| from transformers import PretrainedConfig | |
| class ActEstimatorConfig(PretrainedConfig): | |
| model_type = "ACT-Estimator" | |
| def __init__( | |
| self, | |
| input_shape=(3, 44, 224, 224), | |
| num_classes=9, | |
| max_seq_len=44, | |
| timestamp_dim=1, | |
| d_model=512, | |
| num_heads=8, | |
| dropout=0.1, | |
| feature_map_size=4, | |
| **kwargs | |
| ): | |
| self.input_shape = input_shape | |
| self.num_classes = num_classes | |
| self.max_seq_len = max_seq_len | |
| self.timestamp_dim = timestamp_dim | |
| self.d_model = d_model | |
| self.num_heads = num_heads | |
| self.dropout = dropout | |
| self.feature_map_size = feature_map_size | |
| super().__init__(**kwargs) | |