temporian.save_graph #
save_graph(
inputs: Optional[graph.NamedEventSetNodes],
outputs: graph.NamedEventSetNodes,
path: str,
) -> None
Saves the graph between the inputs
and outputs
EventSetNodes
to a file.
Usage example
>>> evset = tp.event_set(
... timestamps=[1, 2, 3],
... features={"input_feature": [0, 42, 10]}
... )
>>> # Create a graph
>>> a = evset.node()
>>> b = a.moving_sum(2)
>>> b = b.rename("result_feature")
>>> # Check evaluation
>>> b.run({a: evset})
indexes: []
features: [('result_feature', int64)]
events:
(3 events):
timestamps: [1. 2. 3.]
'result_feature': [ 0 42 52]
...
>>> # Save the graph
>>> file_path = tmp_dir / "graph.tem"
>>> tp.save_graph(
... inputs={"input_node": a},
... outputs={"output_node": b},
... path=file_path,
... )
>>> # Load the graph
>>> inputs, outputs = tp.load_graph(path=file_path)
>>> # Evaluate reloaded graph
>>> a_reloaded = inputs["input_node"]
>>> b_reloaded = outputs["output_node"]
>>> b_reloaded.run({a_reloaded: evset})
indexes: []
features: [('result_feature', int64)]
events:
(3 events):
timestamps: [1. 2. 3.]
'result_feature': [ 0 42 52]
...
Parameters:
Name | Type | Description | Default |
---|---|---|---|
inputs |
Optional[NamedEventSetNodes]
|
Input EventSetNodes. If None, the inputs is inferred. In this case, input EventSetNodes have to be named. |
required |
outputs |
NamedEventSetNodes
|
Output EventSetNodes. |
required |
path |
str
|
File path to save to. |
required |