fixed image builds failing with ‘docker.io’ in the URL prefix
• Added progress bar and package size during deploy command
• Added user facing logs to indicate image loading process in worker
• Fix context not being passed when running beta9 deploy
• Fix GPU literal types by using hyphens instead of underscores

​

How to update

pip install --upgrade beam-client

​

ASGI Added to v2

You can now use @asgi decorators in V2 to deploy webservers, like Fast API or Flask apps.

​

New GPUs: RTX 4090s

These are fantastic cards: they outperform A100s for many inference tasks and only cost $0.69.

To use them, just add gpu="RTX4090" to your app:

​

Real-Time Compute Utilization in Dashboard

You can now monitor your real-time compute usage in the dashboard:

​

Download Dozens of App Templates with beam example CLI

We shipped a new CLI to quickly download templates of apps that run on Beam.

You can run beam example list to see all available templates.

To download a template, run beam example download [template-name]