Parameter
Defines a parameter for a flow.
Usage
from metaflow import FlowSpec, Parameter, step
class MyFlow(FlowSpec):
alpha = Parameter('alpha',
help='Learning rate',
default=0.01)
@step
def start(self):
print(f"Alpha value: {self.alpha}")
self.next(self.end)
@step
def end(self):
pass
python myflow.py run --alpha=0.05
Description
Parameters must be instantiated as class variables in flow classes. The parameter is specified on the command line and its value is accessible through a read-only artifact.
Note that the user-visible parameter name can be different from the artifact name.
class MyFlow(FlowSpec):
param = Parameter('myparam') # CLI: --myparam
type argument or to match the type of default, if it is set.
Constructor
Parameter(
name: str,
default: Optional[Union[str, float, int, bool, Dict[str, Any], Callable]] = None,
type: Optional[Union[Type[str], Type[float], Type[int], Type[bool], JSONTypeClass]] = None,
help: Optional[str] = None,
required: Optional[bool] = None,
show_default: Optional[bool] = None,
**kwargs: Dict[str, Any]
)
User-visible parameter name.
default
str | float | int | bool | Dict[str, Any] | Callable
default:"None"
Default value for the parameter. Use a special JSONType class to
indicate that the value must be a valid JSON object. A function
implies that the parameter corresponds to a deploy-time parameter.
The type of the default value is used as the parameter type.
If default is not specified, define the parameter type. Specify
one of str, float, int, bool, or JSONType. If None, defaults
to the type of default or str if none specified.
Help text to show in run --help.
Require that the user specifies a value for the parameter. Note that if
a default is provided, the required flag is ignored.
A value of None is equivalent to False.
If True, show the default value in the help text. A value of None is equivalent
to True.
Examples
String parameter
from metaflow import FlowSpec, Parameter, step
class MyFlow(FlowSpec):
name = Parameter('name',
help='User name',
default='World')
@step
def start(self):
print(f"Hello, {self.name}!")
self.next(self.end)
@step
def end(self):
pass
Numeric parameters
from metaflow import FlowSpec, Parameter, step
class MyFlow(FlowSpec):
alpha = Parameter('alpha',
type=float,
help='Learning rate',
default=0.01)
epochs = Parameter('epochs',
type=int,
help='Number of epochs',
default=10)
@step
def start(self):
print(f"Training for {self.epochs} epochs with alpha={self.alpha}")
self.next(self.end)
@step
def end(self):
pass
Boolean parameter
from metaflow import FlowSpec, Parameter, step
class MyFlow(FlowSpec):
verbose = Parameter('verbose',
type=bool,
help='Enable verbose output',
default=False)
@step
def start(self):
if self.verbose:
print("Verbose mode enabled")
self.next(self.end)
@step
def end(self):
pass
JSON parameter
from metaflow import FlowSpec, Parameter, JSONType, step
class MyFlow(FlowSpec):
config = Parameter('config',
type=JSONType,
help='Configuration object',
default='{"key": "value"}')
@step
def start(self):
print(f"Config: {self.config}")
print(f"Key value: {self.config['key']}")
self.next(self.end)
@step
def end(self):
pass
python myflow.py run --config='{"key": "new_value", "count": 42}'
Required parameter
from metaflow import FlowSpec, Parameter, step
class MyFlow(FlowSpec):
input_file = Parameter('input-file',
help='Path to input file',
required=True)
@step
def start(self):
print(f"Processing {self.input_file}")
self.next(self.end)
@step
def end(self):
pass
Deploy-time parameter
Deploy-time parameters are evaluated when deploying to a scheduler (e.g., AWS Step Functions):
from metaflow import FlowSpec, Parameter, step
import datetime
class MyFlow(FlowSpec):
deploy_time = Parameter('deploy-time',
default=lambda ctx: datetime.datetime.now().isoformat())
@step
def start(self):
print(f"Deployed at: {self.deploy_time}")
self.next(self.end)
@step
def end(self):
pass
Properties
The user-visible name of the parameter.
If set, the parameter value will be split by this separator.
Only valid for string parameters.
Returns True if the parameter is a string type.
Special Types
JSONType
Use JSONType to accept JSON objects as parameters:
from metaflow import JSONType, Parameter
config = Parameter('config', type=JSONType)
