configs/echo_infinity-long.yaml

generator_ckpt: checkpoints/echo_infinity.pt
generator_fsdp_wrap_strategy: size
real_score_fsdp_wrap_strategy: size
fake_score_fsdp_wrap_strategy: size
real_name: Wan2.1-T2V-14B
fake_name: Wan2.1-T2V-1.3B
text_encoder_fsdp_wrap_strategy: size
denoising_step_list:
- 1000
- 750
- 500
- 250
warp_denoising_step: true
ts_schedule: false
num_train_timestep: 1000
timestep_shift: 5.0
guidance_scale: 3.0
denoising_loss_type: flow
mixed_precision: true
seed: 0
wandb_key: your_wandb_key
wandb_entity: your_wandb_entity
wandb_project: Echo-Infinity
sharding_strategy: hybrid_full
lr: 1.0e-05
lr_critic: 2.0e-06
beta1: 0.0
beta2: 0.999
beta1_critic: 0.0
beta2_critic: 0.999
data_path: prompts/vidprom_filtered_extended.txt
switch_prompt_path: prompts/vidprom_filtered_extended_switch.txt
batch_size: 1
gradient_accumulation_steps: 2
ema_weight: 0.99
ema_start_step: 200
total_batch_size: 64
log_iters: 200
max_checkpoints: 50
max_iters: 6000

negative_prompt: '色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走'
dfake_gen_update_ratio: 5
image_or_video_shape:
- 1
- 240
- 16
- 60
- 104
distribution_loss: dmd_switch
global_sink: true
trainer: score_distillation
gradient_checkpointing: true
num_frame_per_block: 3
load_raw_video: false
model_kwargs:
  timestep_shift: 5.0
  local_attn_size: 12
  sink_size: 3
  relative_rope: true
  relative_rope_pmax: 21

vis_interval: 500
vis_ema: false
vis_video_lengths:
- 240
val_batch_size: 1
val_data_path: prompts/vidprom_filtered_extended.txt
val_switch_prompt_path: prompts/vidprom_filtered_extended_switch.txt

slice_last_frames: 21

switch_mode: random_choice
switch_choices:
- 21
- 39
- 57
- 75
- 93
- 111
- 129
- 147
- 165
- 183
- 201

streaming_training: true
streaming_chunk_size: 21
streaming_max_length: 240

streaming_min_new_frame: 18
train_first_chunk: true

last_step_only: false

adapter:
  type: "lora"
  rank: 256
  alpha: 256
  dropout: 0.0
  dtype: "bfloat16"
  apply_to_critic: true
  verbose: true

memory_kwargs:
  enabled: true
  Q_frames: 3
  tokens_per_frame: 1560
  n_encoder_layers: 2
  hidden_dim: 1536
  num_heads: 12
  head_dim: 128
  initializer_range: 0.014
  rope: true
  qk_norm: true
  use_evicted_kv: false
  use_batch_update: false
  use_sink_anchor: false
  use_vib: false
  bptt_clips: 1
  gate_init_bias: 2.0
  encoder_lr_multiplier: 5.0
  normalize_memory_k: false
  memory_recache: true