Skip to main content

​
Overview

Train a custom model with your images (pro)
  • Output Type: lora
  • Base Model: flux-dev
  • Estimated Cost: 0.1 * max_train_steps credits
  • Handler: gcp

​
Parameters

​
Required Parameters

​
name
string
required
Name of your trained model. Does not have to be unique. Used to ‘trigger’ your model in prompts.
  • Label: Model name
​
lora_training_urls
array<image>
required
Images to train your visual model. (Use at least 5 imgs, 10-20 is even better.)
  • Label: Your training images

​
Optional Parameters

​
mode
string
default:"auto"
Type of model to train (leave ‘auto’ if you’re not sure).
  • Label: Training Mode
  • Options: auto, face, object, style
​
max_train_steps
integer
default:3000
Number of training steps. Only increase this if you have lots of images (>50) otherwise the model wil overfit and will not look good.
  • Label: Training steps
  • Minimum: 1000
  • Maximum: 5000
​
lora_rank
integer
default:8
LoRA rank. Higher gives the model more capacity to learn but might lead to overfitting. 8 is almost always perfect.
  • Label: Rank
  • Minimum: 2
  • Maximum: 32
​
learning_rate
float
default:0.00005
Learning rate for training. Higher values converge faster but too high might cause overfitting. Low values are more accurate but take more training steps.
  • Label: Learning rate
  • Minimum: 1e-05
  • Maximum: 0.0002
​
caption_prefix
string
Custom text to visually describe and trigger your model (leave this empty unless you know what you’re doing!)
  • Label: Custom Trigger Text
​
custom_validation_prompts
array<string>
4 custom text prompts to test your model after training. If left empty, these will be auto-generated.
  • Label: Custom validation prompts
​
seed
integer
default:"random"
Set random seed for reproducibility. If blank, will be set to a random value.
You should only set this if you want to start from/copy the seed of a previous image. Unless one is specified, you should leave this blank!
  • Label: Seed
  • Minimum: 0
  • Maximum: 2147483647
⌘I