| data_root_dir: /mnt/blob/open_x | |
| action_model_type: DiT-B | |
| future_action_window_size: 15 | |
| hf_token: HF_TOKEN | |
| image_aug: true | |
| is_resume: false | |
| load_all_data_for_training: true | |
| past_action_window_size: 0 | |
| pretrained_checkpoint: /mnt/blob/vla_model/openvla-7b-prismatic/checkpoints/step-295000-epoch-40-loss=0.2200.pt | |
| repeated_diffusion_steps: 8 | |
| resume_epoch: null | |
| resume_step: null | |
| run_id: prism-dinosiglip-224px+oxe+diffusion+n2+b16+x42--image_aug | |
| run_id_note: null | |
| run_root_dir: /mnt/blob/cogact/ditb_8_lr2e-5_b16_fa15_pa0_shuffle_oxe | |
| save_interval: 2500 | |
| seed: 42 | |
| trackers: | |
| - jsonl | |
| - wandb | |
| vla: | |
| base_vlm: prism-dinosiglip-224px+7b | |
| data_mix: oxe_magic_soup_plus_minus | |
| enable_gradient_checkpointing: true | |
| enable_mixed_precision_training: true | |
| epochs: 100 | |
| expected_world_size: 16 | |
| freeze_llm_backbone: false | |
| freeze_vision_backbone: false | |
| global_batch_size: 256 | |
| learning_rate: 2.0e-05 | |
| lr_scheduler_type: constant | |
| max_grad_norm: 1.0 | |
| max_steps: null | |
| per_device_batch_size: 16 | |
| reduce_in_full_precision: true | |
| shuffle_buffer_size: 250000 | |
| train_strategy: fsdp-full-shard | |
| type: prism-dinosiglip-224px+oxe+diffusion | |
| unfreeze_last_llm_layer: false | |
| vla_id: prism-dinosiglip-224px+oxe+diffusion | |
| warmup_ratio: 0.0 | |
| weight_decay: 0.0 | |
| wandb_entity: null | |
| wandb_project: null | |