Wiro - Train Sdxl Lora icon

Wiro - Train Sdxl Lora

LoRA training is a low-rank adaptation technique used to fine-tune Stable Diffusion XL (SDXL) m

Wiro - Train Sdxl Lora

AI-generated

Overview

This node facilitates the training of a LoRA (Low-Rank Adaptation) model specifically for Stable Diffusion XL (SDXL). It is designed to fine-tune SDXL models by sending training parameters and data to the Wiro API, which handles the training process. The node is useful for users who want to customize and improve SDXL models with their own datasets and training configurations, such as adjusting prompts, batch sizes, learning rates, and precision settings.

Use Case Examples

  1. A user wants to train a custom SDXL LoRA model with specific instance and class prompts, validation images, and training steps to improve image generation quality for a particular style or subject.
  2. A developer automates the training of multiple LoRA models with different hyperparameters and datasets by integrating this node into an n8n workflow.

Properties

Name Meaning
Trained Model Name The name assigned to the trained LoRA model.
Train Model Description A description providing details about the trained model.
Reference Model The base model used as a reference for training.
Reference Model Private Indicates if the reference model is private.
Classification Dataset Dataset used for classification during training.
Instance Prompt Prompt describing the instance to be trained.
Class Prompt Prompt describing the class for training.
Validation Prompt Prompt used for validation during training.
Validation Negative Prompt Negative prompt used for validation.
Num Validation Images Number of images used for validation.
Num Class Images Number of class images used in training.
With Prior Preservation Flag to enable prior preservation during training.
Prior Loss Weight Weight assigned to the prior loss during training.
Seed Seed value for reproducibility of training results.
Resolution Resolution of images used in training.
Center Crop Flag to enable center cropping of images.
Train Text Encoder Flag to enable training of the text encoder.
Train Batch Size Batch size used during training.
Sample Batch Size Batch size used for sampling.
Training Steps Number of training steps to perform.
Gradient Accumulation Steps Number of steps to accumulate gradients.
Gradient Checkpointing Flag to enable gradient checkpointing to save memory.
Learning Rate Learning rate for the optimizer.
Scale Lr Flag to enable learning rate scaling.
Lr Scheduler Learning rate scheduler type.
Lr Warm Up Steps Number of warm-up steps for learning rate scheduler.
Lr Num Cycles Number of cycles for learning rate scheduler.
Lr Power Power parameter for polynomial learning rate scheduler.
Use 8bit Adam Flag to enable 8-bit Adam optimizer.
Adam Beta1 Beta1 parameter for Adam optimizer.
Adam Beta2 Beta2 parameter for Adam optimizer.
Adam Weight Decay Weight decay parameter for Adam optimizer.
Adam Epsilon Epsilon parameter for Adam optimizer.
Max Grad Norm Maximum gradient norm for clipping.
Checkpointing Steps Steps interval for checkpointing during training.
Checkpoints Total Limit Maximum number of checkpoints to keep.
Mixed Precision Mixed precision mode for training.
Prior Generation Precision Precision mode for prior generation.
Lora Rank Rank parameter for LoRA.
Allowtf32 Flag to enable TensorFloat-32 precision.
Selected Folder For Train Data Folder path for training data.
ZIP URL For Training Data URL to a ZIP file containing training data, which will be extracted automatically.

Output

JSON

  • taskid - Identifier of the training task returned by the Wiro API.
  • url - URL to the trained model or result, if training completed successfully.
  • status - Status of the training task, e.g., 'completed' or 'failed'.

Dependencies

  • Wiro API

Troubleshooting

  • If the Wiro API does not return a valid task ID or socket access token, the node throws an error indicating an invalid response. Ensure API credentials are correct and the API endpoint is reachable.
  • Training may fail if required parameters are missing or incorrectly set; verify all required inputs are provided and valid.
  • Network issues or API downtime can cause polling to fail; check network connectivity and Wiro API status.

Links

Discussion