Guides & Reference

The multi-cluster training and inference fabric — train and serve any model across HPC, cloud, and on-prem GPU pools.

How the platform plumbs an FL job, what every submission ships, and the empirical lessons baked into the defaults.

Your auto-provisioned S3 bucket — browse folders, upload datasets, and edit scripts in the browser.

Step-by-step walkthrough of all 7 wizard steps — name, training, federation, model, advanced, validate, launch.

Track running jobs, read training logs, and download the final model checkpoint when training finishes.

Provision a bucket, install resontech, log in, and submit your first FL job.

Every field on ResonTechConfig — required keys, defaults, env-var patterns, self-hosted overrides.

TrainingConfig fixed fields, the extra bag, federation knobs, and previewing rendered configs.

Override the model, executor, or persistor; understand how source extraction works.

Bucket layout, upload mechanics, presigned downloads, and re-running against an existing workspace.

JobState values, polling patterns, output locations, cancellation, and archival.

Known failure modes with actionable fixes — auth, storage, submission, source extraction, networking.

Write model_def.py, wire up fl_train_model(), and adapt your existing PyTorch code for federated learning.

How to structure and shard your dataset into zips so workers receive the right data.

Mount your bucket locally with rclone so you can upload files from your machine.

A catalog of FL recipes spanning vision, language, speech, embeddings, medical, and image generation.

What ResonTech is, why it exists, and how training and inference work in four steps each.

Three GPU cluster types, how the distributed network aggregates compute, and performance characteristics.

How ML researchers, production teams, and enterprises each use ResonTech — pain points, workflow, and benefits.

Origin story, the team, mission, values, and contact information.