model.PieceBranch Class Reference

Inheritance diagram for model.PieceBranch:

Collaboration diagram for model.PieceBranch:

Public Member Functions
	__init__ (self, in_channels, mid_channels=16)
	forward (self, x)

Public Attributes
	conv0 = nn.Conv2d(in_channels, mid_channels, kernel_size=3, padding=1)
	conv1
	conv2 = nn.Conv2d(mid_channels, mid_channels, kernel_size=1)
	act = nn.ReLU(inplace=True)

Detailed Description

The deep, cacheable spatial logic for a single piece type.
Because these are computed BEFORE the mixer, they only cost CPU cycles
when the specific piece type moves.

Constructor & Destructor Documentation

__init__()

model.PieceBranch.__init__	(		self,
			in_channels,
			mid_channels = 16 )

Here is the call graph for this function:

Here is the caller graph for this function:

Member Function Documentation

forward()

model.PieceBranch.forward	(		self,
			x )

Member Data Documentation

act

model.PieceBranch.act = nn.ReLU(inplace=True)

conv0

model.PieceBranch.conv0 = nn.Conv2d(in_channels, mid_channels, kernel_size=3, padding=1)

conv1

model.PieceBranch.conv1

Initial value:

=  nn.Conv2d(
            mid_channels,
            mid_channels,
            kernel_size=3,
            padding=1,
            groups=mid_channels,
        )

conv2

model.PieceBranch.conv2 = nn.Conv2d(mid_channels, mid_channels, kernel_size=1)

---

The documentation for this class was generated from the following file:

• training/model.py