Quickstart

Transformers is designed to be fast and easy to use so that everyone can start learning or building with transformer models.

The number of user-facing abstractions is limited to only three classes for instantiating a model, and two APIs for inference or training. This quickstart introduces you to Transformers’ key features and shows you how to:

• load a pretrained model
• run inference with Pipeline
• fine-tune a model with Trainer

Set up

To start, we recommend creating a Hugging Face account. An account lets you host and access version controlled models, datasets, and Spaces on the Hugging Face Hub, a collaborative platform for discovery and building.

Create a User Access Token and log in to your account.

from huggingface_hub import notebook_login

notebook_login()

Install a machine learning framework.

!pip install torch

Then install an up-to-date version of Transformers and some additional libraries from the Hugging Face ecosystem for accessing datasets and vision models, evaluating training, and optimizing training for large models.

!pip install -U transformers datasets evaluate accelerate timm

Pretrained models

Each pretrained model inherits from three base classes.

We recommend using the AutoClass API to load models and preprocessors because it automatically infers the appropriate architecture for each task and machine learning framework based on the name or path to the pretrained weights and configuration file.

Use from_pretrained() to load the weights and configuration file from the Hub into the model and preprocessor class.

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf", torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")

model_inputs = tokenizer(["The secret to baking a good cake is "], return_tensors="pt").to("cuda")

generated_ids = model.generate(**model_inputs, max_length=30)
tokenizer.batch_decode(generated_ids)[0]
'<s> The secret to baking a good cake is 100% in the preparation. There are so many recipes out there,'

Pipeline

The Pipeline class is the most convenient way to inference with a pretrained model. It supports many tasks such as text generation, image segmentation, automatic speech recognition, document question answering, and more.

Create a Pipeline object and select a task. By default, Pipeline downloads and caches a default pretrained model for a given task. Pass the model name to the model parameter to choose a specific model.

from transformers import pipeline

pipeline = pipeline("text-generation", model="meta-llama/Llama-2-7b-hf", device="cuda")

pipeline("The secret to baking a good cake is ", max_length=50)
[{'generated_text': 'The secret to baking a good cake is 100% in the batter. The secret to a great cake is the icing.\nThis is why we’ve created the best buttercream frosting reci'}]

Trainer

Trainer is a complete training and evaluation loop for PyTorch models. It abstracts away a lot of the boilerplate usually involved in manually writing a training loop, so you can start training faster and focus on training design choices. You only need a model, dataset, a preprocessor, and a data collator to build batches of data from the dataset.

Use the TrainingArguments class to customize the training process. It provides many options for training, evaluation, and more. Experiment with training hyperparameters and features like batch size, learning rate, mixed precision, torch.compile, and more to meet your training needs. You could also use the default training parameters to quickly produce a baseline.

Load a model, tokenizer, and dataset for training.

from transformers import AutoModelForSequenceClassification, AutoTokenizer
from datasets import load_dataset

model = AutoModelForSequenceClassification.from_pretrained("distilbert/distilbert-base-uncased")
tokenizer = AutoTokenizer.from_pretrained("distilbert/distilbert-base-uncased")
dataset = load_dataset("rotten_tomatoes")

Create a function to tokenize the text and convert it into PyTorch tensors. Apply this function to the whole dataset with the map method.

def tokenize_dataset(dataset):
    return tokenizer(dataset["text"])
dataset = dataset.map(tokenize_dataset, batched=True)

Load a data collator to create batches of data and pass the tokenizer to it.

from transformers import DataCollatorWithPadding

data_collator = DataCollatorWithPadding(tokenizer=tokenizer)

Next, set up TrainingArguments with the training features and hyperparameters.

from transformers import TrainingArguments

training_args = TrainingArguments(
    output_dir="distilbert-rotten-tomatoes",
    learning_rate=2e-5,
    per_device_train_batch_size=8,
    per_device_eval_batch_size=8,
    num_train_epochs=2,
    push_to_hub=True,
)

Finally, pass all these separate components to Trainer and call train() to start.

from transformers import Trainer

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=dataset["train"],
    eval_dataset=dataset["test"],
    tokenizer=tokenizer,
    data_collator=data_collator,
)

trainer.train()

Share your model and tokenizer to the Hub with push_to_hub().

trainer.push_to_hub()

Congratulations, you just trained your first model with Transformers!

TensorFlow

Trainer doesn’t work with TensorFlow models, but you can still train a Transformers model implemented in TensorFlow with Keras. Transformers TensorFlow models are a standard tf.keras.Model, which is compatible with Keras’ compile and fit methods.

Load a model, tokenizer, and dataset for training.

from transformers import TFAutoModelForSequenceClassification, AutoTokenizer

model = TFAutoModelForSequenceClassification.from_pretrained("distilbert/distilbert-base-uncased")
tokenizer = AutoTokenizer.from_pretrained("distilbert/distilbert-base-uncased")

Create a function to tokenize the text and convert it into TensorFlow tensors. Apply this function to the whole dataset with the map method.

def tokenize_dataset(dataset):
    return tokenizer(dataset["text"])
dataset = dataset.map(tokenize_dataset)

Transformers provides the prepare_tf_dataset() method to collate and batch a dataset.

tf_dataset = model.prepare_tf_dataset(
    dataset["train"], batch_size=16, shuffle=True, tokenizer=tokenizer
)

Finally, call compile to configure the model for training and fit to start.

from tensorflow.keras.optimizers import Adam

model.compile(optimizer="adam")
model.fit(tf_dataset)

Next steps

Now that you have a better understanding of Transformers and what it offers, it’s time to keep exploring and learning what interests you the most.

• Base classes: Learn more about the configuration, model and processor classes. This will help you understand how to create and customize models, preprocess different types of inputs (audio, images, multimodal), and how to share your model.
• Inference: Explore the Pipeline further, inference and chatting with LLMs, agents, and how to optimize inference with your machine learning framework and hardware.
• Training: Study the Trainer in more detail, as well as distributed training and optimizing training on specific hardware.
• Quantization: Reduce memory and storage requirements with quantization and speed up inference by representing weights with fewer bits.
• Resources: Looking for end-to-end recipes for how to train and inference with a model for a specific task? Check out the task recipes!