pnnx convert nn.Upsample with recompute_scale_factor=True

[nihui]

No description provided.

[nihui]

import torch
import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self):
        super(Model, self).__init__()

        self.up_0 = nn.Upsample(scale_factor=(1,2.44), mode='bilinear', align_corners=False, recompute_scale_factor=False)
        self.up_1 = nn.Upsample(scale_factor=(1,2.44), mode='bilinear', align_corners=False, recompute_scale_factor=True)
        self.up_2 = nn.Upsample(scale_factor=(1,2.44), mode='bilinear', align_corners=True, recompute_scale_factor=False)
        self.up_3 = nn.Upsample(scale_factor=(1,2.44), mode='bilinear', align_corners=True, recompute_scale_factor=True)

    def forward(self, x):
        out0 = self.up_0(x)
        out1 = self.up_1(x)
        out2 = self.up_2(x)
        out3 = self.up_3(x)
        return out0, out1, out2, out3

def test():
    net = Model()
    net.eval()

    torch.manual_seed(0)
    x = torch.rand(1, 1, 1, 2)

    out0, out1, out2, out3 = net(x)

    print(out0)
    print(out1)
    print(out2)
    print(out3)

if __name__ == "__main__":
    if test():
        exit(0)
    else:
        exit(1)

# [nihui@nihuini-LC2 test]$ python test.py
# tensor([[[[0.4963, 0.5275, 0.6389, 0.7504]]]])
# tensor([[[[0.4963, 0.5642, 0.7002, 0.7682]]]])
# tensor([[[[0.4963, 0.5869, 0.6776, 0.7682]]]])
# tensor([[[[0.4963, 0.5869, 0.6776, 0.7682]]]])