| {"data_dir": "/home/jovyan/foundation_datasets_new", "context_len": 4096, "trajectory_sparsity": 800, "ignore_episode_sparsity": false, "preload": false, "ic_treshold": 0.3, "last_frac": null, "exp_treshold": 0.75, "action_emb_dim": 767, "observation_emb_dim": 767, "reward_emb_dim": 2, "hidden_dim": 1536, "transformer_depth": 16, "transformer_heads": 24, "attn_dropout": 0.0, "residual_dropout": 0.0, "normalize_qk": true, "bias": true, "parallel_residual": false, "shared_attention_norm": false, "norm_class": "LayerNorm", "mlp_class": "GptNeoxMLP", "intermediate_size": 6144, "inner_ep_pos_enc": false, "norm_acs": false, "norm_obs": true, "add_bos": true, "trans_prior": false, "optimizer": "Adam", "lr": 5e-05, "betas": [0.9, 0.99], "weight_decay": 0.0001, "precision": "bf16", "clip_grad": 2.5, "grad_accum_steps": 1, "warmup_ratio": 0.006, "local_rank": 0, "epochs": 150, "batch_size": 8, "eval_every": 6, "val_max_trans": 10000, "smooth_win": 8, "save_every": 3, "save_dir": "/home/jovyan/found_model_checkpoints/frl_d16h24h1536c4096_dpt_fm_10domains", "stats_path": "/home/jovyan/foundation_datasets_new/stats.json", "load_ckpt": null, "start_epoch": 0, "seed": 5, "permute": true, "acs_distrib_log_freq": 500, "acs_avg_window_size": 1, "extended_logging_dpt": false, "dataset_config_paths": ["vintix/data/configs/metaworld_config.yaml", "vintix/data/configs/mujoco_config.yaml", "vintix/data/configs/ib_config.yaml", "vintix/data/configs/bidex_config.yaml", "vintix/data/configs/humenv_config.yaml", "vintix/data/configs/kinetix_config.yaml", "vintix/data/configs/metadrive_config.yaml", "vintix/data/configs/citylearn_config.yaml", "vintix/data/configs/controlgym_config.yaml", "vintix/data/configs/sinergym_config.yaml"], "project": "foundation_rl", "group": "dpt_fm", "name": "frl_d16h24h1536c4096_dpt_fm_10domains", "model_type": "FMDPT"} |