Add accelerator_count for VertexAICustomTrainingJob

[jeremy-thomas-roc]

The VertexAICustomTrainingJob class is missing the accelerator_count property, making it impossible to actually attach an accelerator to a custom job, due to the MachineSpec being invalid.

This PR adds the property and passes it to the _build_job_spec method to ensure it is passed along to Google properly.

Example

vertex_ai_job = VertexAICustomTrainingJob(
    type="vertex-ai-custom-training-job",
    image=IMAGE_URI,
    credentials=GcpCredentials.load(GCP_CREDENTIALS_BLOCK_NAME),
    region="us-central1",
    network=NETWORK,
    service_account=SERVICE_ACCOUNT,
    machine_type="n1-standard-8",
    accelerator_type="NVIDIA_TESLA_T4",
    accelerator_count=1
)

Screenshots

The documentation on this class is limited as is, these changes do not reduce their usability or accuracy.

Closes #175

Checklist

• References any related issue by including "Closes #" or "Closes ".
  • If no issue exists and your change is not a small fix, please create an issue first.
• Includes tests or only affects documentation.
• Passes pre-commit checks.
  • Run pre-commit install && pre-commit run --all locally for formatting and linting.
• Includes screenshots of documentation updates.
  • Run mkdocs serve view documentation locally.
• Summarizes PR's changes in CHANGELOG.md

[desertaxle]

Thanks for the contribution @jeremy-thomas-roc! Would you be able to add a test covering this new functionality? A test that checks that the machine spec is correctly created when a accelerator type and count are provided should suffice.

[jeremy-thomas-roc]

@desertaxle added a simple test to make sure the values are passed through _build_job_spec, is this enough, or did you also want a test to ensure the .run() works as well?

[desertaxle]

LGTM!