| defaults: | |
| - base_config | |
| - override data: base | |
| - override hydra/job_logging: custom | |
| - _self_ | |
| hydra: | |
| run: | |
| dir: ./output/${exp_id} | |
| output_subdir: train-${now:%Y-%m-%d_%H-%M-%S}-hydra | |
| ema: | |
| start: 0 | |
| mini_train: False | |
| example_train: False | |
| enable_grad_scaler: False | |
| vgg_oversample_rate: 4 | |
| log_text_interval: 200 | |
| log_extra_interval: 20_000 | |
| val_interval: 5_000 | |
| eval_interval: 20_000 | |
| save_eval_interval: 40_000 | |
| save_weights_interval: 10_000 | |
| save_checkpoint_interval: 10_000 | |
| save_copy_iterations: [40000,60000,80000,100000,150000,200000,220000,240000,260000,280000,300000] | |
| batch_size: 320 | |
| eval_batch_size: 32 # per-GPU | |
| num_iterations: 220_000 | |
| learning_rate: 1.0e-4 | |
| linear_warmup_steps: 1_000 | |
| lr_schedule: step | |
| lr_schedule_steps: [200_000, 240_000, 270_000] | |
| lr_schedule_gamma: 0.1 | |
| clip_grad_norm: 1.0 | |
| weight_decay: 1.0e-6 |