Bitdeer AI Cloud Python SDK

Overview

The Bitdeer AI Cloud Python SDK provides a simple and efficient interface for managing cloud resources and services, like training jobs. It allows users to create, list, get details, and manage training jobs with ease. This SDK communicates with the server using gRPC and provides a range of functionalities for handling training jobs, including creation, retrieval, listing, deletion, suspension, and resumption.

Installation

To install the Bitdeer AI Cloud Python SDK, you can use pip:

pip install bitdeer-ai

Usage of Training Service

Initialization

To interact with training service, you need to initialize the TrainingClient object with the host address of target host and an API Key for authentication.

from bitdeer_ai.training.client import TrainingClient

# Initialize the client
client = TrainingClient(host='api.bitdeer.ai:443', token='API-KEY')

Creating a Training Job

To create a training job, use the create_training_job method. You need to provide various parameters such as project_id, job_name, job_type, worker_spec, num_workers, and optional parameters like worker_image, working_dir, volume_name, volume_mount_path etc.

from training.training_pb2 import JobType

job = client.create_training_job(
    project_id='your_project_id',
    job_name='example_job',
    job_type='your_job_type',
    worker_spec='spec_of_worker',
    num_workers=2,
    worker_image='worker_image_url',
    working_dir='/path/to/working/dir',
    volume_name='volume_name',
    volume_mount_path='/mount/path'
)
print(f'Training job created with ID: {job.training_job_id}')

Retrieving a Training Job

To retrieve details of a specific training job, use the get_training_job method with the training_job_id.

job = client.get_training_job(training_job_id='your_training_job_id')
print(f'Job Name: {job.job_name}')

Listing Training Jobs

To list all training jobs, use the list_training_jobs method.

jobs = client.list_training_jobs()
for job in jobs.training_jobs:
    print(f'Job ID: {job.training_job_id}, Job Name: {job.job_name}')

Deleting a Training Job

To delete a specific training job, use the delete_training_job method with the training_job_id.

client.delete_training_job(training_job_id='your_training_job_id')
print('Training job deleted successfully.')

Suspending a Training Job

To suspend an active training job, use the suspend_training_job method with the training_job_id.

client.suspend_training_job(training_job_id='your_training_job_id')
print('Training job suspended successfully.')

Resuming a Training Job

To resume a suspended training job, use the resume_training_job method with the training_job_id.

client.resume_training_job(training_job_id='your_training_job_id')
print('Training job resumed successfully.')

Getting Training Job Workers

To get details of workers associated with a specific training job, use the get_training_job_workers method with the training_job_id.

workers = client.get_training_job_workers(training_job_id='your_training_job_id')
for worker in workers.workers:
    print(f'Worker Name: {worker.name}')

Getting Training Job Logs

To stream logs of a specific training job, use the get_training_job_logs method with the training_job_id, worker_name, and follow flag.

logs = client.get_training_job_logs(training_job_id='your_training_job_id', worker_name='worker_name', follow=True)
for log in logs:
    print(log.message)

Error Handling

The SDK raises various exceptions to handle errors:

• RuntimeError: Raised when there is a failure in creating or deleting a training job.

Make sure to handle these exceptions in your code to ensure smooth operation.

try:
    job = client.create_training_job(
        project_id='your_project_id',
        job_name='example_job',
        job_type='your_job_type',
        worker_spec='spec_of_worker',
        num_workers=2
    )
except RuntimeError as e:
    print(f'Runtime Error: {e}')