| global_seed: 0 | |
| data: | |
| batch_size: 16 | |
| eval_batch_size: 5 | |
| image_size: 256 | |
| num_workers: 8 | |
| imagenet_train_index: "imagenet_train_index_overall.json" | |
| imagenet_train_tar: "ILSVRC2012_img_train.tar" | |
| imagenet_val_index: "imagenet_val_index_overall.json" | |
| imagenet_val_tar: "ILSVRC2012_img_val.tar" | |
| model: | |
| enable_mup: True | |
| patch_size: 4 | |
| mup_width: 6 | |
| enc_mup_width: 4 | |
| fix_initial_norms: True | |
| fix_norm_mode: l2 | |
| quantization_type: lfq | |
| code_length: 256 | |
| context_dim: 18 | |
| codebook_size_for_entropy: 9 | |
| entropy_loss_weight: 0.0025 | |
| commit_loss_weight: 0.000625 | |
| enc_depth: 8 | |
| dec_depth: 16 | |
| enable_cfg: True | |
| ema_decay: 0.9999 | |
| posttrain_sample: False | |
| posttrain_sample_enable_cfg: False | |
| posttrain_sample_k: 8 | |
| opt: | |
| # Floats need to be specified with decimals to get loaded as such | |
| lr: 1.0e-4 | |
| n_grad_acc: 2 | |
| schedule: fat_lognormal | |
| freeze_encoder: False | |
| freeze_encoder_after: 200000 | |
| weight_decay: 0.0 | |
| log_norms: True | |
| lpips_weight: 0.1 | |
| lpips_mode: "vgg" | |
| beta1: .9 | |
| beta2: .95 | |
| trainer: | |
| enable_bfloat16: True | |
| log_every: 100 | |
| checkpoint_every: 5000 | |
| max_steps: 10000000000000 | |
| keep_every: 200000 | |
| gs_checkpoint_bucket: "" | |
| eval: | |
| reconstruction: True | |
| state_dict_key: model_ema_state_dict | |
| eval_dir: "" | |
| eval_baseline: "" | |
| continuous: True | |
| force_ckpt_path: null | |
| subsample_rate: 1 | |
| sampling: | |
| sample_steps: 25 | |
| schedule: "pow_0.25" | |
| cfg: 1.5 | |
| mode: 'rf' | |
| cfg_interval: "(.17,1.02)" | |