Fix the multi-output, dict-input, parameter counting and calculation overflow problem.

[cainmagi]

Update report

1. Fix the bug of parameter number calculation when there are more than one output variables, including both sequence case and dict case (mentioned in Cannot get the summary #162).
2. Make multiple output variables split into multiple lines.
3. Remove the last line break of summary_string().
4. Enable argument device to accept both str and torch.device.
5. Fix a bug when the model requires batch_size to be a specific number.
6. Fix a bug caused by multiple input cases when dtypes=None.
7. Add text auto wrap when the layer name is too long.
8. Support counting all parameters instead of weight and bias (a different solution of Fix parameter count #142, the package does not count "torch.nn.parameter" #148).
9. Drop the np.sum/prod to fix the overflow problem during calculating the total size (mentioned in RuntimeWarning: overflow encountered in long_scalars #158).
10. Fix the bug caused by layers with dict input values (mentioned in Cannot get the summary #162).
11. Add docstring.

Example for verifying this update

The following code is not compatible with the base repository:

import torch
import torch.nn as nn
from torchsummary import summary

class VeryLongNameSimpleMultiConv(nn.Module):
    def __init__(self):
        super(VeryLongNameSimpleMultiConv, self).__init__()
        self.features_1 = nn.Sequential(
            nn.Conv2d(1, 1, kernel_size=3, stride=1, padding=1),
            nn.ReLU(),
        )
        self.features_2 = nn.Sequential(
            nn.Conv2d(1, 2, kernel_size=3, stride=1, padding=1),
            nn.ReLU(),
        )

    def forward(self, x):
        x1 = self.features_1(x)
        x2 = self.features_2(x)
        return x1, x2
    
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = VeryLongNameSimpleMultiConv().to(device)

summary(model, (1, 16, 16))

Now the output is:

----------------------------------------------------------------
        Layer (type)               Output Shape         Param #
================================================================
            Conv2d-1            [-1, 1, 16, 16]              10
              ReLU-2            [-1, 1, 16, 16]               0
            Conv2d-3            [-1, 2, 16, 16]              20
              ReLU-4            [-1, 2, 16, 16]               0
VeryLong...ltiConv-5            [-1, 1, 16, 16]               0
                                [-1, 2, 16, 16]
================================================================
Total params: 30
Trainable params: 30
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.00
Forward/backward pass size (MB): 0.02
Params size (MB): 0.00
Estimated Total Size (MB): 0.02
----------------------------------------------------------------