Contents Menu Expand Light mode Dark mode Auto light/dark, in light mode Auto light/dark, in dark mode Skip to content
Early implementation: the bounded RadioML2018 pretraining path is verified; train/validation, evaluation, distributed training, and production-scale workflows are not supported.
RF Foundation Models
RF Foundation Models

Start Here

  • Overview
  • Maturity matrix

Current Reader Guide

  • Getting started
    • Prepare a source checkout
    • Compose the bounded experiment
    • First bounded training run
  • Concepts
    • Architecture
    • Semantic RF records
    • Configuration
    • Canonical datasets
    • Models and objectives
    • Training loop
    • Checkpointing
    • Experiment tracking
    • Reproducibility
    • Execution
  • How-to guides
    • Resume training
  • Reference
    • Maturity matrix
    • Project layout
    • Command-line interface
    • Canonical dataset contract
    • Bounded pretraining configuration
    • Checkpoint format
    • Storage layout
    • Logging and observability
    • Determinism
    • Metrics
    • Gemini Enterprise Agent Platform serverless training
    • API reference
      • Data API
      • Configuration API
      • Model and training API
      • Execution API
Back to top
View this page
Edit this page

Concepts¶

Concept pages explain the current bounded application and the limitations that matter when using it.

Core concepts¶

  • Architecture

  • Semantic RF records

  • Configuration

  • Canonical datasets

  • Models and objectives

  • Training loop

  • Checkpointing

  • Experiment tracking

  • Reproducibility

Execution boundary¶

  • Execution distinguishes current single-process provider execution from unsupported distributed training.

Next
Architecture
Previous
First bounded training run
Copyright © 2026, Superpose
Made with Sphinx and @pradyunsg's Furo
On this page
  • Concepts
    • Core concepts
    • Execution boundary