ResNet basic 3D stem module. It performs spatiotemporal Convolution, BN, and activation following by a spatiotemporal pooling. Conv3d ↓ Normalization ↓ Activation ↓ Pool3d :param in_channels: Input channel size of the convolution. :type in_channels: int :param out_channels: Output channel size of the convolution. :type out_channels: int :param conv_kernel_size: Convolutional kernel size(s). :type conv_kernel_size: Tuple :param conv_stride: Convolutional stride size(s). :type conv_stride: Tuple :param conv_padding: Convolutional padding size(s). :type conv_padding: Tuple :param conv_bias: Convolutional bias. If true, adds a learnable bias to the output. :type conv_bias: bool :param conv: Callable used to build the convolution layer. :type conv: callable :param pool: A callable that constructs pooling layer, options include: nn.AvgPool3d, nn.MaxPool3d, and None (not performing pooling). :type pool: Callable :param pool_kernel_size: Pooling kernel size(s). :type pool_kernel_size: Tuple :param pool_stride: Pooling stride size(s). :type pool_stride: Tuple :param pool_padding: Pooling padding size(s). :type pool_padding: Tuple :param norm: A callable that constructs normalization layer, options include nn.BatchNorm3d, None (not performing normalization). :type norm: Callable :param norm_eps: Normalization epsilon. :type norm_eps: float :param norm_momentum: Normalization momentum. :type norm_momentum: float :param activation: A callable that constructs activation layer, options include: nn.ReLU, nn.Softmax, nn.Sigmoid, and None (not performing activation). :type activation: Callable.