pretrained_model_path: "../models/Model/stable-diffusion-v1-5"

motion_module_path: '../models/live2diff.ckpt'
depth_model_path: '../models/dpt_hybrid_384.pt'

unet_additional_kwargs:
  cond_mapping: true
  use_inflated_groupnorm:          true
  use_motion_module              : true
  motion_module_resolutions      : [ 1,2,4,8 ]
  unet_use_cross_frame_attention : false
  unet_use_temporal_attention    : false

  motion_module_type: Streaming
  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 : 24
    temporal_attention_dim_div         : 1
    zero_initialize                    : true

    attention_class_name               : 'stream'

    attention_kwargs:
      window_size: 16
      sink_size: 8

noise_scheduler_kwargs:
  num_train_timesteps: 1000
  beta_start:    0.00085
  beta_end:      0.012
  beta_schedule: "linear"
  steps_offset:  1
  clip_sample:   False

third_party_dict:
  dreambooth: "../models/Model/aziibpixelmix_v10.safetensors"
  lora_list:
    - lora: '../models/LoRA/kFeltedReV.safetensors'
      lora_alpha: 1
  clip_skip: 2

num_inference_steps: 50
t_index_list: [35, 40, 45]

prompt: "masterpiece, best quality, 1gril"