ConvNet¶
- class torchrl.modules.ConvNet(in_features: int | None = None, depth: int | None = None, num_cells: Sequence[int] | int = None, kernel_sizes: Union[Sequence[int], int] = 3, strides: Sequence[int] | int = 1, paddings: Sequence[int] | int = 0, activation_class: Type[nn.Module] | Callable = <class 'torch.nn.modules.activation.ELU'>, activation_kwargs: dict | List[dict] | None = None, norm_class: Type[nn.Module] | Callable | None = None, norm_kwargs: dict | List[dict] | None = None, bias_last_layer: bool = True, aggregator_class: Type[nn.Module] | Callable | None = <class 'torchrl.modules.models.utils.SquashDims'>, aggregator_kwargs: dict | None = None, squeeze_output: bool = False, device: DEVICE_TYPING | None = None)[source]¶
一个卷积神经网络。
- 参数:
in_features (int, optional) – 输入特征数。如果为
None
,则使用LazyConv2d
模块作为第一层。;depth (int, optional) – 网络的深度。深度为 1 将生成具有所需输入大小的单个线性层网络,其输出大小等于
num_cells
参数的最后一个元素。如果没有指定深度,则深度信息应包含在num_cells
参数中(见下文)。如果num_cells
是一个可迭代对象并且指定了depth
,则两者应该匹配:len(num_cells)
必须等于depth
。num_cells (int 或 Sequence of int, optional) – 输入和输出之间每一层中的单元格数。如果提供整数,则每一层将具有相同的单元格数。如果提供一个可迭代对象,则线性层
out_features
将与num_cells
的内容匹配。默认值为[32, 32, 32]
。kernel_sizes (int, sequence of int, optional) – 卷积网络的内核大小。如果为可迭代对象,则长度必须与
num_cells
或深度参数定义的深度匹配。默认值为3
。strides (int 或 sequence of int, optional) – 卷积网络的步长。如果为可迭代对象,则长度必须与
num_cells
或深度参数定义的深度匹配。默认值为1
。activation_class (Type[nn.Module] or callable, optional) – 要使用的激活类或构造函数。默认值为
Tanh
.activation_kwargs (dict 或 list of dicts, optional) – 要与激活类一起使用的关键字参数。也可以传递长度为
depth
的关键字参数列表,每层一个元素。norm_class (Type 或 callable, optional) – 归一化类或构造函数(如果有)。
norm_kwargs (dict 或 list of dicts, optional) – 要与归一化层一起使用的关键字参数。也可以传递长度为
depth
的关键字参数列表,每层一个元素。bias_last_layer (bool) – 如果为
True
,则最后一个线性层将具有偏差参数。默认值为True
。aggregator_class (Type[nn.Module] or callable) – 要在链末尾使用的聚合器类或构造函数。默认值为
torchrl.modules.utils.models.SquashDims
;aggregator_kwargs (dict, optional) –
aggregator_class
的关键字参数。squeeze_output (bool) – 输出是否应该压缩其单例维度。默认值为
False
。device (torch.device, optional) – 用于创建模块的设备。
示例
>>> # All of the following examples provide valid, working MLPs >>> cnet = ConvNet(in_features=3, depth=1, num_cells=[32,]) # MLP consisting of a single 3 x 6 linear layer >>> print(cnet) ConvNet( (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1)) (1): ELU(alpha=1.0) (2): SquashDims() ) >>> cnet = ConvNet(in_features=3, depth=4, num_cells=32) >>> print(cnet) ConvNet( (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1)) (1): ELU(alpha=1.0) (2): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1)) (3): ELU(alpha=1.0) (4): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1)) (5): ELU(alpha=1.0) (6): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1)) (7): ELU(alpha=1.0) (8): SquashDims() ) >>> cnet = ConvNet(in_features=3, num_cells=[32, 33, 34, 35]) # defines the depth by the num_cells arg >>> print(cnet) ConvNet( (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1)) (1): ELU(alpha=1.0) (2): Conv2d(32, 33, kernel_size=(3, 3), stride=(1, 1)) (3): ELU(alpha=1.0) (4): Conv2d(33, 34, kernel_size=(3, 3), stride=(1, 1)) (5): ELU(alpha=1.0) (6): Conv2d(34, 35, kernel_size=(3, 3), stride=(1, 1)) (7): ELU(alpha=1.0) (8): SquashDims() ) >>> cnet = ConvNet(in_features=3, num_cells=[32, 33, 34, 35], kernel_sizes=[3, 4, 5, (2, 3)]) # defines kernels, possibly rectangular >>> print(cnet) ConvNet( (0): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1)) (1): ELU(alpha=1.0) (2): Conv2d(32, 33, kernel_size=(4, 4), stride=(1, 1)) (3): ELU(alpha=1.0) (4): Conv2d(33, 34, kernel_size=(5, 5), stride=(1, 1)) (5): ELU(alpha=1.0) (6): Conv2d(34, 35, kernel_size=(2, 3), stride=(1, 1)) (7): ELU(alpha=1.0) (8): SquashDims() )