Skip to content

register_function_parameters

Function parameter registration method.

It registers the parameters involved in some data processing functions, e.g., BatchNorm1d.

Parameters:

Name Type Description Default
model

The rpn model.

 required
func_list

The list of data processing functions.

 required

Returns:

Type Description
None

This functon doesn't have any return values.
Source code in tinybig/util/util.py 
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
def register_function_parameters(model, func_list):
    """
    Function parameter registration method.

    It registers the parameters involved in some data processing functions, e.g., BatchNorm1d.

    Parameters
    ----------
    model: Any
        The rpn model.
    func_list: list
        The list of data processing functions.

    Returns
    -------
    None
        This functon doesn't have any return values.
    """
    if not hasattr(func_list, '__iter__'):
        func_list = [func_list]
    for idx, function in enumerate(func_list):
        if hasattr(function, 'weight') and hasattr(function, 'bias'):
            model.register_parameter(f'layer_{idx}_weight', function.weight)
            model.register_parameter(f'layer_{idx}_bias', function.bias)
            if hasattr(function, 'running_mean') and hasattr(function, 'running_var'):
                model.register_buffer(f'layer_{idx}_running_mean', function.running_mean)
                model.register_buffer(f'layer_{idx}_running_var', function.running_var)