Method register_buffer

python/paddle/nn/layer/layers.py:1609–1676

Registers a tensor as buffer into the layer. `buffer` is a non-trainable tensor and will not be updated by optimizer, but is necessary for evaluation and inference. For example, the mean and variance in BatchNorm layers. The registered buffer is persistable by defau

(
        self, name: str, tensor: Tensor, persistable: bool = True
    )

    @param_one_alias(["persistable", "persistent"])
    def register_buffer(
        self, name: str, tensor: Tensor, persistable: bool = True
    ) -> None:
        """
        Registers a tensor as buffer into the layer.

        `buffer` is a non-trainable tensor and will not be updated by optimizer,
        but is necessary for evaluation and inference. For example, the mean and variance in BatchNorm layers.
        The registered buffer is persistable by default, and will be saved into
        `state_dict` alongside parameters. If set persistable=False, it registers
        a non-persistable buffer, so that it will not be a part of `state_dict` .

        Buffers can be accessed as attributes using given names.

        Parameters:
            name (string): name of the buffer. The buffer can be accessed
                from this layer using the given name
            tensor (Tensor): the tensor to be registered as buffer.
            persistable (bool): whether the buffer is part of this layer's
                state_dict.

        Returns:
            None

        Examples:
            .. code-block:: pycon

                >>> import numpy as np
                >>> import paddle

                >>> linear = paddle.nn.Linear(10, 3)
                >>> value = np.array([0]).astype("float32")
                >>> buffer = paddle.to_tensor(value)
                >>> linear.register_buffer("buf_name", buffer, persistable=True)

                >>> # get the buffer by attribute.
                >>> print(linear.buf_name)
                Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True,
                [0.])

        """

        if '_buffers' not in self.__dict__:
            raise ValueError("super().__init__() should be called first")
        elif not isinstance(name, str):
            raise TypeError(
                f"The name of buffer should be a string, but received {type(name).__name__}."
            )
        elif '.' in name:
            raise KeyError(
                "The name of buffer can not contain `.`, "
                "because when you access the newly added buffer in the "
                "form of `self.**.**`, it will cause AttributeError."
            )
        elif name == '':
            raise KeyError("The name of buffer can not be empty.")
        elif hasattr(self, name) and name not in self._buffers:
            raise KeyError(f"attribute '{name}' already exists.")