hybrid_forward 参数问题

最新推荐文章于 2022-09-07 20:12:55 发布

原创

最新推荐文章于 2022-09-07 20:12:55 发布 · 1.5k 阅读

3 ·

CC 4.0 BY-SA版权

下面代码是全连接层的实现。可以看到 hybrid_forward 的输入参数为：
(self, F, x, weight, bias=None)。F是NDArray或者Symbol.
X是输入的值。我们返现 weight, bias=None 并没有传入，而是通过
在__init__里用self.weight = self.params.get(…)以后，这个weight就会自动被传到hybrid_forward里。

    class Dense(HybridBlock):
    def __init__(self, units, activation=None, use_bias=True, flatten=True,
                 weight_initializer=None, bias_initializer='zeros',
                 in_units=0, **kwargs):
        super(Dense, self).__init__(**kwargs)
        self._flatten = flatten
        with self.name_scope():
            self._units = units
            self._in_units = in_units
            self.weight = self.params.get('weight', shape=(units, in_units),
                                          init=weight_initializer,
                                          allow_deferred_init=True)
            if use_bias:
                self.bias = self.params.get('bias', shape=(units,),
                                            init=bias_init