Fine-tune a model | Dagnam.AI Docs

Start from a curated base model, apply a shipped method to your own data, and get a trained model version.

Fine-tune a model Fine-tuning starts from a curated base model, applies a shipped method to a dataset you choose, and registers the result as a model version in My models. This page covers what fine-tuning does, how to start a run, the run page, and where your models and their versions live. What fine-tuning does From My models , Fine-tune a model opens a form for the methods the platform currently ships, each built for one kind of base model and one dataset shape. A base labeled gated cannot be used yet; choose a different one. Access Fine-tuning runs on Dagnam's cloud GPUs, which need cloud GPU access, available on request during the beta. Ask through Support /support . Start a run The run page The run page shows its status the same statuses as any training job, see Training overview /docs/training/overview job-statuses , a progress bar, and the run's estimated credit cost, shown as up to N credits . This is an estimate, not a cap: the credits charged follow how long the run actually runs. If the run fails, the error is shown. A Frozen settings section lists exactly the hyperparameters the run was recorded with: immutable, and the reproducible record of what produced the resulting model. Your models and versions My models lists the models you have registered. Opening a model shows: - Versions : each version's number and status draft, uploading, validating, ready, failed or revoked . - Lineage : for the first ready version in the list, what it was derived from: each relation and the parent version it points to. - Evaluation results : a table comparing each ready version's scores. Deploy a version Once a version is ready, its row carries a Deploy button. See Deployments overview /docs/deployments/overview .