| models: |
| student: |
| name: "Self-Forcing" |
| size: "1.3B" |
| config_path: "./self_forcing/configs/self_forcing_dmd.yaml" |
| checkpoint_path: "/mnt/vepfs/base2/stream/luoyang/checkpoints/self_forcing_dmd.pt" |
| device: "cuda:0" |
| use_ema: true |
|
|
| discriminator: |
| name: "VideoAlign" |
| path: "/mnt/vepfs/base2/stream/luoyang/checkpoints/Qwen2-VL-2B-Instruct" |
| pretrained_checkpoint_dir: "/mnt/vepfs/base2/stream/luoyang/checkpoints/VideoReward" |
| pretrained_checkpoint_step: 11352 |
| warmup_checkpoint: "/mnt/vepfs/base2/stream/luoyang/checkpoints/disc_videoalign_sf_openvid1m_warmup/checkpoint-best" |
| lora_rank: 64 |
| lora_alpha: 128 |
| lora_dropout: 0.05 |
| num_frames: 16 |
| max_pixels: 151200 |
| fps: 1.0 |
|
|
| lora: |
| rank: 256 |
| alpha: 256 |
| dropout: 0.0 |
| target_modules: ['CausalWanAttentionBlock'] |
|
|
| training: |
| mixed_precision: "bf16" |
| gad_nft: |
| epochs: 100 |
| max_steps: 300 |
| warmup_steps: 0 |
| batch_size: 3 |
| micro_batch_size: 1 |
| student_learning_rate: 1.0e-5 |
| freeze_discriminator: true |
| discriminator_learning_rate: 0 |
| gradient_accumulation_steps: 1 |
| inference_batch_size: 4 |
| decode_batch_size: 4 |
| reward_inference_batch_size: 4 |
| discriminator_update_interval: 1 |
| save_steps: 60 |
| logging_steps: 20 |
| stat_clear_interval: 1000 |
| ema_weight: 0.99 |
| ema_start_step: 0 |
| use_ema_for_eval: true |
| clip_grad_norm: 1.0 |
| gc_interval: 200 |
| beta1: 0.9 |
| beta2: 0.999 |
| weight_decay: 1.0e-4 |
| beta1_critic: 0.9 |
| beta2_critic: 0.999 |
| vis_interval: 1 |
| eval_freq: 2 |
| teacher_v_loss_weight: 0 |
| beta: 0.1 |
| decay_type: 0 |
| grpo: |
| num_samples_per_prompt: 8 |
| grpo_epochs: 1 |
| kl_coef: 1.0e-4 |
| adv_clip_max: 5.0 |
| use_per_prompt_stats: true |
| global_std: true |
| timestep_fraction: 1 |
|
|
| data: |
| teacher_videos_dir: "/mnt/vepfs/base2/stream/luoyang/GAD-Video-Data-200-OpenVid-1M" |
| video_format: "mp4" |
| use_subfolder_format: true |
| train_split_ratio: 0.8 |
| split_seed: 42 |
| num_frames: 81 |
| height: 480 |
| width: 832 |
| fps: 16 |
| max_prompt_length: 512 |
| num_workers: 4 |
|
|
| checkpoint: |
| output_dir: "/mnt/vepfs/base2/stream/luoyang/checkpoints/afd_student_openvid1m_lr1e-5" |
| save_total_limit: 10 |
| resume_from: null |
|
|
| logging: |
| output_dir: "/mnt/vepfs/base2/stream/luoyang/logs/afd_student_openvid1m" |
| wandb_project: "video-gad-student" |
| wandb_run_name: null |
|
|
| wandb: |
| project: "video-gad-student" |
| entity: null |
| key: null |
| save_dir: "./wandb_logs" |
|
|
| disable_wandb: false |
| seed: 42 |
| experiment_name: "afd_student_openvid1m_lr1e-5_frozen_disc" |
|
|
