| config: {'root_dir': '/home/leonmkim/fish_leon', 'replay_buffer_size': 150000, 'replay_buffer_num_workers': 2, 'nstep': 3, 'batch_size': 128, 'seed': 0, 'dataset_shuffle_seed': 0, 'device': 'cuda', 'save_video': True, 'save_train_video': True, 'use_tb': True, 'use_wandb': True, 'wandb_run_id': '497_0', 'wandb_notes': '497_0_req_580_0', 'eval': True, 'true_action_history': False, 'process_contact_features': True, 'obs_type': 'pixels', 'use_color': True, 'use_depth': True, 'use_masks': False, 'mask_list': ['EE_obj_mask'], 'mask_representation': 'channels', 'crop_hw': [144, 144], 'crop_down_offset': 48, 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'add_crop_binary_mask': False, 'add_coord_conv_map': False, 'use_context_color': False, 'use_context_depth': False, 'use_context_segmask': False, 'context_color_crop_type': None, 'context_depth_crop_type': None, 'context_segmask_crop_type': None, 'context_add_crop_binary_mask': False, 'context_add_coord_conv_map': False, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'max_contact_prob': 0.1, 'max_depth': 2.0, 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'dtc_adaptive_normalization': False, 'mask_normals_within_sdf': True, 'adaptive_normals_mask': True, 'learnable_contact_preprocess_params': True, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'contact_estimation_model_ckpt_path': '~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt', 'encoder_type': 'small', 'debug_timestamps': False, 'open_loop': False, 'action_trajectories': True, 'stop_after_action': False, 'interpolation_frequency': 25, 'policy_frequency': 5, 'wait_for_new_camera_frames': True, 'baseline': False, 'train_demo_idxs_list_or_num': -1, 'log_train_every_steps': 25, 'name_of_expert_demo': '112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'expert_dataset_dirpath': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'store_dataset_in_memory': False, 'expert_dataset': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'action_key': 'action_trajectory_25hz', 'semantic_demo_grouping_name': 'semantic_demo_grouping.yaml', 'semantic_demo_grouping': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/semantic_demo_grouping.yaml', 'include_groups_list': ['german_twodim'], 'expert_dataset_config': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demo_config.yaml', 'name_of_valid_demo': '112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act', 'valid_dataset_dir': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'valid_demo_idxs_list_or_num': None, 'val_num_groups': 0, 'load_bc': True, 'checkpoint_epoch_list': [99, 199, 299, 399, 499, 599, 699, 799, 899, 999, 1249, 1499, 1749, 1999, 2999, 3999, 4999, 5999, 6999, 7999, 8999, 9999], 'snapshot_root_dir': '/mnt/kostas-graid/datasets/extrinsic_contact_data/FISH', 'save_snapshot': True, 'save_last_snapshot': True, 'save_snapshot_when_done': True, 'top_k_checkpoints': 5, 'save_snapshot_link_to_weights_dir': 'deprecated', 'bc_regularize': False, 'bc_weight_type': 'qfilter', 'experiment_dir': './exp_local/frankagym_pixels/FrankaInsertion-v1/497_0', 'agent': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgent', 'name': 'diffusion_policy', 'load_checkpoint': True, 'device': 'cuda', 'n_obs_steps': 1, 'suite_name': 'frankagym', 'obs_type': 'pixels', 'enable_arm': True, 'enable_camera': True, 'use_tb': True, 'desired_image_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'config': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgentConfig', 'compile': False, 'device': 'cuda', 'cam_resize_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'policy_frequency': 5, 'interpolation_frequency': 25, 'policy_cfg': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig', 'n_obs_steps': 1, 'horizon': 36, 'n_action_steps': 36, 'output_shapes': {'action': [7]}, 'input_normalization_modes': {'observation.image': 'mean_std', 'observation.state': 'min_max', 'observation.action_history': 'min_max'}, 'output_normalization_modes': {'action': 'min_max'}, 'vision_backbone': 'resnet18', 'pretrained_backbone_weights': None, 'transforms': [{'_target_': 'torchaug.transforms.RandomAffine', 'degrees': [-5, 5], 'translate': [0.05, 0.05], 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}, {'_target_': 'torchaug.transforms.RandomColorJitter', 'brightness': 0.3, 'contrast': 0.4, 'saturation': 0.5, 'hue': 0.08, 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}], 'use_group_norm': True, 'spatial_softmax_num_keypoints': 32, 'action_history_encoder_config': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig', 'in_channels': 7, 'out_channels': 32, 'history_length': 6, 'kernel_size': 5, 'downsample_kernel_size': 3, 'downsample_stride': 2, 'downsample_padding': 1}, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'diffusion_step_embed_dim': 128, 'use_film_scale_modulation': True, 'noise_scheduler_type': 'DDIM', 'beta_schedule': 'squaredcos_cap_v2', 'beta_start': 0.0001, 'beta_end': 0.02, 'prediction_type': 'epsilon', 'clip_sample': True, 'clip_sample_range': 1.0, 'num_train_timesteps': 50, 'num_inference_steps': 10, 'do_mask_loss_for_padding': False, 'input_shapes': {'observation.image': [13, 180, 240], 'context_observation.image': [13, 180, 240], 'observation.state': [8], 'observation.action_history': [7]}}, 'train_cfg': {'_target_': 'utils.TrainConfig', 'lr': 0.0001, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'adam_betas': [0.95, 0.999], 'adam_eps': 1e-08, 'adam_weight_decay': 1e-06, 'grad_clip_norm': 10, 'offline_steps': 1000000, 'use_amp': True}, 'observation_cfg': {'_target_': 'agent.encoder.VisualFeatureSet', 'use_depth': True, 'use_color': True, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}, 'context_input_config': {'_target_': 'agent.encoder.ContextInputConfig', 'use_color': False, 'use_depth': False, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}}, 'mask_soft_approx_scheduler_config': {'_target_': 'agent.encoder.MaskSoftApproxSchedulerConfig', 'num_steps': 40000, 'initial_value': 10.0, 'final_value': 1000.0, 'interpolation_scheme': 'cosine'}, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'max_contact_prob': 0.1, 'mask_normals_within_sdf': True, 'dtc_adaptive_normalization': False, 'adaptive_normals_mask': True, 'max_depth': 2.0, 'image_shape': [13, 180, 240], 'learnable_contact_preprocess_params': True, 'learning_rate': 0.0001, 'weight_decay': 0.0, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'zero_centered': False}}}, 'suite': {'suite': 'frankagym', 'name': 'frankagym', 'frame_stack': 1, 'action_repeat': 1, 'discount': 0.99, 'hidden_dim': 1024, 'num_train_frames': 2010, 'num_seed_frames': 260, 'num_train_epochs': 5000, 'validate_every_epochs': 100, 'validate_diffusion_on_action_loss_every_epochs': 500, 'train_eval_diffusion_on_action_loss_every_epochs': 500, 'check_topk_every_epochs': 10, 'save_snapshot_every_epochs': 5000, 'eval_every_frames': 2000, 'num_eval_episodes': 5, 'save_snapshot': True, 'wait_for_user_to_start_episode': True, 'task_make_fn': {'_target_': 'suite.frankagym.make', 'name': 'FrankaInsertion-v1', 'height': 240, 'width': 320, 'frame_stack': 1, 'action_repeat': 1, 'seed': 0, 'enable_arm': True, 'enable_gripper': True, 'start_with_gripper_open': True, 'enable_camera': True, 'path_to_depth_extrinsics': '~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy', 'contact_estimation_model_ckpt_path': '~/fish_leon/contact_estimation/artifacts/175604_2/checkpoints/epoch=09-val_loss=0.00.ckpt', 'x_limit': [0.2, 0.7], 'y_limit': [-0.4, 0.4], 'z_limit': [-0.05, 0.55], 'device': 'cuda', 'interpolation_frequency': 25, 'policy_frequency': 5, 'debug_timestamps': False, 'stop_after_action': False, 'open_loop': False, 'wait_for_new_camera_frames': True, 'action_key': 'action_trajectory_25hz', 'action_trajectory_horizon': 36, 'action_trajectories': True, 'path_to_zarr_dataset': '/home/leonmkim/fish_leon/FISH/expert_demos/frankagym/FrankaInsertion-v1/112_240x320_all_twodim_left_to_right_annotated_start_idx_5hz_zstd7_EE_pxl_coords_expert_demos_imp_act/demos.zarr', 'agent_policy_cfg': {'_target_': 'agent.diffusion_policy.DiffusionPolicyAgentConfig', 'compile': False, 'device': 'cuda', 'cam_resize_shape': [13, 180, 240], 'orig_cam_shape': [3, 240, 320], 'policy_frequency': 5, 'interpolation_frequency': 25, 'policy_cfg': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.DiffusionConfig', 'n_obs_steps': 1, 'horizon': 36, 'n_action_steps': 36, 'output_shapes': {'action': [7]}, 'input_normalization_modes': {'observation.image': 'mean_std', 'observation.state': 'min_max', 'observation.action_history': 'min_max'}, 'output_normalization_modes': {'action': 'min_max'}, 'vision_backbone': 'resnet18', 'pretrained_backbone_weights': None, 'transforms': [{'_target_': 'torchaug.transforms.RandomAffine', 'degrees': [-5, 5], 'translate': [0.05, 0.05], 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}, {'_target_': 'torchaug.transforms.RandomColorJitter', 'brightness': 0.3, 'contrast': 0.4, 'saturation': 0.5, 'hue': 0.08, 'batch_transform': True, 'num_chunks': -1, 'batch_inplace': True}], 'use_group_norm': True, 'spatial_softmax_num_keypoints': 32, 'action_history_encoder_config': {'_target_': 'lerobot.common.policies.diffusion.configuration_diffusion.Unet1dEncoderConfig', 'in_channels': 7, 'out_channels': 32, 'history_length': 6, 'kernel_size': 5, 'downsample_kernel_size': 3, 'downsample_stride': 2, 'downsample_padding': 1}, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'diffusion_step_embed_dim': 128, 'use_film_scale_modulation': True, 'noise_scheduler_type': 'DDIM', 'beta_schedule': 'squaredcos_cap_v2', 'beta_start': 0.0001, 'beta_end': 0.02, 'prediction_type': 'epsilon', 'clip_sample': True, 'clip_sample_range': 1.0, 'num_train_timesteps': 50, 'num_inference_steps': 10, 'do_mask_loss_for_padding': False, 'input_shapes': {'observation.image': [13, 180, 240], 'context_observation.image': [13, 180, 240], 'observation.state': [8], 'observation.action_history': [7]}}, 'train_cfg': {'_target_': 'utils.TrainConfig', 'lr': 0.0001, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'adam_betas': [0.95, 0.999], 'adam_eps': 1e-08, 'adam_weight_decay': 1e-06, 'grad_clip_norm': 10, 'offline_steps': 1000000, 'use_amp': True}, 'observation_cfg': {'_target_': 'agent.encoder.VisualFeatureSet', 'use_depth': True, 'use_color': True, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}, 'context_input_config': {'_target_': 'agent.encoder.ContextInputConfig', 'use_color': False, 'use_depth': False, 'mask_input_dict': {'_target_': 'agent.encoder.MaskInputDict', 'enable': False, 'representation': 'channels', 'mask_list': ['EE_obj_mask']}, 'crop_input_config': {'_target_': 'agent.encoder.CropInputConfig', 'color_crop_type': None, 'depth_crop_type': None, 'segmask_crop_type': None, 'crop_hw': [144, 144], 'crop_down_offset': 48, 'add_crop_binary_mask': False, 'add_coord_conv_map': False}}, 'mask_soft_approx_scheduler_config': {'_target_': 'agent.encoder.MaskSoftApproxSchedulerConfig', 'num_steps': 40000, 'initial_value': 10.0, 'final_value': 1000.0, 'interpolation_scheme': 'cosine'}, 'use_contact_map': False, 'use_sdf_maps': False, 'use_normals_maps': False, 'which_objects': 'both', 'grasped_dtc_max_value': 0.2, 'env_dtc_max_value': 0.4, 'grasped_normals_mask_max_dtc_value': 0.2, 'env_normals_mask_max_dtc_value': 0.4, 'clamp_dtc': True, 'max_contact_prob': 0.1, 'mask_normals_within_sdf': True, 'dtc_adaptive_normalization': False, 'adaptive_normals_mask': True, 'max_depth': 2.0, 'image_shape': [13, 180, 240], 'learnable_contact_preprocess_params': True, 'learning_rate': 0.0001, 'weight_decay': 0.0, 'contact_model_name': 'local_multitask_outhd64all_home_crop_h144w144d48_mask_ctxtmask_seed_220979_epoch_9', 'zero_centered': False}}, 'true_action_history': False}}, 'num_train_frames_bc': 50000, 'num_train_frames_drq': 1100000, 'stddev_schedule_drq': 'linear(1.0,0.1,100000)', 'task_name': 'FrankaInsertion-v1', 'num_train_frames_vinn': 25000, 'num_train_frames_diffusion': 1000000, 'num_train_epochs_bc': 5000, 'num_train_epochs_diffusion': 15000, 'validate_every_epochs_bc': 5, 'validate_every_epochs_diffusion': 250, 'validate_diffusion_on_action_loss_every_epochs': 250, 'train_eval_diffusion_on_action_loss_every_epochs': 250, 'check_topk_every_epochs': 5, 'check_topk_every_epochs_diffusion': 250, 'save_snapshot_every_epochs_diffusion': 1500, 'x_limit': [0.2, 0.7], 'y_limit': [-0.4, 0.4], 'z_limit': [-0.05, 0.55], 'home_displacement': [0.55, 0.0, 0.55, 180.0, 0.0, 0.0], 'enable_gripper': True, 'start_with_gripper_open': True, 'offset_mask': [1, 1, 1, 1, 1, 1], 'path_to_depth_extrinsics': '~/fish_leon/FISH/cfgs/camera_poses/camera_poses_L515/20240904-122305/color_tf_world.npy', 'feature_type': '180x240_1_RGB_D_2.0_acthst_hst6_out32_dwnkrnl3_dwnstrd2_dwnpd1', 'save_buffer': True, 'num_eval': 5, 'random_start': False, 'eval_starts': '/home/leonmkim/fish_leon/FISH/eval_starts/frankagym_pixels/FrankaInsertion-v1', 'num_valid_demos': None, 'load_checkpoint': True, 'checkpoint_epoch': 12000, 'load_residual_weight': False, 'checkpoint_root_dir': '/home/leonmkim/fish_leon/FISH', 'checkpoint_weight_dir': '/home/leonmkim/fish_leon/FISH/exp_local/frankagym_pixels/FrankaInsertion-v1/497_0', 'residual_weight': '/home/leonmkim/fish_leon/FISH/weights/frankagym_pixels/FrankaInsertion-v1/weight.pt', 'final_experiment_dir': './exp_local/frankagym_pixels/FrankaInsertion-v1/497_0/201837'} |
