| { | |
| "d_model": 512, | |
| "nhead": 8, | |
| "num_layers": 6, | |
| "dim_feedforward": 2048, | |
| "dropout": 0.2, | |
| "encoder_type": "hal_flat", | |
| "d_intra": 256, | |
| "encoder_nlayers": 2, | |
| "k_query": 1, | |
| "scaled_emb": false, | |
| "max_seq_len": 256, | |
| "pos_enc": "relpos", | |
| "rope_theta": 10000.0, | |
| "rope_learnable_freqs": false, | |
| "xpos_scale_base": 0.0, | |
| "attn_variant": "standard", | |
| "n_kv_heads": 0, | |
| "use_rmsnorm": false, | |
| "use_swiglu": false, | |
| "stick_loss": "clusters", | |
| "btn_loss": "focal", | |
| "no_opp_inputs": true, | |
| "no_self_inputs": false, | |
| "head_hidden": 256, | |
| "n_stick_clusters": 37, | |
| "n_shoulder_bins": 3, | |
| "autoregressive_heads": true, | |
| "hal_mode": true, | |
| "lean_features": false, | |
| "hal_minimal_features": true, | |
| "hal_controller_encoding": true, | |
| "n_controller_combos": 7, | |
| "num_stages": 6, | |
| "num_ports": 4, | |
| "num_characters": 27, | |
| "num_actions": 396, | |
| "num_costumes": 6, | |
| "num_c_dirs": 9, | |
| "num_proj_types": 103, | |
| "num_proj_subtypes": 40, | |
| "model_preset": "hal", | |
| "run_name": "cptfalcon-7class-v2" | |
| } |