def symbolic(g, input, pad, value=0):
paddings = prepare_onnx_paddings(len(input.type().sizes()), pad)
return g.op("Pad",
input,
pads_i=paddings,
mode_s="constant",
value_f=value)
def symbolic(g, input, pad, value=0):
paddings = prepare_onnx_paddings(len(input.type().sizes()), pad)
return g.appendNode(
g.create("Pad",
[input]).is_("paddings",
paddings).s_("mode",
"constant").f_("value", value))
def replication_pad(g, input, padding):
from torch.autograd._functions.utils import prepare_onnx_paddings
mode = "edge"
paddings = prepare_onnx_paddings(len(input.type().sizes()), padding)
return g.op("Pad", input, pads_i=paddings, mode_s=mode)
def test_prepare_onnx_paddings(self):
sizes = [2, 3, 4]
pad = [1, 2, 3, 4]
paddings = prepare_onnx_paddings(len(sizes), pad)
self.assertEqual(paddings, [0, 3, 1, 0, 4, 2])
def constant_pad_nd(g, input, padding, value):
from torch.autograd._functions.utils import prepare_onnx_paddings
mode = "constant"
paddings = prepare_onnx_paddings(len(input.type().sizes()), padding)
return g.op("Pad", input, pads_i=paddings, mode_s=mode, value_f=value)
def test_prepare_onnx_paddings(self):
sizes = [2, 3, 4]
pad = [1, 2, 3, 4]
paddings = prepare_onnx_paddings(len(sizes), pad)
self.assertEqual(paddings, [0, 3, 1, 0, 4, 2])
def replication_pad(g, input, padding):
from torch.autograd._functions.utils import prepare_onnx_paddings
mode = "edge"
paddings = prepare_onnx_paddings(len(input.type().sizes()), padding)
return g.op("Pad", input, pads_i=paddings, mode_s=mode)
def replicationpad_symbolic(g, input, *params):
mode = "edge"
paddings = prepare_onnx_paddings(len(input.type().sizes()), params)
return g.op("Pad", input, paddings_i=paddings, mode_s=mode)
def replicationpad_symbolic(g, input, *params):
mode = "edge"
paddings = prepare_onnx_paddings(len(input.type().sizes()), params)
return g.op("Pad", input, pads_i=paddings, mode_s=mode)
def symbolic(g, input: Variable, padding: Union[int, Tuple[int]]):
paddings = prepare_onnx_paddings(len(input.type().sizes()), pad)
return g.op("Pad", input, pads_i=paddings, mode_s="reflect")
def symbolic(g, input, pad, value=0):
paddings = prepare_onnx_paddings(len(input.type().sizes()), pad)
return g.op("Pad", input, paddings_i=paddings, mode_s="constant", value_f=value)
def reflectionpad_symbolic(g, input, *params):
mode = "reflect"
paddings = prepare_onnx_paddings(input, params)
return g.op("Pad", input, paddings_i=paddings, mode_s=mode)
