Studio
The Ertas Studio canvas: how runs are built, executed, evaluated, and iterated on.
Studio is where you fine-tune. The canvas is a visual editor: you compose a run by dropping Action Modules and connecting them to base models, datasets, training configs, and LoRA configs. When the recipe is complete, you press play, watch the Run panel, and download the result.
The pages below walk through Studio the way you actually use it: pick a model, configure the run, train it, evaluate it, iterate, and manage the project across many runs.
Picking A Base Model
Read the picking a base model guide.
Configuring A Run
Read the configuring a run guide.
Training
Read the training guide.
Evaluating A Model
Read the evaluating a model guide.
Iterating
Read the iterating guide.
Managing Projects
Read the managing projects guide.
Credits And Usage
Read the credits and usage guide.
Storage
Read the storage guide.
Handling Failures
Read the handling failures guide.
Parallel Runs
Read the parallel runs guide.
Training Tips
Read the training tips guide.
How the canvas is organised
Every Studio session has the same shape:
- A canvas in Build mode where you compose recipes by drag and drop.
- One or more Action Modules (Fine-Tune or Train) per recipe.
- A Run panel that opens in Run mode and shows live job status, logs, and metrics.
- A persistent project in the left rail that holds the canvas state, run history, and downloadable artifacts.
If you have not touched Studio yet, run the Quickstart before reading further. Most pages in this section assume you have at least one completed run to look at.