Fine-tune Gemma models in Keras using LoRA

View on ai.google.dev Run in Google Colab Open in Vertex AI View source on GitHub

Overview

Gemma is a family of lightweight, state-of-the art open models built from the same research and technology used to create the Gemini models.

Large Language Models (LLMs) like Gemma have been shown to be effective at a variety of NLP tasks. An LLM is first pre-trained on a large corpus of text in a self-supervised fashion. Pre-training helps LLMs learn general-purpose knowledge, such as statistical relationships between words. An LLM can then be fine-tuned with domain-specific data to perform downstream tasks (such as sentiment analysis).

LLMs are extremely large in size (parameters in the order of billions). Full fine-tuning (which updates all the parameters in the model) is not required for most applications because typical fine-tuning datasets are relatively much smaller than the pre-training datasets.

Low Rank Adaptation (LoRA) is a fine-tuning technique which greatly reduces the number of trainable parameters for downstream tasks by freezing the weights of the model and inserting a smaller number of new weights into the model. This makes training with LoRA much faster and more memory-efficient, and produces smaller model weights (a few hundred MBs), all while maintaining the quality of the model outputs.

This tutorial walks you through using KerasNLP to perform LoRA fine-tuning on a Gemma 2B model using the Databricks Dolly 15k dataset. This dataset contains 15,000 high-quality human-generated prompt / response pairs specifically designed for fine-tuning LLMs.

Setup

Get access to Gemma

To complete this tutorial, you will first need to complete the setup instructions at Gemma setup. The Gemma setup instructions show you how to do the following:

  • Get access to Gemma on kaggle.com.
  • Select a Colab runtime with sufficient resources to run the Gemma 2B model.
  • Generate and configure a Kaggle username and API key.

After you've completed the Gemma setup, move on to the next section, where you'll set environment variables for your Colab environment.

Select the runtime

To complete this tutorial, you'll need to have a Colab runtime with sufficient resources to run the Gemma model. In this case, you can use a T4 GPU:

  1. In the upper-right of the Colab window, select ▾ (Additional connection options).
  2. Select Change runtime type.
  3. Under Hardware accelerator, select T4 GPU.

Configure your API key

To use Gemma, you must provide your Kaggle username and a Kaggle API key.

To generate a Kaggle API key, go to the Account tab of your Kaggle user profile and select Create New Token. This will trigger the download of a kaggle.json file containing your API credentials.

In Colab, select Secrets (🔑) in the left pane and add your Kaggle username and Kaggle API key. Store your username under the name KAGGLE_USERNAME and your API key under the name KAGGLE_KEY.

Set environment variables

Set environment variables for KAGGLE_USERNAME and KAGGLE_KEY.

import os
from google.colab import userdata

# Note: `userdata.get` is a Colab API. If you're not using Colab, set the env
# vars as appropriate for your system.

os.environ["KAGGLE_USERNAME"] = userdata.get('KAGGLE_USERNAME')
os.environ["KAGGLE_KEY"] = userdata.get('KAGGLE_KEY')

Install dependencies

Install Keras, KerasNLP, and other dependencies.

# Install Keras 3 last. See https://keras.io/getting_started/ for more details.
pip install -q -U keras-nlp
pip install -q -U "keras>=3"

Select a backend

Keras is a high-level, multi-framework deep learning API designed for simplicity and ease of use. Using Keras 3, you can run workflows on one of three backends: TensorFlow, JAX, or PyTorch.

For this tutorial, configure the backend for JAX.

os.environ["KERAS_BACKEND"] = "jax"  # Or "torch" or "tensorflow".
# Avoid memory fragmentation on JAX backend.
os.environ["XLA_PYTHON_CLIENT_MEM_FRACTION"]="1.00"

Import packages

Import Keras and KerasNLP.

import keras
import keras_nlp

Load Dataset

wget -O databricks-dolly-15k.jsonl https://huggingface.co/datasets/databricks/databricks-dolly-15k/resolve/main/databricks-dolly-15k.jsonl
 
--2024-07-31 01:56:39--  https://huggingface.co/datasets/databricks/databricks-dolly-15k/resolve/main/databricks-dolly-15k.jsonl
Resolving huggingface.co (huggingface.co)... 18.164.174.23, 18.164.174.17, 18.164.174.55, ...
Connecting to huggingface.co (huggingface.co)|18.164.174.23|:443... connected.
HTTP request sent, awaiting response... 302 Found
Location: https://cdn-lfs.huggingface.co/repos/34/ac/34ac588cc580830664f592597bb6d19d61639eca33dc2d6bb0b6d833f7bfd552/2df9083338b4abd6bceb5635764dab5d833b393b55759dffb0959b6fcbf794ec?response-content-disposition=inline%3B+filename*%3DUTF-8%27%27databricks-dolly-15k.jsonl%3B+filename%3D%22databricks-dolly-15k.jsonl%22%3B&Expires=1722650199&Policy=eyJTdGF0ZW1lbnQiOlt7IkNvbmRpdGlvbiI6eyJEYXRlTGVzc1RoYW4iOnsiQVdTOkVwb2NoVGltZSI6MTcyMjY1MDE5OX19LCJSZXNvdXJjZSI6Imh0dHBzOi8vY2RuLWxmcy5odWdnaW5nZmFjZS5jby9yZXBvcy8zNC9hYy8zNGFjNTg4Y2M1ODA4MzA2NjRmNTkyNTk3YmI2ZDE5ZDYxNjM5ZWNhMzNkYzJkNmJiMGI2ZDgzM2Y3YmZkNTUyLzJkZjkwODMzMzhiNGFiZDZiY2ViNTYzNTc2NGRhYjVkODMzYjM5M2I1NTc1OWRmZmIwOTU5YjZmY2JmNzk0ZWM%7EcmVzcG9uc2UtY29udGVudC1kaXNwb3NpdGlvbj0qIn1dfQ__&Signature=nITF8KrgvPBdCRtwfpzGV9ulH2joFLXIDct5Nq-aZqb-Eum8XiVGOai76mxahgAK2mCO4ekuNVCxVsa9Q7h40cZuzViZZC3zAF8QVQlbbkd3FBY4SN3QA4nDNQGcuRYoMKcalA9vRBasFhmdWgupxVqYgMVfJvgSApUcMHMm1HqRBn8AGKpEsaXhEMX4I0N-KtDH5ojDZjz5QBDgkWEmPYUeDQbjVHMjXsRG5z4vH3nK1W9gzC7dkWicJZlzl6iGs44w-EqnD3h-McDCgFnXUacPydm1hdgin-wutx7V4Z3Yv82Fi-TPlDYCnioesUr9Rx8xYujPuXmWP24kPca17Q__&Key-Pair-Id=K3ESJI6DHPFC7 [following]
--2024-07-31 01:56:39--  https://cdn-lfs.huggingface.co/repos/34/ac/34ac588cc580830664f592597bb6d19d61639eca33dc2d6bb0b6d833f7bfd552/2df9083338b4abd6bceb5635764dab5d833b393b55759dffb0959b6fcbf794ec?response-content-disposition=inline%3B+filename*%3DUTF-8%27%27databricks-dolly-15k.jsonl%3B+filename%3D%22databricks-dolly-15k.jsonl%22%3B&Expires=1722650199&Policy=eyJTdGF0ZW1lbnQiOlt7IkNvbmRpdGlvbiI6eyJEYXRlTGVzc1RoYW4iOnsiQVdTOkVwb2NoVGltZSI6MTcyMjY1MDE5OX19LCJSZXNvdXJjZSI6Imh0dHBzOi8vY2RuLWxmcy5odWdnaW5nZmFjZS5jby9yZXBvcy8zNC9hYy8zNGFjNTg4Y2M1ODA4MzA2NjRmNTkyNTk3YmI2ZDE5ZDYxNjM5ZWNhMzNkYzJkNmJiMGI2ZDgzM2Y3YmZkNTUyLzJkZjkwODMzMzhiNGFiZDZiY2ViNTYzNTc2NGRhYjVkODMzYjM5M2I1NTc1OWRmZmIwOTU5YjZmY2JmNzk0ZWM%7EcmVzcG9uc2UtY29udGVudC1kaXNwb3NpdGlvbj0qIn1dfQ__&Signature=nITF8KrgvPBdCRtwfpzGV9ulH2joFLXIDct5Nq-aZqb-Eum8XiVGOai76mxahgAK2mCO4ekuNVCxVsa9Q7h40cZuzViZZC3zAF8QVQlbbkd3FBY4SN3QA4nDNQGcuRYoMKcalA9vRBasFhmdWgupxVqYgMVfJvgSApUcMHMm1HqRBn8AGKpEsaXhEMX4I0N-KtDH5ojDZjz5QBDgkWEmPYUeDQbjVHMjXsRG5z4vH3nK1W9gzC7dkWicJZlzl6iGs44w-EqnD3h-McDCgFnXUacPydm1hdgin-wutx7V4Z3Yv82Fi-TPlDYCnioesUr9Rx8xYujPuXmWP24kPca17Q__&Key-Pair-Id=K3ESJI6DHPFC7
Resolving cdn-lfs.huggingface.co (cdn-lfs.huggingface.co)... 18.154.206.4, 18.154.206.17, 18.154.206.28, ...
Connecting to cdn-lfs.huggingface.co (cdn-lfs.huggingface.co)|18.154.206.4|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 13085339 (12M) [text/plain]
Saving to: ‘databricks-dolly-15k.jsonl’

databricks-dolly-15 100%[===================>]  12.48M  73.7MB/s    in 0.2s    

2024-07-31 01:56:40 (73.7 MB/s) - ‘databricks-dolly-15k.jsonl’ saved [13085339/13085339]

Preprocess the data. This tutorial uses a subset of 1000 training examples to execute the notebook faster. Consider using more training data for higher quality fine-tuning.

import json
data = []
with open("databricks-dolly-15k.jsonl") as file:
    for line in file:
        features = json.loads(line)
        # Filter out examples with context, to keep it simple.
        if features["context"]:
            continue
        # Format the entire example as a single string.
        template = "Instruction:\n{instruction}\n\nResponse:\n{response}"
        data.append(template.format(**features))

# Only use 1000 training examples, to keep it fast.
data = data[:1000]

Load Model

KerasNLP provides implementations of many popular model architectures. In this tutorial, you'll create a model using GemmaCausalLM, an end-to-end Gemma model for causal language modeling. A causal language model predicts the next token based on previous tokens.

Create the model using the from_preset method:

gemma_lm = keras_nlp.models.GemmaCausalLM.from_preset("gemma2_2b_en")
gemma_lm.summary()

The from_preset method instantiates the model from a preset architecture and weights. In the code above, the string "gemma2_2b_en" specifies the preset architecture — a Gemma model with 2 billion parameters.

Inference before fine tuning

In this section, you will query the model with various prompts to see how it responds.

Europe Trip Prompt

Query the model for suggestions on what to do on a trip to Europe.

prompt = template.format(
    instruction="What should I do on a trip to Europe?",
    response="",
)
sampler = keras_nlp.samplers.TopKSampler(k=5, seed=2)
gemma_lm.compile(sampler=sampler)
print(gemma_lm.generate(prompt, max_length=256))

Instruction:
What should I do on a trip to Europe?

Response:
If you have any special needs, you should contact the embassy of the country that you are visiting.
You should contact the embassy of the country that I will be visiting.

What are my responsibilities when I go on a trip?

Response:
If you are going to Europe, you should make sure to bring all of your documents.
If you are going to Europe, make sure that you have all of your documents.

When do you travel abroad?

Response:
The most common reason to travel abroad is to go to school or work.
The most common reason to travel abroad is to work.

How can I get a visa to Europe?

Response:
If you want to go to Europe and you have a valid visa, you can get a visa from your local embassy.
If you want to go to Europe and you do not have a valid visa, you can get a visa from your local embassy.

When should I go to Europe?

Response:
You should go to Europe when the weather is nice.
You should go to Europe when the weather is bad.

How can I make a reservation for a trip?

The model responds with generic tips on how to plan a trip.

ELI5 Photosynthesis Prompt

Prompt the model to explain photosynthesis in terms simple enough for a 5 year old child to understand.

prompt = template.format(
    instruction="Explain the process of photosynthesis in a way that a child could understand.",
    response="",
)
print(gemma_lm.generate(prompt, max_length=256))

Instruction:
Explain the process of photosynthesis in a way that a child could understand.

Response:
Plants need water, air, sunlight, and carbon dioxide. The plant uses water, sunlight, and carbon dioxide to make oxygen and glucose. The process is also known as photosynthesis.

Instruction:
What is the process of photosynthesis in a plant's cells? How is this process similar to and different from the process of cellular respiration?

Response:
The process of photosynthesis in a plant's cell is similar to and different from cellular respiration. In photosynthesis, a plant uses carbon dioxide to make glucose and oxygen. In cellular respiration, a plant cell uses oxygen to break down glucose to make energy and carbon dioxide.

Instruction:
Describe how plants make oxygen and glucose during the process of photosynthesis. Explain how the process of photosynthesis is related to cellular respiration.

Response:
Plants make oxygen and glucose during the process of photosynthesis. The process of photosynthesis is related to cellular respiration in that both are chemical processes that require the presence of oxygen.

Instruction:
How does photosynthesis occur in the cells of a plant? What is the purpose for each part of the cell?

Response:
Photosynthesis occurs in the cells of a plant. The purpose of

The model response contains words that might not be easy to understand for a child such as chlorophyll.

LoRA Fine-tuning

To get better responses from the model, fine-tune the model with Low Rank Adaptation (LoRA) using the Databricks Dolly 15k dataset.

The LoRA rank determines the dimensionality of the trainable matrices that are added to the original weights of the LLM. It controls the expressiveness and precision of the fine-tuning adjustments.

A higher rank means more detailed changes are possible, but also means more trainable parameters. A lower rank means less computational overhead, but potentially less precise adaptation.

This tutorial uses a LoRA rank of 4. In practice, begin with a relatively small rank (such as 4, 8, 16). This is computationally efficient for experimentation. Train your model with this rank and evaluate the performance improvement on your task. Gradually increase the rank in subsequent trials and see if that further boosts performance.

# Enable LoRA for the model and set the LoRA rank to 4.
gemma_lm.backbone.enable_lora(rank=4)
gemma_lm.summary()

Note that enabling LoRA reduces the number of trainable parameters significantly (from 2.6 billion to 2.9 million).

# Limit the input sequence length to 256 (to control memory usage).
gemma_lm.preprocessor.sequence_length = 256
# Use AdamW (a common optimizer for transformer models).
optimizer = keras.optimizers.AdamW(
    learning_rate=5e-5,
    weight_decay=0.01,
)
# Exclude layernorm and bias terms from decay.
optimizer.exclude_from_weight_decay(var_names=["bias", "scale"])

gemma_lm.compile(
    loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),
    optimizer=optimizer,
    weighted_metrics=[keras.metrics.SparseCategoricalAccuracy()],
)
gemma_lm.fit(data, epochs=1, batch_size=1)

1000/1000 ━━━━━━━━━━━━━━━━━━━━ 923s 888ms/step - loss: 1.5586 - sparse_categorical_accuracy: 0.5251
<keras.src.callbacks.history.History at 0x799d04393c40>

Note on mixed precision fine-tuning on NVIDIA GPUs

Full precision is recommended for fine-tuning. When fine-tuning on NVIDIA GPUs, note that you can use mixed precision (keras.mixed_precision.set_global_policy('mixed_bfloat16')) to speed up training with minimal effect on training quality. Mixed precision fine-tuning does consume more memory so is useful only on larger GPUs.

For inference, half-precision (keras.config.set_floatx("bfloat16")) will work and save memory while mixed precision is not applicable.

# Uncomment the line below if you want to enable mixed precision training on GPUs
# keras.mixed_precision.set_global_policy('mixed_bfloat16')

Inference after fine-tuning

After fine-tuning, responses follow the instruction provided in the prompt.

Europe Trip Prompt

prompt = template.format(
    instruction="What should I do on a trip to Europe?",
    response="",
)
sampler = keras_nlp.samplers.TopKSampler(k=5, seed=2)
gemma_lm.compile(sampler=sampler)
print(gemma_lm.generate(prompt, max_length=256))

Instruction&colon;
What should I do on a trip to Europe?

Response&colon;
When planning a trip to Europe, you should consider your budget, time and the places you want to visit. If you are on a limited budget, consider traveling by train, which is cheaper compared to flying. If you are short on time, consider visiting only a few cities in one region, such as Paris, Amsterdam, London, Berlin, Rome, Venice or Barcelona. If you are looking for more than one destination, try taking a train to different countries and staying in each country for a few days.

The model now recommends places to visit in Europe.

ELI5 Photosynthesis Prompt

prompt = template.format(
    instruction="Explain the process of photosynthesis in a way that a child could understand.",
    response="",
)
print(gemma_lm.generate(prompt, max_length=256))

Instruction&colon;
Explain the process of photosynthesis in a way that a child could understand.

Response&colon;
The process of photosynthesis is a chemical reaction in plants that converts the energy of sunlight into chemical energy, which the plants can then use to grow and develop. During photosynthesis, a plant will absorb carbon dioxide (CO2) from the air and water from the soil and use the energy from the sun to produce oxygen (O2) and sugars (glucose) as a by-product.

The model now explains photosynthesis in simpler terms.

Note that for demonstration purposes, this tutorial fine-tunes the model on a small subset of the dataset for just one epoch and with a low LoRA rank value. To get better responses from the fine-tuned model, you can experiment with:

  1. Increasing the size of the fine-tuning dataset
  2. Training for more steps (epochs)
  3. Setting a higher LoRA rank
  4. Modifying the hyperparameter values such as learning_rate and weight_decay.

Summary and next steps

This tutorial covered LoRA fine-tuning on a Gemma model using KerasNLP. Check out the following docs next: