unet.yaml

unet_additional_kwargs:
  use_inflated_groupnorm: true
  unet_use_cross_frame_attention: false
  unet_use_temporal_attention: false
  use_motion_module: true
  use_audio_module: true
  motion_module_resolutions:
    - 1
    - 2
    - 4
    - 8
  motion_module_mid_block: true
  motion_module_decoder_only: false
  motion_module_type: Vanilla
  motion_module_kwargs:
    num_attention_heads: 8
    num_transformer_block: 1
    attention_block_types:
      - Temporal_Self
      - Temporal_Self
    temporal_position_encoding: true
    temporal_position_encoding_max_len: 32
    temporal_attention_dim_div: 1
  audio_attention_dim: 768
  stack_enable_blocks_name:
    - "up"
    - "down"
    - "mid"
  stack_enable_blocks_depth: [0,1,2,3]

enable_zero_snr: true

noise_scheduler_kwargs:
  beta_start: 0.00085
  beta_end: 0.012
  beta_schedule: "linear"
  clip_sample: false
  steps_offset: 1
  ### Zero-SNR params
  prediction_type: "v_prediction"
  rescale_betas_zero_snr: True
  timestep_spacing: "trailing"

sampler: DDIM