Fix the bug of parameter number calculation when there are more than one output variables, including both sequence case and dict case (mentioned in #162).
Make multiple output variables split into multiple lines.
Remove the last line break of summary_string().
Enable argument device to accept both str and torch.device.
Fix a bug when the model requires batch_size to be a specific number.
Fix a bug caused by multiple input cases when dtypes=None.
Add text auto wrap when the layer name is too long.
Support counting all parameters instead of weight and bias (a different solution of #142, #148).
Drop the np.sum/prod to fix the overflow problem during calculating the total size (mentioned in #158).
Fix the bug caused by layers with dict input values (mentioned in #162).
Add docstring.
Example for verifying this update
The following code is not compatible with the base repository:
Update report
summary_string()
.device
to accept both str andtorch.device
.batch_size
to be a specific number.dtypes=None
.weight
andbias
(a different solution of #142, #148).np.sum/prod
to fix the overflow problem during calculating the total size (mentioned in #158).Example for verifying this update
The following code is not compatible with the base repository:
Now the output is: