| { | |
| "activation_function": "gelu", | |
| "architectures": [ | |
| "PatchTSTForPretraining" | |
| ], | |
| "attention_dropout": 0.0, | |
| "bias": true, | |
| "channel_attention": true, | |
| "channel_consistent_masking": false, | |
| "context_length": 512, | |
| "channel_rope": true, | |
| "d_model": 512, | |
| "distribution_output": "student_t", | |
| "do_mask_input": true, | |
| "dropout": 0.0, | |
| "ff_dropout": 0.0, | |
| "ffn_dim": 512, | |
| "head_dropout": 0.0, | |
| "huber_delta": 1.0, | |
| "init_std": 0.02, | |
| "loss": "mse", | |
| "mask_type": "random", | |
| "mask_value": 0, | |
| "max_wavelength": 500, | |
| "mode": "pretrain", | |
| "model_type": "patchtst", | |
| "norm_eps": 1e-05, | |
| "norm_type": "rmsnorm", | |
| "num_attention_heads": 8, | |
| "num_forecast_mask_patches": 3, | |
| "num_hidden_layers": 8, | |
| "num_input_channels": 1, | |
| "num_parallel_samples": 100, | |
| "num_poly_feats": 256, | |
| "num_rff": 496, | |
| "num_targets": 1, | |
| "output_range": null, | |
| "patch_length": 16, | |
| "patch_stride": 16, | |
| "path_dropout": 0.0, | |
| "poly_feat_degree": 2, | |
| "pooling_type": "max", | |
| "positional_dropout": 0.0, | |
| "positional_encoding_type": "sincos", | |
| "pre_norm": true, | |
| "prediction_length": 128, | |
| "pretrained_encoder_path": null, | |
| "random_mask_ratio": 0.5, | |
| "rff_trainable": true, | |
| "rope_percent": 0.75, | |
| "scaling": "std", | |
| "share_embedding": true, | |
| "share_projection": true, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.40.1", | |
| "unmasked_channel_indices": null, | |
| "use_cls_token": false, | |
| "use_dynamics_embedding": false | |
| } | |