An input tensor of shape (channels height width) passes through an Inception module with these four branches:

• 4 filters of size ;
• 2 filters of size , followed by 4 filters of size ;
• 1 filter of size , followed by 2 filters of size ;
• max-pooling, followed by 2 filters of size .

All operations use stride and padding that preserves the spatial dimensions. After the forward pass of the module, what is the output shape?