context

class verta.runtime.context(validate: Optional[bool] = False)

Context manager for aggregating key-value pairs into a custom log entry.

This class should not be instantiated directly in your model code. For all models deployed in Verta, the predict() method is wrapped inside a context by default.

For local testing, a single instance can be instantiated as in the provided example.

Parameters:

validate (bool, default False) – If True, each individual call to log() will verify that the value provided is JSON serializable.

Raises:

RuntimeError – If this context manager instance is nested inside the context of an existing instance.

Examples

 import json
 from verta import runtime

 with runtime.context(validate=True) as ctx:  # validate values are JSON-serializable
     my_model.predict(x)  # your model, with your calls to runtime.log()
     print(json.dumps(ctx.logs()))

 # Same output after exiting the context manager:
 print(json.dumps(ctx.logs()))

logs() → Dict[str, Any]

Return the currently stored, thread-local logs from inside this context manager. If called after exiting the context manager, the final collection of logs captured at the time of exit is returned.

Returns:

logs (Dict[str, Any]) – Dictionary of logs collected within this context manager.