Run training on Amazon SageMaker

This guide will show you how to train a 🤗 Transformers model with the SageMaker Python SDK. Learn how to:

Install and setup your training environment.
Prepare a training script.
Create a ModelTrainer.
Run training with the train method.
Access your trained model.
Perform distributed training.
Create a spot instance.
Load a training script from a GitHub repository.
Collect training metrics.

Installation and setup

Before you can train a 🤗 Transformers model with SageMaker, you need to sign up for an AWS account. If you don’t have an AWS account yet, learn more here.

Once you have an AWS account, get started using one of the following:

To start training locally, you need to setup an appropriate IAM role.

Upgrade to the latest sagemaker version:

pip install 'sagemaker>=3.0.0'

These docs and examples use the SageMaker Python SDK v3, which introduces a new framework-agnostic API built around ModelTrainer (training) and ModelBuilder (inference), replacing the v2 HuggingFace and HuggingFaceModel classes. Install it with pip install "sagemaker>=3.0.0".

SageMaker environment

Setup your SageMaker environment as shown below:

from sagemaker.core.helper.session_helper import Session, get_execution_role
sess = Session()
role = get_execution_role()

Note: The execution role is only available when running a notebook within SageMaker. If you run get_execution_role in a notebook not on SageMaker, expect a region error.

Local environment

Setup your local environment as shown below:

import boto3
from sagemaker.core.helper.session_helper import Session

iam_client = boto3.client('iam')
role = iam_client.get_role(RoleName='role-name-of-your-iam-role-with-right-permissions')['Role']['Arn']
sess = Session()

Prepare a 🤗 Transformers fine-tuning script

Our training script is very similar to a training script you might run outside of SageMaker. However, you can access useful properties about the training environment through various environment variables (see here for a complete list), such as:

SM_MODEL_DIR: A string representing the path to which the training job writes the model artifacts. After training, artifacts in this directory are uploaded to S3 for model hosting. SM_MODEL_DIR is always set to /opt/ml/model.
SM_NUM_GPUS: An integer representing the number of GPUs available to the host.
SM_CHANNEL_XXXX: A string representing the path to the directory that contains the input data for the specified channel. For example, when you specify train and test channels in the ModelTrainer via input_data_config, the environment variables are set to SM_CHANNEL_TRAIN and SM_CHANNEL_TEST.

The hyperparameters defined in the ModelTrainer are passed as named arguments and processed by ArgumentParser().

import transformers
import datasets
import argparse
import os

if __name__ == "__main__":

    parser = argparse.ArgumentParser()

    # hyperparameters sent by the client are passed as command-line arguments to the script
    parser.add_argument("--epochs", type=int, default=3)
    parser.add_argument("--per_device_train_batch_size", type=int, default=32)
    parser.add_argument("--model_name_or_path", type=str)

    # data, model, and output directories
    parser.add_argument("--model-dir", type=str, default=os.environ["SM_MODEL_DIR"])
    parser.add_argument("--training_dir", type=str, default=os.environ["SM_CHANNEL_TRAIN"])
    parser.add_argument("--test_dir", type=str, default=os.environ["SM_CHANNEL_TEST"])

Note that SageMaker doesn’t support argparse actions. For example, if you want to use a boolean hyperparameter, specify type as bool in your script and provide an explicit True or False value.

Look train.py file for a complete example of a 🤗 Transformers training script.

Training Output Management

If output_dir in the TrainingArguments is set to ‘/opt/ml/model’ the Trainer saves all training artifacts, including logs, checkpoints, and models. Amazon SageMaker archives the whole ‘/opt/ml/model’ directory as model.tar.gz and uploads it at the end of the training job to Amazon S3. Depending on your Hyperparameters and TrainingArguments this could lead to a large artifact (> 5GB), which can slow down deployment for Amazon SageMaker Inference. You can control how checkpoints, logs, and artifacts are saved by customization the TrainingArguments. For example by providing save_total_limit as TrainingArgument you can control the limit of the total amount of checkpoints. Deletes the older checkpoints in output_dir if new ones are saved and the maximum limit is reached.

In addition to the options already mentioned above, there is another option to save the training artifacts during the training session. Amazon SageMaker supports Checkpointing, which allows you to continuously save your artifacts during training to Amazon S3 rather than at the end of your training. To enable Checkpointing you need to provide a CheckpointConfig(s3_uri=...) pointing to an Amazon S3 location on the ModelTrainer and set output_dir to /opt/ml/checkpoints. Note: If you set output_dir to /opt/ml/checkpoints make sure to call trainer.save_model("/opt/ml/model") or model.save_pretrained(“/opt/ml/model”)/tokenizer.save_pretrained("/opt/ml/model") at the end of your training to be able to deploy your model seamlessly to Amazon SageMaker for Inference.

Create a ModelTrainer

Run 🤗 Transformers training scripts on SageMaker by creating a ModelTrainer. The ModelTrainer handles end-to-end SageMaker training. There are several parameters you should define:

source_code specifies the fine-tuning script (entry_script) and its directory (source_dir).
compute specifies the Amazon instance(s) to launch. Refer here for a complete list of instance types.
training_image is the training container image, retrieved with image_uris.retrieve.
hyperparameters specifies training hyperparameters. View additional available hyperparameters in train.py file.

The following code sample shows how to train with a custom script train.py with three hyperparameters (epochs, per_device_train_batch_size, and model_name_or_path):

from sagemaker.train.model_trainer import ModelTrainer
from sagemaker.train.configs import SourceCode, Compute
from sagemaker.core import image_uris
from sagemaker.core.helper.session_helper import Session, get_execution_role

# set up the SageMaker session and execution role
sess = Session()
role = get_execution_role()

# hyperparameters which are passed to the training job (as `--key value` CLI args)
hyperparameters = {
    'epochs': 1,
    'per_device_train_batch_size': 32,
    'model_name_or_path': 'distilbert-base-uncased',
}

instance_type = 'ml.g6.12xlarge'

# Retrieve the Hugging Face PyTorch training DLC image URI
training_image = image_uris.retrieve(
    framework="huggingface",
    region=sess.boto_region_name,
    version="4.49.0",
    base_framework_version="pytorch2.5.1",
    py_version="py311",
    image_scope="training",
    instance_type=instance_type,
)

# create the ModelTrainer
huggingface_estimator = ModelTrainer(
    sagemaker_session=sess,
    role=role,
    training_image=training_image,
    source_code=SourceCode(
        source_dir='./scripts',
        entry_script='train.py',
    ),
    compute=Compute(
        instance_type=instance_type,
        instance_count=1,
    ),
    hyperparameters=hyperparameters,
)

If you are running a TrainingJob locally, define instance_type='local' or instance_type='local_gpu' for GPU usage. Note that this will not work with SageMaker Studio.

Execute training

Start your TrainingJob by calling train on a ModelTrainer. Specify your input training data as channels via input_data_config. Each channel’s data_source can be a:

S3 URI such as s3://my-bucket/my-training-data.
FileSystemInput for Amazon Elastic File System or FSx for Lustre.

Each channel is mounted inside the container at /opt/ml/input/data/<channel_name>. Call train to begin training:

from sagemaker.train.configs import InputData

huggingface_estimator.train(
    input_data_config=[
        InputData(channel_name="train", data_source="s3://<your-bucket>/imdb/train"),
        InputData(channel_name="test", data_source="s3://<your-bucket>/imdb/test"),
    ]
)

SageMaker starts and manages all the required EC2 instances and initiates the TrainingJob by running:

/opt/conda/bin/python train.py --epochs 1 --model_name_or_path distilbert-base-uncased --per_device_train_batch_size 32

Access trained model

Once training is complete, you can access your model through the AWS console or download it directly from S3. The S3 URI of the trained model artifacts is available on the completed training job:

import boto3
from urllib.parse import urlparse

# S3 URI where the trained model artifacts (model.tar.gz) are located
model_data = huggingface_estimator._latest_training_job.model_artifacts.s3_model_artifacts

parsed = urlparse(model_data)
boto3.client("s3").download_file(
    parsed.netloc,                 # bucket
    parsed.path.lstrip("/"),       # key
    "model.tar.gz",                # local path where the artifact is saved
)

Distributed training

SageMaker provides two strategies for distributed training: data parallelism and model parallelism. Data parallelism splits a training set across several GPUs, while model parallelism splits a model across several GPUs.

Data parallelism

The Hugging Face Trainer supports distributed data parallel training. With ModelTrainer you launch your script with torchrun by passing a Torchrun config to the distributed parameter. Set process_count_per_node to the number of GPUs per instance (ml.p3dn.24xlarge has 8):

from sagemaker.train.model_trainer import ModelTrainer
from sagemaker.train.configs import SourceCode, Compute
from sagemaker.train.distributed import Torchrun
from sagemaker.core import image_uris
from sagemaker.core.helper.session_helper import Session, get_execution_role

# set up the SageMaker session and execution role
sess = Session()
role = get_execution_role()

# hyperparameters which are passed to the training job (as `--key value` CLI args)
hyperparameters = {
    'epochs': 1,
    'per_device_train_batch_size': 32,
    'model_name_or_path': 'distilbert-base-uncased',
}

instance_type = 'ml.p3dn.24xlarge'

training_image = image_uris.retrieve(
    framework="huggingface",
    region=sess.boto_region_name,
    version="4.49.0",
    base_framework_version="pytorch2.5.1",
    py_version="py311",
    image_scope="training",
    instance_type=instance_type,
)

# create the ModelTrainer with torchrun for distributed data parallelism
huggingface_estimator = ModelTrainer(
    sagemaker_session=sess,
    role=role,
    training_image=training_image,
    source_code=SourceCode(source_dir='./scripts', entry_script='train.py'),
    compute=Compute(instance_type=instance_type, instance_count=2),
    distributed=Torchrun(process_count_per_node=8),
    hyperparameters=hyperparameters,
)

📓 Open the sagemaker-notebook.ipynb notebook for an example of how to run the data parallelism library with TensorFlow.

Model parallelism

The Hugging Face [Trainer] also supports model parallelism through the SageMaker Model Parallelism library (SMP). With ModelTrainer you enable it by passing an SMP config to Torchrun. SMP provides tensor parallelism, context parallelism and sharded data parallelism:

from sagemaker.train.model_trainer import ModelTrainer
from sagemaker.train.configs import SourceCode, Compute
from sagemaker.train.distributed import Torchrun, SMP
from sagemaker.core import image_uris
from sagemaker.core.helper.session_helper import Session, get_execution_role

# set up the SageMaker session and execution role
sess = Session()
role = get_execution_role()

# hyperparameters which are passed to the training job (as `--key value` CLI args)
hyperparameters = {
    'epochs': 1,
    'per_device_train_batch_size': 32,
    'model_name_or_path': 'distilbert-base-uncased',
}

instance_type = 'ml.p3dn.24xlarge'

training_image = image_uris.retrieve(
    framework="huggingface",
    region=sess.boto_region_name,
    version="4.49.0",
    base_framework_version="pytorch2.5.1",
    py_version="py311",
    image_scope="training",
    instance_type=instance_type,
)

# create the ModelTrainer with torchrun + SMP for model parallelism
huggingface_estimator = ModelTrainer(
    sagemaker_session=sess,
    role=role,
    training_image=training_image,
    source_code=SourceCode(source_dir='./scripts', entry_script='train.py'),
    compute=Compute(instance_type=instance_type, instance_count=2),
    distributed=Torchrun(
        process_count_per_node=8,
        smp=SMP(
            tensor_parallel_degree=2,
            hybrid_shard_degree=1,
        ),
    ),
    hyperparameters=hyperparameters,
)

📓 Open the sagemaker-notebook.ipynb notebook for an example of how to run the model parallelism library.

Spot instances

The Hugging Face extension for the SageMaker Python SDK means we can benefit from fully-managed EC2 spot instances. This can help you save up to 90% of training costs!

Note: Unless your training job completes quickly, we recommend you use checkpointing with managed spot training. In this case, you need to define the checkpoint_s3_uri.

Set enable_managed_spot_training=True on Compute and define max_wait_time_in_seconds and max_runtime_in_seconds on StoppingCondition to use spot instances:

from sagemaker.train.model_trainer import ModelTrainer
from sagemaker.train.configs import SourceCode, Compute, StoppingCondition, CheckpointConfig
from sagemaker.core import image_uris
from sagemaker.core.helper.session_helper import Session, get_execution_role

# set up the SageMaker session and execution role
sess = Session()
role = get_execution_role()

# hyperparameters which are passed to the training job
hyperparameters = {
    'epochs': 1,
    'train_batch_size': 32,
    'model_name': 'distilbert-base-uncased',
    'output_dir': '/opt/ml/checkpoints',
}

instance_type = 'ml.g6.12xlarge'

training_image = image_uris.retrieve(
    framework="huggingface",
    region=sess.boto_region_name,
    version="4.49.0",
    base_framework_version="pytorch2.5.1",
    py_version="py311",
    image_scope="training",
    instance_type=instance_type,
)

# create the ModelTrainer
huggingface_estimator = ModelTrainer(
    sagemaker_session=sess,
    role=role,
    training_image=training_image,
    source_code=SourceCode(source_dir='./scripts', entry_script='train.py'),
    compute=Compute(
        instance_type=instance_type,
        instance_count=1,
        enable_managed_spot_training=True,   # use fully-managed spot instances
    ),
    # max_wait_time_in_seconds should be equal to or greater than max_runtime_in_seconds
    stopping_condition=StoppingCondition(
        max_runtime_in_seconds=1000,
        max_wait_time_in_seconds=3600,
    ),
    checkpoint_config=CheckpointConfig(s3_uri=f's3://{sess.default_bucket()}/checkpoints'),
    hyperparameters=hyperparameters,
)

# Training seconds: 874
# Billable seconds: 262
# Managed Spot Training savings: 70.0%

📓 Open the sagemaker-notebook.ipynb notebook for an example of how to use spot instances.

Git repository

The v2 git_config parameter is not available in ModelTrainer. To run a training script that lives in a GitHub repository (such as the 🤗 Transformers example scripts), clone the repository locally first and point source_dir/entry_script at the checked-out files. Choose a branch that matches the Transformers version of your training image.

Tip: Save your model to S3 by setting output_dir=/opt/ml/model in the hyperparameter of your training script.

# clone the repo locally, matching the transformers version of your training image
git clone --branch v4.49.0 https://github.com/huggingface/transformers.git

from sagemaker.train.model_trainer import ModelTrainer
from sagemaker.train.configs import SourceCode, Compute
from sagemaker.core import image_uris
from sagemaker.core.helper.session_helper import Session, get_execution_role

# set up the SageMaker session and execution role
sess = Session()
role = get_execution_role()

# hyperparameters which are passed to the training job (as `--key value` CLI args)
hyperparameters = {
    'epochs': 1,
    'per_device_train_batch_size': 32,
    'model_name_or_path': 'distilbert-base-uncased',
}

instance_type = 'ml.g6.12xlarge'

# Retrieve the Hugging Face PyTorch training DLC image URI
training_image = image_uris.retrieve(
    framework="huggingface",
    region=sess.boto_region_name,
    version="4.49.0",
    base_framework_version="pytorch2.5.1",
    py_version="py311",
    image_scope="training",
    instance_type=instance_type,
)

# create the ModelTrainer pointing at the cloned example directory
huggingface_estimator = ModelTrainer(
    sagemaker_session=sess,
    role=role,
    training_image=training_image,
    source_code=SourceCode(
        source_dir='transformers/examples/pytorch/text-classification',
        entry_script='run_glue.py',
        requirements='requirements.txt',
    ),
    compute=Compute(instance_type='ml.g6.12xlarge', instance_count=1),
    hyperparameters=hyperparameters,
)

SageMaker metrics

SageMaker metrics automatically parses training job logs for metrics and sends them to CloudWatch. If you want SageMaker to parse the logs, you must specify the metric’s name and a regular expression for SageMaker to use to find the metric. With ModelTrainer you attach them using with_metric_definitions:

from sagemaker.train.model_trainer import ModelTrainer
from sagemaker.train.configs import SourceCode, Compute, MetricDefinition
from sagemaker.core import image_uris
from sagemaker.core.helper.session_helper import Session, get_execution_role

# set up the SageMaker session and execution role
sess = Session()
role = get_execution_role()

# hyperparameters which are passed to the training job (as `--key value` CLI args)
hyperparameters = {
    'epochs': 1,
    'per_device_train_batch_size': 32,
    'model_name_or_path': 'distilbert-base-uncased',
}

instance_type = 'ml.g6.12xlarge'

training_image = image_uris.retrieve(
    framework="huggingface",
    region=sess.boto_region_name,
    version="4.49.0",
    base_framework_version="pytorch2.5.1",
    py_version="py311",
    image_scope="training",
    instance_type=instance_type,
)

# define metrics definitions
metric_definitions = [
    MetricDefinition(name="train_runtime", regex="train_runtime.*=\D*(.*?)$"),
    MetricDefinition(name="eval_accuracy", regex="eval_accuracy.*=\D*(.*?)$"),
    MetricDefinition(name="eval_loss", regex="eval_loss.*=\D*(.*?)$"),
]

# create the ModelTrainer
huggingface_estimator = ModelTrainer(
    sagemaker_session=sess,
    role=role,
    training_image=training_image,
    source_code=SourceCode(source_dir='./scripts', entry_script='train.py'),
    compute=Compute(instance_type=instance_type, instance_count=1),
    hyperparameters=hyperparameters,
).with_metric_definitions(metric_definitions)

📓 Open the notebook for an example of how to capture metrics in SageMaker.

Update on GitHub

Deploying on AWS