dmx.compressor.modeling.nn.experimental

Classes

Conv1dScatter(in_channels, out_channels, ...)	This is an alternative version of the DmxModule .nn.Conv1d, without calling torch.nn.functional.conv1d(), but torch.scatter() and torch.matmul() instead.
Conv1dUnfold(in_channels, out_channels, ...)	This is an alternative version of the DmxModule .nn.Conv1d, without calling torch.nn.functional.conv1d(), but torch.nn.functional.unfold() and torch.matmul() instead.
Conv2dGather(in_channels, out_channels, ...)	This is an alternative version of the DmxModule.nn.Conv2d, without calling torch.nn.functional.conv2d(), but torch.gather() and torch.matmul() instead.
Conv2dUnfold(in_channels, out_channels, ...)	This is an alternative version of the DmxModule.nn.Conv2d, without calling torch.nn.functional.conv2d(), but torch.nn.functional.unfold() and torch.matmul() instead.

class dmx.compressor.modeling.nn.experimental.Conv1dScatter(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', **kwargs)

Bases: Conv1d

This is an alternative version of the DmxModule .nn.Conv1d, without calling torch.nn.functional.conv1d(), but torch.scatter() and torch.matmul() instead.

to_compiler_graph() → Graph

Returns a compiler friendly graph

training: bool

class dmx.compressor.modeling.nn.experimental.Conv1dUnfold(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', **kwargs)

Bases: Conv1d

This is an alternative version of the DmxModule .nn.Conv1d, without calling torch.nn.functional.conv1d(), but torch.nn.functional.unfold() and torch.matmul() instead.

to_compiler_graph() → Graph

Returns a compiler friendly graph

training: bool

class dmx.compressor.modeling.nn.experimental.Conv2dGather(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', **kwargs)

Bases: Conv2d

This is an alternative version of the DmxModule.nn.Conv2d, without calling torch.nn.functional.conv2d(), but torch.gather() and torch.matmul() instead.

to_compiler_graph() → Graph

Returns a compiler friendly graph

training: bool

class dmx.compressor.modeling.nn.experimental.Conv2dUnfold(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', **kwargs)

Bases: Conv2d

This is an alternative version of the DmxModule.nn.Conv2d, without calling torch.nn.functional.conv2d(), but torch.nn.functional.unfold() and torch.matmul() instead.

to_compiler_graph() → Graph

Returns a compiler friendly graph

training: bool