MCPcopy Create free account
hub / github.com/YesianRohn/TextSSR / ResidualTemporalBlock1D

Class ResidualTemporalBlock1D

diffusers/src/diffusers/models/resnet.py:424–468  ·  view source on GitHub ↗

Residual 1D block with temporal convolutions. Parameters: inp_channels (`int`): Number of input channels. out_channels (`int`): Number of output channels. embed_dim (`int`): Embedding dimension. kernel_size (`int` or `tuple`): Size of the convolving kernel.

Source from the content-addressed store, hash-verified

422
423# unet_rl.py
424class ResidualTemporalBlock1D(nn.Module):
425 """
426 Residual 1D block with temporal convolutions.
427
428 Parameters:
429 inp_channels (`int`): Number of input channels.
430 out_channels (`int`): Number of output channels.
431 embed_dim (`int`): Embedding dimension.
432 kernel_size (`int` or `tuple`): Size of the convolving kernel.
433 activation (`str`, defaults `mish`): It is possible to choose the right activation function.
434 """
435
436 def __init__(
437 self,
438 inp_channels: int,
439 out_channels: int,
440 embed_dim: int,
441 kernel_size: Union[int, Tuple[int, int]] = 5,
442 activation: str = "mish",
443 ):
444 super().__init__()
445 self.conv_in = Conv1dBlock(inp_channels, out_channels, kernel_size)
446 self.conv_out = Conv1dBlock(out_channels, out_channels, kernel_size)
447
448 self.time_emb_act = get_activation(activation)
449 self.time_emb = nn.Linear(embed_dim, out_channels)
450
451 self.residual_conv = (
452 nn.Conv1d(inp_channels, out_channels, 1) if inp_channels != out_channels else nn.Identity()
453 )
454
455 def forward(self, inputs: torch.Tensor, t: torch.Tensor) -> torch.Tensor:
456 """
457 Args:
458 inputs : [ batch_size x inp_channels x horizon ]
459 t : [ batch_size x embed_dim ]
460
461 returns:
462 out : [ batch_size x out_channels x horizon ]
463 """
464 t = self.time_emb_act(t)
465 t = self.time_emb(t)
466 out = self.conv_in(inputs) + rearrange_dims(t)
467 out = self.conv_out(out)
468 return out + self.residual_conv(inputs)
469
470
471class TemporalConvLayer(nn.Module):

Callers 4

__init__Method · 0.85
__init__Method · 0.85
__init__Method · 0.85
__init__Method · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected