Your own PyTorch code must run on managed infrastructure. What fits?
Why C is the answer
Agent Platform custom training
Custom training runs your code on managed infrastructure.
In detail
- Custom training runs your own PyTorch code on managed infrastructure.
- You keep your code, and hand over the infrastructure.
Use case
An illustrative scenario, not a real incident.
A team has PyTorch code, and custom training runs it on managed machines.
See it in different ways
stateDiagram-v2 [*] --> Task Task --> PrebuiltApi: a common task Task --> BigQueryMl: data in BigQuery, SQL team Task --> AutoML: custom model, no code Task --> CustomTraining: your own code
flowchart LR S0["Package code"] --> S1["Submit a training job"] --> S2["Collect the model"]
Key point: Custom training runs your code.
Why the other options are wrong
- A. Model Garden one-click deployment
- That deploys existing models.
- B. AutoML
- AutoML does not run your code.
- D. BigQuery ML with a SQL statement
- BigQuery ML does not run your code.