tracker_project_name: memo
output_dir: outputs
resume_from_checkpoint: null
model_name_or_path: memoavatar/memo
vae: stabilityai/sd-vae-ft-mse
gradient_checkpointing: true
gradient_accumulation_steps: 1
train_batch_size: 2
max_train_steps: 3500
num_train_epochs: -1
enable_xformers_memory_efficient_attention: true
checkpoints_total_limit: 3
robust_training: true
learning_rate: 1e-5
max_grad_norm: 1.0
scale_lr: false
lr_scheduler: constant
lr_warmup_steps: 0
seed: 42
mixed_precision: bf16
use_8bit_adam: false
allow_tf32: true
use_ema: false
adam_beta1: 0.9
adam_beta2: 0.999
adam_weight_decay: 0.01
adam_epsilon: 1e-08
dataloader_num_workers: 16
prefetch_factor: 4
checkpointing_steps: 5000
data:
  width: 512
  height: 512
  num_past_frames: 16
  dynamic_past_frames: false
  n_sample_frames: 16
  audio_margin: 2
  metadata_paths:
    - assets/embedding/metadata.jsonl
weighting_scheme: logit_normal
logit_mean: 0.0
logit_std: 1.0 
mode_scale: 1.29
noise_scheduler_kwargs:
  num_train_timesteps: 1000
train_reference_net: true
train_diffusion_net: true
train_image_proj: true
train_audio_proj: true
trainable_modules:  
  - to_q
  - to_k
  - to_v
uncond_img_ratio: 0.05
uncond_audio_ratio: 0.05
start_ratio: 0.05