​
From The Adaptive UI

You can launch a training in the adaptive UI. In the “Trainings” tab, you can hit the “New Training” button and select “Custom Recipe” as the training method. This will lead you to a wizard to launch your recipe based on its content and config. You have 3 steps:
  1. Select the recipe that you want to run
  2. Select the values for the config. The Adaptive UI automatically create widgets to input values based on the types of your config fields. (See the image below)
  3. Review your custom recipe and launch the job

Config parameters selection in the UI

Above is an example of parameter selection in the UI for a config defined in the custom recipe as:
custom_recipe.py
class MyConfig(InputConfig):
    # Training parameters
    learning_rate: float = 0.001
    batch_size: int = 32
    epochs: int = 10

    # Model parameters
    model_name: AdaptiveModel
    max_length: int = 2048

​
From the SDK

The Adaptive SDK provides a comprehensive interface for managing custom recipes. You can upload your recipe files, list existing recipes, and launch training jobs programmatically. First, initialize the Adaptive SDK client: 
from adaptive_sdk import Adaptive

# Initialize client
client = Adaptive(
    base_url="[ADAPTIVE_URL]",
    api_key="[ADAPTIVE_API_KEY]"
)

# Set default use case (optional)
client.set_default_use_case("my-use-case")

​
Uploading Custom Recipes

Use adaptive.custom_recipes.upload() to upload your recipe files to the platform. 
# Upload a recipe file
recipe_data = client.custom_recipes.upload(
    file_path="/path/to/your/recipe.py",
    custom_script_key="my-recipe",
    name="PPO Training Recipe"
)

print(f"Uploaded recipe with ID: {recipe_data.id}")

​
Listing Recipes

Use adaptive.custom_recipes.list() to view all available recipes. To list all recipes: 
# List all recipes in the default use case
recipes = client.custom_recipes.list()

for recipe in recipes:
    print(f"Recipe: {recipe.name} (Key: {recipe.key})")
To list all recipes of a specific use-case: 
# List recipes in a specific use case
recipes = client.custom_recipes.list(use_case="production")

for recipe in recipes:
    print(f"Recipe: {recipe.name} (Key: {recipe.key})")

​
Launching Training Jobs

Use adaptive.custom_recipes.run_training_recipe() to launch training jobs with your uploaded recipes. Arguments to pass to the recipe config should be given in the recipe_args argument and should perfectly match the config defined in the recipe. 
# Launch training with a dataset
job = client.custom_recipes.run_training_recipe(
    model="llama-3.1-8b-instruct",
    recipe_key="ppo-training-recipe",
    output_name="ppo-trained-model",
    recipe_args = {
        "lr": 0.02,
        "batch_size": 8
    }
)