Frequently asked questions

A workbench for language model experiments. It covers the path from raw data to a tuned model you can call or download: dataset preparation, fine-tuning and distillation, evaluation, and deployment or export.

You need to know your task and your data. The pipeline ships with defaults that produce a reasonable run, and the knobs are there once you want to change them.

Open-weight models, tuned with QLoRA or LoRA by default, with full fine-tuning available when adapters aren't enough. Small and mid-size models are the focus, because they are the ones you can afford to iterate on and realistically self-host.

A split is held back before training and never trained on. The tuned model and the base model are then scored against it and compared, so a run that didn't help shows up as a run that didn't help.

Yes. Adapters and GGUF builds can be exported and run on your own infrastructure, a laptop, or an edge device. Serving it here is a convenience, not a requirement.

Your data and the models you train from it are yours, isolated to your workspace, and are not used to train anything of ours or shared with other users. The privacy page has the detail.

Yes. The full source — the training pipeline, data generation, evaluation harness, and this site — is public on GitHub, so you can read exactly how your data is handled. Issues and pull requests are welcome.

Not currently. Ekcron is an experimental engineering project, not a commercial service, and it isn't operated by any registered business entity. There is nothing to pay for, and it comes with no warranties or availability guarantees. Treat it as an experimental environment rather than production infrastructure.

Nothing today. Compute is finite, so runs may be queued or capped to keep the environment usable, but there is no billing while the project is experimental.

Create an account, bring a small dataset, and run one short job end to end. Starting small is the fastest way to see whether the pipeline suits your task.