config/easyanimate_video_v5.1_magvit_qwen.yaml

transformer_additional_kwargs:
  transformer_type:                           "EasyAnimateTransformer3DModel"
  after_norm:                                 false
  time_position_encoding_type:                "3d_rope"
  resize_inpaint_mask_directly:               true
  enable_text_attention_mask:                 true
  enable_clip_in_inpaint:                     false
  add_ref_latent_in_control_model:            true

vae_kwargs:
  vae_type: "AutoencoderKLMagvit"
  mini_batch_encoder: 4
  mini_batch_decoder: 1
  slice_mag_vae: false
  slice_compression_vae: false
  cache_compression_vae: false
  cache_mag_vae: true

text_encoder_kwargs:
  enable_multi_text_encoder: false
  replace_t5_to_llm: true