Contact UsGet started

Using data from Object Storage to train a model in Yandex DataSphere

Written by
Updated at October 17, 2024

You will create a bucket in Object Storage, mount it to the Windows file system using rclone, and upload data to this bucket to train a model in Yandex DataSphere.

To use data from Object Storage to train a model in DataSphere:

  1. Prepare your infrastructure.
  2. Create a static access key.
  3. Create a bucket.
  4. Set up a connection to Object Storage.
  5. Mount the bucket.
  6. Prepare data for model training.
  7. Create an S3 connector.
  8. Attach your bucket to a project.
  9. Train the model.

If you no longer need the resources you created, delete them.

Getting started

Before getting started, register in Yandex Cloud, set up a community, and link your billing account to it.

  1. On the DataSphere home page, click Try for free and select an account to log in with: Yandex ID or your working account in the identity federation (SSO).
  2. Select the Yandex Cloud Organization organization you are going to use in Yandex Cloud.
  3. Create a community.
  4. Link your billing account to the DataSphere community you are going to work in. Make sure that you have a billing account linked and its status is ACTIVE or TRIAL_ACTIVE. If you do not have a billing account yet, create one in the DataSphere interface.

The cost of training a model based on data from Object Storage includes:

Prepare the infrastructure

Log in to the Yandex Cloud management console and select the organization you use to access DataSphere. On the Yandex Cloud Billing page, make sure you have a billing account linked.

If you have an active billing account, you can create or select a folder to deploy your infrastructure in, on the cloud page.

Note

If you use an identity federation to access Yandex Cloud, billing details might be unavailable to you. In this case, contact your Yandex Cloud organization administrator.

Set up a working environment

Download and install the winfsp distribution from the developer website.

Download it from the developer website and unpack to the working directory on your local computer:

Create a folder

  1. In the management console, select a cloud and click Create folder.
  2. Give your folder a name, e.g., data-folder.
  3. Click Create.

Create a service account for Object Storage

To access a bucket in Object Storage, you need a service account with the storage.editor role.

  1. In the management console, go to data-folder.
  2. In the Service accounts tab, click Create service account.
  3. Enter a name for the service account, e.g., datasphere-sa.
  4. Click Add role and assign the service account the storage.editor role.
  5. Click Create.

Create a static access key

To access Object Storage from DataSphere, you need a static key.

  1. In the management console, navigate to the folder the service account belongs to.
  2. At the top of the screen, go to the Service accounts tab.
  3. Select the datasphere-sa service account.
  4. In the top panel, click Create new key.
  5. Select Create static access key.
  6. Specify the static key description and click Create.
  7. Save the ID and private key. After you close the dialog, the private key value will become unavailable.

  1. Create a static access key for the datasphere-sa service account:

    yc iam access-key create --service-account-name datasphere-sa

    Result:

    access_key:
  id: aje6t3vsbj8l********
  service_account_id: ajepg0mjt06s********
  created_at: "2022-07-18T14:37:51Z"
  key_id: 0n8X6WY6S24N7Oj*****
secret: JyTRFdqw8t1kh2-OJNz4JX5ZTz9Dj1rI9hx*****

    For more information about the yc iam access-key create command, see the CLI reference.

  2. Save the key_id and the secret key. You will not be able to get the key value again.

To create a static access key, use the create method for the AccessKey resource.

Create a bucket

  1. In the management console, select the folder where you want to create your bucket.
  2. In the list of services, select Object Storage.
  3. At the top right, click Create bucket.
  4. In the ** Name** field, enter a name for the bucket consistent with the naming conventions.
  5. In the Object read access, Object listing access, and Read access to settings fields, select Restricted.
  6. Click Create bucket.

  1. If you do not have the AWS CLI yet, install and configure it.

  2. Enter the bucket name following the naming conventions:

    aws --endpoint-url https://storage.yandexcloud.net \
  s3 mb s3://<bucket_name>

    Result:

    make_bucket: <bucket_name>

Note

Terraform uses a service account to interact with Object Storage. Assign to the service account the required role, e.g., storage.admin, for the folder where you are going to create resources.

If you don't have Terraform, install it and configure the Yandex Cloud provider.

  1. Describe the parameters for creating a service account and access key in the configuration file:

    ...
// Creating a service account
resource "yandex_iam_service_account" "sa" {
  name = "<service_account_name>"
}

// Assigning a role to a service account
resource "yandex_resourcemanager_folder_iam_member" "sa-admin" {
  folder_id = "<folder_ID>"
  role      = "storage.admin"
  member    = "serviceAccount:${yandex_iam_service_account.sa.id}"
}

// Creating a static access key
resource "yandex_iam_service_account_static_access_key" "sa-static-key" {
  service_account_id = yandex_iam_service_account.sa.id
  description        = "static access key for object storage"
}

  2. Add a section with bucket parameters to the configuration file and enter the bucket name following the naming conventions:

    resource "yandex_storage_bucket" "<bucket_name>" {
  access_key = yandex_iam_service_account_static_access_key.sa-static-key.access_key
  secret_key = yandex_iam_service_account_static_access_key.sa-static-key.secret_key
  bucket     = "<bucket_name>"
}

    For more information about the yandex_storage_bucket resource, see the Terraform provider documentation.

  3. Create resources:

    1. In the terminal, change to the folder where you edited the configuration file.

    2. Make sure the configuration file is correct using the command:

      terraform validate

      If the configuration is correct, the following message is returned:

      Success! The configuration is valid.

    3. Run the command:

      terraform plan

      The terminal will display a list of resources with parameters. No changes are made at this step. If the configuration contains errors, Terraform will point them out.

    4. Apply the configuration changes:

      terraform apply

    5. Confirm the changes: type yes in the terminal and press Enter.

    Terraform will create all the required resources. You can check the new resources and their configuration using the management console.

Use the create REST API method for the Bucket resource, the BucketService/Create gRPC API call, or the create S3 API method.

Set up a connection to Object Storage

To move data from your local disk to Object Storage, configure the rclone utility.

  1. Open the command line in your working directory as an admin user and run the following command:

    .\psexec -i -s cmd.exe

  2. In the console that opens, run whoami and make sure the session has been started by the system user.

  3. Go to the working directory and run a configuration session for rclone:

    rclone.exe config

  4. Follow the prompts to create a new connection profile:

    1. Start creating a new profile by entering n in the terminal.
    2. Enter the connection name: s3-connect.
    3. Select the storage type by entering 4 in the terminal.
    4. Select a provider by entering 1 in the terminal.
    5. Select manual entry of credentials by entering 1 in the terminal.
    6. Enter the secret key ID in the terminal.
    7. Enter the secret key value in the terminal.
    8. Specify the region by entering ru-central1 in the terminal.
    9. Specify the endpoint by entering storage.yandexcloud.net in the terminal.
    10. You can leave the default values for all other settings by pressing Enter to skip them.

Note

You can perform advanced connection setup if needed. To do this, type y at the Edit advanced config? step. For more information about advanced settings, see the rclone documentation.

Mount the bucket to the Windows file system

  1. Check your connection to the bucket. In the same command line you used to configure the connection, run the following command with the bucket name specified:

    rclone.exe ls s3-connect:<bucket_name>

    If the configuration is set up correctly, the objects in the bucket will be listed in the console.

  2. Mount the bucket to the file system specifying the bucket name and an available drive letter in the file system:

    rclone.exe mount s3-connect:<bucket_name> <disk_letter>: --vfs-cache-mode full

    You will see a new disk with the objects from the bucket in Windows Explorer.

Prepare data for model training

  1. Download a CSV file with diabetes prediction data. A dataset is created based on Kaggle data for 2015.
  2. Copy the file to the mounted bucket.

Attach your bucket to a project

To connect to the bucket from DataSphere, you need an S3 connector:

  1. Select the relevant project in your community or on the DataSphere homepage in the Recent projects tab.

  2. In the top-right corner, click Create resource. In the pop-up window, select S3 Connector.
  3. Fill in the fields as follows:
    • Name: Name of the new connector, e.g., s3-datasphere-connect.
    • Endpoint: Object Storage host, which is https://storage.yandexcloud.net/.
    • Bucket: Name of your bucket.
    • Mount name: Name of the volume for mounting the bucket into the project file system.
    • Static access key ID used to connect to storage.
    • In the Static access key field, click Create. In the window that opens, enter the secret name and the secret key being used to connect to storage.
  4. Click Create.
  5. Go to the S3 connector page and click Activate. Once connected, the bucket will be available in the JupyterLab UI in the list on the S3 Mounts tab, and you will be able to view it as a file system.

Train the model

Clone the Git repository containing the diabetes_catboost.ipynb notebook with a sample for CatBoost model training:

  1. Open the DataSphere project:

    1. Select the relevant project in your community or on the DataSphere homepage in the Recent projects tab.

    2. Click Open project in JupyterLab and wait for the loading to complete.
    3. Open the notebook tab.
  2. In the top menu, click Git and select Clone.
  3. In the window that opens, enter https://github.com/yandex-cloud-examples/yc-datasphere-s3-mount.git and click Clone.

In the diabetes_catboost.ipynb notebook, you will connect to the diabetes_data.csv dataset in the bucket you mounted and train the CatBoost model:

  1. Install the required modules:

    %pip install catboost
%pip install scikit-learn
%pip install ipywidgets

  2. Import the required libraries:

    import pandas as pd
from catboost import CatBoostClassifier
from sklearn.metrics import classification_report
from sklearn.model_selection import train_test_split

  3. Upload data for model training:

    data = pd.read_csv('<path_to_dataset>')
data.head()

    To get the diabetes_data.csv file path, open the S3 Mounts tab, right-click it, and select Copy path.

  4. Separate training factors from the target variable:

    X = data.loc[:, data.columns != 'Diabetes_binary']
y = data['Diabetes_binary']

  5. Divide the sample into training and test ones:

    X_train, X_validation, y_train, y_validation = train_test_split(X, y, train_size=0.75, random_state=42)

  6. Create a classifier named СatBoost and set hyperparameters:

    model = CatBoostClassifier(
  iterations=100,
  random_seed=42,
  logging_level='Verbose'
)

  7. Start model training:

    model.fit(
  X_train, y_train,
  eval_set=(X_validation, y_validation),
  plot=True
)

  8. Output the trained model quality metrics:

    print(classification_report(y_validation, model.predict(X_validation)))

How to delete the resources you created

To stop paying for the resources you created:

Previous
Regular recognition of audio files in a bucket
Next
Connecting to Object Storage from VPC