feed

Product	Supported
Atlas A3 training product/Atlas A3 inference product	√
Atlas A2 training product/Atlas A2 inference product	√
Atlas 200I/500 A2 inference product	x
Atlas inference product	x
Atlas training product	x

Inputs data to a graph. A serializable input is supported.

feed(feed_dict, timeout=-1, partial_inputs = False) -> int

Parameter	Data Type	Description
feed_dict	Dict[FlowData, Any]	The key is a FlowData node, and the value can be any input that can be serialized. To connect to the NPU model, the input must be of the dataflow.Tensor type.
timeout	int	Timeout period for inputting data, in ms. The value range is [0, 2147483647), and the value -1 indicates that no timeout will occur.
partial_inputs	bool	Whether feed_dict supports containing only partial inputs of the model whenever the feed API is called. The value can be: True: The feed_dict parameter can contain only partial inputs of the model. False: The feed_dict parameter must contain all inputs of the model. The default value is False.

Parameter

Data Type

Description

feed_dict

Dict[FlowData, Any]

The key is a FlowData node, and the value can be any input that can be serialized.

To connect to the NPU model, the input must be of the dataflow.Tensor type.

timeout

int

Timeout period for inputting data, in ms. The value range is [0, 2147483647), and the value -1 indicates that no timeout will occur.

partial_inputs

bool

Whether feed_dict supports containing only partial inputs of the model whenever the feed API is called. The value can be:

True: The feed_dict parameter can contain only partial inputs of the model.
False: The feed_dict parameter must contain all inputs of the model.

The default value is False.

0 is returned in normal scenarios.

A specific error code is returned and error logs are printed in abnormal scenarios.

import dataflow as df
graph = df.FlowGraph(...)
graph.feed(...)

If an input of the non-dataflow.Tensor type is used, decorators @pyflow and @method must be used together.

Parent topic: dataflow.FlowGraph