params:
    image_size: 64
    model_channels: 704
    num_res_blocks: 3
    channel_mult: "1,2,3,4"
    attention_resolutions: "32,16,8"
    num_heads: 1
    num_head_channels: 64
    num_heads_upsample: -1
    use_scale_shift_norm: true
    dropout: 0.0
    resblock_updown: true
    encoder_dim: 4096
    encoder_channels: 2816
    in_channels: 3
    out_channels: 6
    activation: "gelu"
    att_pool_heads: 64
    disable_self_attentions: false
    precision: "16"