com.microsoft - Pad

Pad - 1

Version

• name: Pad (GitHub)

• domain: com.microsoft

• since_version: 1

• function:

• support_level: SupportType.COMMON

• shape inference: True

This version of the operator has been available since version 1 of domain com.microsoft.

Summary

Attributes

• mode - STRING : Three modes: constant`(default) - pads with a given constant value, `reflect - pads with the reflection of the vector mirrored on the first and last values of the vector along each axis, edge - pads with the edge values of array

Inputs

Between 2 and 3 inputs.

• data (heterogeneous) - T:

• pads (heterogeneous) - tensor(int64):

• value (optional, heterogeneous) - T:

Outputs

• output (heterogeneous) - T:

Type Constraints

• T in ( tensor(double), tensor(float), tensor(float16) ): Constrain input and output types to float tensors.

Examples

_constant_pad

import numpy as np
import onnx

node = onnx.helper.make_node(
    "Pad", inputs=["x", "pads", "value"], outputs=["y"], mode="constant"
)
x = np.random.randn(1, 3, 4, 5).astype(np.float32)
pads = np.array([0, 0, 1, 3, 0, 0, 2, 4]).astype(
    np.int64
)  # pad order [x1_begin, x2_begin, ..., x1_end, x2_end, ...]
value = np.float32(1.2)
y = pad_impl(x, pads, "constant", 1.2)

expect(node, inputs=[x, pads, value], outputs=[y], name="test_constant_pad")

_reflection_edge_and_wrap_pad

import numpy as np
import onnx

for mode in ["edge", "reflect", "wrap"]:
    node = onnx.helper.make_node(
        "Pad", inputs=["x", "pads"], outputs=["y"], mode=mode
    )
    x = np.random.randn(1, 3, 4, 5).astype(np.int32)
    pads = np.array([0, 0, 1, 1, 0, 0, 1, 1]).astype(
        np.int64
    )  # pad order [x1_begin, x2_begin, ..., x1_end, x2_end, ...]
    y = pad_impl(x, pads, mode)

    expect(node, inputs=[x, pads], outputs=[y], name=f"test_{mode}_pad")

_constant_pad_axes

import numpy as np
import onnx

node = onnx.helper.make_node(
    "Pad", inputs=["x", "pads", "value", "axes"], outputs=["y"], mode="constant"
)
x = np.random.randn(1, 3, 4, 5).astype(np.float32)
pads = np.array([0, 3, 0, 4]).astype(
    np.int64
)  # pad order [x1_begin, x2_begin, ..., x1_end, x2_end, ...]
value = np.float32(1.2)
axes = np.array([1, 3], dtype=np.int64)
y = pad_impl(
    x,
    pads,
    "constant",
    1.2,
    [1, 3],
)

expect(
    node,
    inputs=[x, pads, value, axes],
    outputs=[y],
    name="test_constant_pad_axes",
)