|
{ |
|
"accumuate_gradient_steps": 2, |
|
"comment": "", |
|
"count_flops": false, |
|
"customized_dataset_mixture_file": "", |
|
"dataloader": { |
|
"batch_size": 64, |
|
"drop_last": true, |
|
"num_workers": 1, |
|
"persistent_workers": false, |
|
"pin_memory": false |
|
}, |
|
"dataset": { |
|
"_target_": "hpt_pretrain.dataset.traj_dataset.TrajDataset", |
|
"action_horizon": 8, |
|
"continue_generate": false, |
|
"data_augmentation": false, |
|
"data_ratio": 1, |
|
"dataset_encoder_postfix": "_resnet", |
|
"dataset_postfix": "_traj100000", |
|
"episode_cnt": 100000, |
|
"heldout_dataset": false, |
|
"horizon": 5, |
|
"image_encoder": "resnet", |
|
"normalize_state": true, |
|
"observation_horizon": 4, |
|
"pad_after": 0, |
|
"pad_before": 0, |
|
"precompute_feat": true, |
|
"regenerate": false, |
|
"step_cnt": 10000000, |
|
"use_disk": true, |
|
"use_heldout_dataset": true, |
|
"use_multiview": false, |
|
"val_ratio": 0.1 |
|
}, |
|
"dataset_groups": "", |
|
"dataset_shuffle": true, |
|
"ddp_dataloader": { |
|
"drop_last": false, |
|
"num_workers": 8, |
|
"persistent_workers": false, |
|
"pin_memory": false, |
|
"prefetch_factor": 6 |
|
}, |
|
"ddp_val_dataloader": { |
|
"drop_last": false, |
|
"num_workers": 8, |
|
"persistent_workers": false, |
|
"pin_memory": false, |
|
"prefetch_factor": 2 |
|
}, |
|
"debug_distributed": false, |
|
"debug_submitit": false, |
|
"dist_backend": "nccl", |
|
"dist_on_itp": false, |
|
"dist_url": "file:///checkpoint/xinleic/experiments/855d1a9c880549309c4f301b0fe381be_init", |
|
"distributed": true, |
|
"domains": "austin_sailor_dataset_converted_externally_to_rlds, stanford_hydra_dataset_converted_externally_to_rlds, austin_buds_dataset_converted_externally_to_rlds, austin_sirius_dataset_converted_externally_to_rlds, berkeley_mvp_converted_externally_to_rlds, berkeley_rpt_converted_externally_to_rlds, iamlab_cmu_pickup_insert_converted_externally_to_rlds, utaustin_mutex, imperialcollege_sawyer_wrist_cam, stanford_mask_vit_converted_externally_to_rlds, language_table, kuka, bc_z, robo_net, dlr_sara_pour_converted_externally_to_rlds, stanford_robocook_converted_externally_to_rlds, cmu_play_fusion, bridge, furniture_bench_dataset_converted_externally_to_rlds, ucsd_pick_and_place_dataset_converted_externally_to_rlds, usc_cloth_sim_converted_externally_to_rlds, stanford_kuka_multimodal_dataset_converted_externally_to_rlds, roboturk, kaist_nonprehensile_converted_externally_to_rlds, asu_table_top_converted_externally_to_rlds, utokyo_xarm_pick_and_place_converted_externally_to_rlds, berkeley_cable_routing, droid, uiuc_d3field, robo_set, qut_dexterous_manpulation, nyu_door_opening_surprising_effectiveness, nyu_franka_play_dataset_converted_externally_to_rlds, mimic_play, maniskill_dataset_converted_externally_to_rlds, columbia_cairlab_pusht_real, conq_hose_manipulation, dlr_edan_shared_control_converted_externally_to_rlds, berkeley_gnm_sac_son, berkeley_autolab_ur5, aloha_mobile, agent_aware_affordances", |
|
"dset_w_temperature": 2.0, |
|
"effective_batch_size": 256, |
|
"effective_total_epochs": 100, |
|
"epoch_size": 10, |
|
"fsdp": false, |
|
"gpu": 0, |
|
"head": { |
|
"_target_": "hpt_pretrain.models.policy_head.MLP", |
|
"dropout": true, |
|
"input_dim": 512, |
|
"output_dim": 48, |
|
"tanh_end": true, |
|
"widths": [ |
|
256, |
|
128 |
|
] |
|
}, |
|
"job_dir": "logs/", |
|
"local_rank": 1, |
|
"log_dir": "output/05_05_2024_20_21_54_727275", |
|
"log_interval": 10, |
|
"lr_scheduler": { |
|
"T_max": 800000, |
|
"_target_": "torch.optim.lr_scheduler.CosineAnnealingLR", |
|
"eta_min": 1e-06 |
|
}, |
|
"network": { |
|
"_target_": "hpt_pretrain.models.policy.Policy", |
|
"action_horizon": 8, |
|
"drop_path": 0.1, |
|
"embed_dim": 512, |
|
"mae_loss_scale": 0.0, |
|
"masked_autoencoding": false, |
|
"num_blocks": 16, |
|
"num_heads": 8, |
|
"token_postprocessing": "mean", |
|
"use_domain_embedding": false, |
|
"use_modality_embedding": true, |
|
"weight_init_style": "pytorch" |
|
}, |
|
"ngpus": 8, |
|
"nodelist": "learnlab,learnfair,scavenge", |
|
"nodes": 4, |
|
"optimizer": { |
|
"_target_": "torch.optim.AdamW", |
|
"eps": 1e-06, |
|
"lr": 0.001, |
|
"weight_decay": 0.05 |
|
}, |
|
"optimizer_misc": { |
|
"nontrunk_lr_scale": 0.5 |
|
}, |
|
"output_dir": "output/05_05_2024_20_21_54_727275", |
|
"partition": "learnlab", |
|
"rank": 0, |
|
"resume": "logs/", |
|
"save_wb_checkpoint": true, |
|
"script_name": "run_resnet_30dataset_traj100000_embed_512_batch2048_800k", |
|
"seed": 0, |
|
"slurm_job_id": "27155766", |
|
"stem": { |
|
"add_pos_embedding_to_state": false, |
|
"crossattn_dim_head": 64, |
|
"crossattn_heads": 8, |
|
"crossattn_latent": { |
|
"image": 16, |
|
"state": 16 |
|
}, |
|
"crossattn_modality_dropout": 0.1, |
|
"image": { |
|
"_target_": "hpt_pretrain.models.policy_stem.MLP", |
|
"input_dim": 512, |
|
"num_of_copy": 1, |
|
"output_dim": 512, |
|
"widths": [ |
|
128 |
|
] |
|
}, |
|
"image_encoder": "resnet", |
|
"modalities": [ |
|
"image", |
|
"state" |
|
], |
|
"modality_embed_dim": 512, |
|
"normalize_state": true, |
|
"num_blocks": 1, |
|
"observation_horizon": 4, |
|
"random_horizon_masking": true, |
|
"state": { |
|
"_target_": "hpt_pretrain.models.policy_stem.MLP", |
|
"input_dim": 8, |
|
"output_dim": 512, |
|
"widths": [ |
|
128 |
|
] |
|
}, |
|
"state_embedding_dim": 1 |
|
}, |
|
"submitit_autoresume": true, |
|
"task_per_gpu": 1, |
|
"timeout": 4320, |
|
"total_num_sample": 0, |
|
"total_num_traj": 0, |
|
"train": { |
|
"epoch_iters": 1000, |
|
"freeze_trunk": false, |
|
"load_pretrain_trunk_only": false, |
|
"max_validation_size": 10, |
|
"pretrained_dir": "05_05_2024_20_21_54_727275", |
|
"total_epochs": 30000, |
|
"total_iters": 800000, |
|
"use_accumulation": false, |
|
"validation_iters": 100 |
|
}, |
|
"trunk_acc_gradient_step": 1, |
|
"trunk_separate_opt": false, |
|
"use_volta32": true, |
|
"val_dataloader": { |
|
"num_workers": 1, |
|
"persistent_workers": false, |
|
"pin_memory": false |
|
}, |
|
"warmup_lr": { |
|
"lr": 1e-10, |
|
"step": 1000 |
|
}, |
|
"wb_cont_run": "3tv0j6ts", |
|
"wb_tag": "default", |
|
"world_size": 32 |
|
} |