Fine-tuning tutorial

This tutorial will help you get started with the Gemini API tuning service using either the Python SDK or the REST API using curl. The examples show how to tune the text model behind the Gemini API text generation service.

View on ai.google.dev Try a Colab notebook View notebook on GitHub

Limitations

Before tuning a model, you should be aware of the following limitations:

Fine-tuning datasets

Fine-tuning datasets for Gemini 1.5 Flash have the following limitations:

  • The maximum input size per example is 40,000 characters.
  • The maximum output size per example is 5,000 characters.
  • Only input-output pair examples are supported. Chat-style multi-turn conversations are not supported.

Tuned models

Tuned models have the following limitations:

  • The input limit of a tuned Gemini 1.5 Flash model is 40,000 characters.
  • JSON mode is not supported with tuned models.
  • Only text input is supported.

Before you begin: Set up your project and API key

Before calling the Gemini API, you need to set up your project and configure your API key.

List tuned models

You can check your existing tuned models with the tunedModels.list method.

# Sending a page_size is optional
curl -X GET https://generativelanguage.googleapis.com/v1beta/tunedModels?page_size=5 \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer ${access_token}" \
    -H "x-goog-user-project: ${project_id}" > tuned_models.json

jq .tunedModels[].name < tuned_models.json

# Send the nextPageToken to get the next page.
page_token=$(jq .nextPageToken < tuned_models.json | tr -d '"')

if [[ "$page_token" != "null"" ]]; then
curl -X GET https://generativelanguage.googleapis.com/v1beta/tunedModels?page_size=5\&page_token=${page_token}?key=$GOOGLE_API_KEY \
    -H "Content-Type: application/json"  > tuned_models2.json
jq .tunedModels[].name < tuned_models.json
fi
tuned_models.sh

Create a tuned model

To create a tuned model, you need to pass your dataset to the model in the tunedModels.create method.

For this example, you will tune a model to generate the next number in the sequence. For example, if the input is 1, the model should output 2. If the input is one hundred, the output should be one hundred one.

curl -X POST "https://generativelanguage.googleapis.com/v1beta/tunedModels?key=$GOOGLE_API_KEY" \
    -H 'Content-Type: application/json' \
    -d '
      {
        "display_name": "number generator model",
        "base_model": "models/gemini-1.5-flash-001-tuning",
        "tuning_task": {
          "hyperparameters": {
            "batch_size": 2,
            "learning_rate": 0.001,
            "epoch_count":5,
          },
          "training_data": {
            "examples": {
              "examples": [
                {
                    "text_input": "1",
                    "output": "2",
                },{
                    "text_input": "3",
                    "output": "4",
                },{
                    "text_input": "-3",
                    "output": "-2",
                },{
                    "text_input": "twenty two",
                    "output": "twenty three",
                },{
                    "text_input": "two hundred",
                    "output": "two hundred one",
                },{
                    "text_input": "ninety nine",
                    "output": "one hundred",
                },{
                    "text_input": "8",
                    "output": "9",
                },{
                    "text_input": "-98",
                    "output": "-97",
                },{
                    "text_input": "1,000",
                    "output": "1,001",
                },{
                    "text_input": "10,100,000",
                    "output": "10,100,001",
                },{
                    "text_input": "thirteen",
                    "output": "fourteen",
                },{
                    "text_input": "eighty",
                    "output": "eighty one",
                },{
                    "text_input": "one",
                    "output": "two",
                },{
                    "text_input": "three",
                    "output": "four",
                },{
                    "text_input": "seven",
                    "output": "eight",
                }
              ]
            }
          }
        }
      }' | tee tunemodel.json

# Check the operation for status updates during training.
# Note: you can only check the operation on v1/
operation=$(cat tunemodel.json | jq ".name" | tr -d '"')
tuning_done=false

while [[ "$tuning_done" != "true" ]];
do
  sleep 5
  curl -X GET "https://generativelanguage.googleapis.com/v1/${operation}?key=$GOOGLE_API_KEY" \
    -H 'Content-Type: application/json' \
     2> /dev/null > tuning_operation.json

  complete=$(jq .metadata.completedPercent < tuning_operation.json)
  tput cuu1
  tput el
  echo "Tuning...${complete}%"
  tuning_done=$(jq .done < tuning_operation.json)
done

# Or get the TunedModel and check it's state. The model is ready to use if the state is active.
modelname=$(cat tunemodel.json | jq ".metadata.tunedModel" | tr -d '"')
curl -X GET  https://generativelanguage.googleapis.com/v1beta/${modelname}?key=$GOOGLE_API_KEY \
    -H 'Content-Type: application/json' > tuned_model.json

cat tuned_model.json | jq ".state"
tuned_models.sh

The optimal values for epoch count, batch size, and learning rate are dependent on your dataset and other constraints of your use case. To learn more about these values, see Advanced tuning settings and Hyperparameters.

Your tuned model is immediately added to the list of tuned models, but its state is set to "creating" while the model is tuned.

Try the model

You can use the tunedModels.generateContent method and specify the name of the tuned model to test its performance.

curl -X POST https://generativelanguage.googleapis.com/v1beta/$modelname:generateContent?key=$GOOGLE_API_KEY \
    -H 'Content-Type: application/json' \
    -d '{
        "contents": [{
        "parts": [{
          "text": "LXIII"
          }]
        }]
        }' 2> /dev/null
tuned_models.sh

Delete the model

You can clean up your tuned model list by deleting models you no longer need. Use the tunedModels.delete method to delete a model. If you canceled any tuning jobs, you may want to delete those as their performance may be unpredictable.

curl -X DELETE https://generativelanguage.googleapis.com/v1beta/${modelname}?key=$GOOGLE_API_KEY \
    -H 'Content-Type: application/json' 
tuned_models.sh