| # Important: Be careful when modifying this file! The fields in file will be overridden by the dataset dependent config file in the `configurations/dataset_experiment` folder so consider making changes there instead! | |
| defaults: | |
| - base_pytorch_algo | |
| - backbone: dit3d | |
| # dataset-dependent configurations | |
| x_shape: ${dataset.observation_shape} | |
| max_frames: ${dataset.max_frames} | |
| n_frames: ${dataset.n_frames} | |
| frame_skip: ${dataset.frame_skip} | |
| context_frames: ${dataset.context_length} | |
| latent: ${dataset.latent} | |
| data_mean: ${dataset.data_mean} | |
| data_std: ${dataset.data_std} | |
| external_cond_dim: ${dataset.external_cond_dim} | |
| external_cond_stack: ${dataset.external_cond_stack} | |
| external_cond_processing: ${dataset.external_cond_processing} | |
| # training hyperparameters | |
| compile: False | |
| weight_decay: 1e-3 | |
| optimizer_beta: [0.9, 0.99] | |
| lr_scheduler: | |
| name: constant_with_warmup | |
| num_warmup_steps: 5000 | |
| noise_level: random_independent | |
| uniform_future: | |
| enabled: false | |
| fixed_context: | |
| enabled: false | |
| indices: null # defaults to context_frames | |
| dropout: 0 | |
| variable_context: | |
| enabled: false | |
| prob: 0 | |
| dropout: 0 | |
| # sampling scheme | |
| chunk_size: -1 | |
| scheduling_matrix: full_sequence | |
| replacement: noisy_scale | |
| diffusion: | |
| is_continuous: False | |
| timesteps: 1000 | |
| beta_schedule: cosine | |
| schedule_fn_kwargs: | |
| shift: 1.0 | |
| use_causal_mask: False | |
| clip_noise: 20.0 | |
| # training | |
| objective: pred_v | |
| loss_weighting: | |
| strategy: fused_min_snr | |
| snr_clip: 5.0 | |
| cum_snr_decay: 0.9 | |
| # sampling | |
| sampling_timesteps: 50 | |
| ddim_sampling_eta: 0.0 | |
| # (For full sequence diffusion) | |
| reconstruction_guidance: 0.0 | |
| # vae | |
| vae: | |
| pretrained_path: null | |
| pretrained_kwargs: {} | |
| use_fp16: True | |
| batch_size: 2 | |
| checkpoint: | |
| reset_optimizer: False | |
| strict: True | |
| # video generation tasks | |
| tasks: | |
| prediction: | |
| enabled: True | |
| history_guidance: | |
| name: conditional | |
| keyframe_density: null | |
| sliding_context_len: null | |
| interpolation: | |
| enabled: False | |
| history_guidance: | |
| name: conditional | |
| max_batch_size: null | |
| # logging | |
| logging: | |
| deterministic: 0 | |
| loss_freq: 100 | |
| grad_norm_freq: 100 | |
| max_num_videos: 8 | |
| n_metrics_frames: null | |
| metrics: | |
| - fvd | |
| - is | |
| - fid | |
| - lpips | |
| - mse | |
| - psnr | |
| - ssim | |
| metrics_batch_size: 16 | |
| sanity_generation: False # generate video samples during sanity check | |
| raw_dir: null | |