This pull request introduces support for the H100 GPU as a hardware option for Replicate cloud jobs, improves the accuracy of hardware cost calculations, and updates the UI to allow users to easily select between L40S and H100 base hardware options. The changes also ensure that hardware pricing is consistent and visible throughout the application, and add corresponding tests to verify these behaviors.

Backend: Hardware Options and Pricing

• Added H100 as a supported hardware option, replacing A100, and updated cost values for both L40S and H100 in DEFAULT_HARDWARE_INFO and related cost calculations. [1] [2] [3] [4]
• Enhanced the provider enrichment logic to calculate and expose per-hour and per-second costs for both L40S and H100 hardware profiles, including support for multi-GPU profiles. [1] [2]
• Updated type hints and profile listing logic to support richer hardware profile metadata. [1] [2]

Frontend: UI Improvements for Hardware Selection

• Added UI components to cloud_tab.html and onboarding flows to let users select between L40S and H100, displaying accurate cost information for each. [1] [2] [3]
• Updated the hero CTA to reference H100 instead of A100.
• Implemented new JavaScript methods for fetching, selecting, and displaying base hardware options and their costs, with persistence of user selection.

Testing

• Added and updated tests to verify hardware options and cost calculations, as well as frontend logic for hardware selection and display. [1] [2]