| import torch | |
| import torch.nn as nn | |
| class IdentityProjector(nn.Module): | |
| def __init__( | |
| self, | |
| input_dim: int, | |
| output_dim: int, | |
| ): | |
| super().__init__() | |
| self.placeholder_param = nn.Parameter(torch.zeros(1)) | |
| def forward(self, *args): | |
| return args[0] | |
| def configure_optimizers(self, weight_decay, lr, betas): | |
| return torch.optim.AdamW( | |
| self.parameters(), lr=lr, betas=betas, weight_decay=weight_decay | |
| ) | |