Method __init__

sat/vae_modules/autoencoder.py:434–458

(self, embed_dim: int, **kwargs)

class AutoencodingEngineLegacy(AutoencodingEngine):
    def __init__(self, embed_dim: int, **kwargs):
        self.max_batch_size = kwargs.pop("max_batch_size", None)
        ddconfig = kwargs.pop("ddconfig")
        ckpt_path = kwargs.pop("ckpt_path", None)
        ckpt_engine = kwargs.pop("ckpt_engine", None)
        super().__init__(
            encoder_config={
                "target": "sgm.modules.diffusionmodules.model.Encoder",
                "params": ddconfig,
            },
            decoder_config={
                "target": "sgm.modules.diffusionmodules.model.Decoder",
                "params": ddconfig,
            },
            **kwargs,
        )
        self.quant_conv = torch.nn.Conv2d(
            (1 + ddconfig["double_z"]) * ddconfig["z_channels"],
            (1 + ddconfig["double_z"]) * embed_dim,
            1,
        )
        self.post_quant_conv = torch.nn.Conv2d(embed_dim, ddconfig["z_channels"], 1)
        self.embed_dim = embed_dim

        self.apply_ckpt(default(ckpt_path, ckpt_engine))

    def get_autoencoder_params(self) -> list:
        params = super().get_autoencoder_params()