pytorchvideo_trainer/pytorchvideo_trainer/module/byol.py [35:75]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        backbone_mmt: nn.Module,
        projector: Optional[nn.Module] = None,
        projector_mmt: Optional[nn.Module] = None,
    ) -> None:
        """
        Args:
            backbone (nn.Module): backbone for byol, input shape depends on the forward
                input size. Standard inputs include `B x C`, `B x C x H x W`, and
                `B x C x T x H x W`.
            projector (nn.Module): An mlp with 2 to 3 hidden layers,
                with (synchronized) BatchNorm and ReLU activation.
            backbone_mmt (nn.Module): backbone for byol, input shape depends on the forward
                input size. Standard inputs include `B x C`, `B x C x H x W`, and
                `B x C x T x H x W`.
            projector_mmt (nn.Module): Am mlp with 2 to 3 hidden layers,
                with (synchronized) BatchNorm and ReLU activation.
            predictor (nn.Module): predictor MLP of BYOL of similar structure as the
                projector MLP.
            mmt (float): momentum update ratio for the momentum backbone.
        """
        super().__init__()

        self.mmt: float = mmt
        if projector is not None:
            backbone = nn.Sequential(
                backbone,
                projector,
            )
        init_net_weights(backbone)
        self.backbone = backbone

        if projector_mmt is not None:
            backbone_mmt = nn.Sequential(
                backbone_mmt,
                projector_mmt,
            )
        init_net_weights(backbone_mmt)
        self.backbone_mmt = backbone_mmt

        for p in self.backbone_mmt.parameters():
            p.requires_grad = False
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



pytorchvideo_trainer/pytorchvideo_trainer/module/moco_v2.py [85:124]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        backbone_mmt: nn.Module,
        projector: Optional[nn.Module] = None,
        projector_mmt: Optional[nn.Module] = None,
    ) -> None:
        """
        Args:
            backbone (nn.Module): backbone for byol, input shape depends on the forward
                input size. Standard inputs include `B x C`, `B x C x H x W`, and
                `B x C x T x H x W`.
            projector (nn.Module): An mlp with 2 to 3 hidden layers,
                with (synchronized) BatchNorm and ReLU activation.
            backbone_mmt (nn.Module): backbone for byol, input shape depends on the forward
                input size. Standard inputs include `B x C`, `B x C x H x W`, and
                `B x C x T x H x W`.
            projector_mmt (nn.Module): Am mlp with 2 to 3 hidden layers,
                with (synchronized) BatchNorm and ReLU activation.
            mmt (float): momentum update ratio for the momentum backbone.
        """
        super().__init__()

        self.mmt: float = mmt

        if projector is not None:
            backbone = nn.Sequential(
                backbone,
                projector,
            )
        init_net_weights(backbone)
        self.backbone = backbone

        if projector_mmt is not None:
            backbone_mmt = nn.Sequential(
                backbone_mmt,
                projector_mmt,
            )
        init_net_weights(backbone_mmt)
        self.backbone_mmt = backbone_mmt

        for p in self.backbone_mmt.parameters():
            p.requires_grad = False
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -