Wiro - Train Flux Lora Portrait icon

Wiro - Train Flux Lora Portrait

Fast LoRA training is a technique used to fine-tune FLUX models, enabling personalized model ou

Wiro - Train Flux Lora Portrait

AI-generated

Overview

This node performs fast LoRA training to fine-tune FLUX models for personalized model output. It is useful for users who want to train custom models using their own image datasets, specifying parameters like model name, trigger word, training steps, and learning rate. For example, a user can train a model on a set of portrait images to generate personalized art or enhancements.

Use Case Examples

  1. Training a custom portrait model with 40 images and 4000 training steps to generate personalized artwork.
  2. Using a unique trigger word to ensure the model recognizes a specific subject without overlap with common words.

Properties

Name Meaning
Trained Model Name The name of the model to be trained, used to identify the training session.
Trigger Word A unique keyword combined with a classifier word to describe the subject, ensuring no overlap with common words.
Training Steps The total number of training steps to complete, recommended around 100 steps per sample image, typically between 500 and 4000.
Learning Rate The learning rate controls the training quality; a low rate with more steps is recommended for better results.
Select A Folder For Training Data The folder path containing images and optional caption files used for training the model.
ZIP URL For Training Data A URL to a ZIP file containing images and optional captions, which is automatically extracted and used for training.

Output

JSON

  • taskid - The unique identifier for the training task.
  • url - The URL to the trained model or result, if training completed successfully.
  • status - The status of the training task, indicating if it completed successfully or failed.

Dependencies

  • Requires an API key and secret for authentication with the Wiro API service.

Troubleshooting

  • If the API response does not include a valid task ID or socket access token, the node throws an error indicating an invalid response from the Wiro API.
  • Training may fail if the provided training steps or learning rate are not within recommended ranges; adjust these parameters accordingly.
  • Ensure the training data folder or ZIP URL contains at least 10 images with optional captions to avoid training issues.

Links

Discussion