| model: | |
| model_id: dcae | |
| sample_size: [360, 640] | |
| channels: 3 | |
| latent_size: 32 | |
| latent_channels: 16 | |
| ch_0: 256 | |
| ch_max: 2048 | |
| encoder_blocks_per_stage: [4, 4, 4, 8] | |
| decoder_blocks_per_stage: [4, 4, 4, 8] | |
| use_middle_block: false | |
| do_channel_mask: false | |
| train: | |
| trainer_id: rec | |
| data_id: video_dir_loader | |
| data_kwargs: | |
| source: | |
| - /mnt/data/datasets/extracted_tars/kbm/fps/*/*.mp4 | |
| - /mnt/data/datasets/extracted_tars/kbm/3ps/*/*.mp4 | |
| - /mnt/data/datasets/extracted_tars/kbm/other/*/*.mp4 | |
| target_size: [360, 640] | |
| target_batch_size: 32 | |
| batch_size: 4 | |
| epochs: 200 | |
| #opt: AdamW | |
| #opt_kwargs: | |
| # lr: 3.0e-5 | |
| # weight_decay: 1.0e-4 | |
| # betas: [0.9, 0.95] | |
| # eps: 1.0e-15 | |
| opt: Muon | |
| opt_kwargs: | |
| lr: 1.0e-3 | |
| momentum: 0.95 | |
| adamw_lr: 1.0e-5 | |
| adamw_wd: 1.0e-2 | |
| adamw_eps: 1.0e-6 | |
| adamw_betas: [0.9, 0.95] | |
| adamw_keys: | |
| - encoder.conv_in | |
| - encoder.conv_out | |
| - encoder.conv_out_logvar | |
| - decoder.conv_in | |
| - decoder.conv_out | |
| - .up. | |
| - .down. | |
| - .residuals. | |
| lpips_type: convnext | |
| loss_weights: | |
| kl: 3.0e-7 | |
| lpips: 12.0 | |
| l2: 1.0 | |
| dwt: 0.25 | |
| scheduler: LinearWarmup | |
| scheduler_kwargs: | |
| warmup_steps: 3000 | |
| min_lr: 1.0e-5 | |
| checkpoint_dir: checkpoints/waypoint_1_vae_owlc_f16_c16 | |
| resume_ckpt: checkpoints/waypoint_1_vae_owlc_f16_c16/step_200000.pt | |
| sample_interval: 1000 | |
| save_interval: 5000 | |
| wandb: | |
| name: shahbuland | |
| project: new_vaes_v2 | |
| run_name: waypoint_1_vae_owlc_f16_c16_highres |