How to use LoRA trainer and not default to Dreambooth?

[basicfeatures]

train.rb

I always end up using Dreambooth instead of lora-trainer. Maybe you know of a solution?

Cheers!

# GENERATE MUSIC ARTIST PROMO PHOTOS
#   https://warp.net/, https://baardlunde.com/ & https://dannyhiele.com/

# --

# gem install --user-install specific_install
# gem specific_install --user-install https://github.com/dreamingtulpa/replicate-ruby.git

require "replicate"
require "pry"

# Optimize images before zipping them:
#   https://replicate.com/cjwbw/backgroundmatting
#   https://replicate.com/tencentarc/gfpgan

artists = [
  "jotep"
# "tamba",
# "janhb",
# "trond"
]

Replicate.configure do |config|
  config.api_token = "XXX"
end

model = Replicate.client.retrieve_model("cloneofsimo/lora-training", version: :latest)

artists.each do |artist|
  puts "Starting `#{ artist }`..."

  training = Replicate.client.create_training(
    task: "face",
    input: {
      instance_data: "https://mysite.com/#{ artist }.zip",
      instance_prompt: "photo of a #{ artist } man",
      class_prompt: "photo of a man",
    },
    model: ("basicfeatures/mod-#{ artist }")
  )
end

[basicfeatures]

As mentioned by @Pwntus on Discord training = Replicate.client.create_training should probably be more like training = model.create_training.

Closing for now.

[basicfeatures]

Full error: https://clbin.com/6Tjkj

train.rb

Replicate.configure do |config|
  config.api_token = "XXX"
end

artists.each do |artist|
  puts "Starting `#{ artist }`..."

  model = Replicate.client.retrieve_model("cloneofsimo/lora-training", version: :latest)

  training = model.create_training(
    task: "face",
    input: {
      instance_data: "https://mysite.com/#{ artist }.zip",
      instance_prompt: "photo of a #{ artist } man",
      class_prompt: "photo of a man"
    },
    model: ("basicfeatures/mod-#{ artist }")
  )
end

[basicfeatures]

Meh. Btw peep my latest replicate-ruby script 🙏 😎