(
self,
extract_features,
attention_mask=None,
mask_time_indices=None,
output_attentions=None,
output_hidden_states=None,
return_dict=None,
)
| 76 | return extract_features |
| 77 | |
| 78 | def encode( |
| 79 | self, |
| 80 | extract_features, |
| 81 | attention_mask=None, |
| 82 | mask_time_indices=None, |
| 83 | output_attentions=None, |
| 84 | output_hidden_states=None, |
| 85 | return_dict=None, |
| 86 | ): |
| 87 | self.config.output_attentions = True |
| 88 | |
| 89 | output_hidden_states = ( |
| 90 | output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states |
| 91 | ) |
| 92 | return_dict = return_dict if return_dict is not None else self.config.use_return_dict |
| 93 | |
| 94 | if attention_mask is not None: |
| 95 | # compute reduced attention_mask corresponding to feature vectors |
| 96 | attention_mask = self._get_feature_vector_attention_mask( |
| 97 | extract_features.shape[1], attention_mask, add_adapter=False |
| 98 | ) |
| 99 | |
| 100 | |
| 101 | hidden_states, extract_features = self.feature_projection(extract_features) |
| 102 | hidden_states = self._mask_hidden_states( |
| 103 | hidden_states, mask_time_indices=mask_time_indices, attention_mask=attention_mask |
| 104 | ) |
| 105 | |
| 106 | encoder_outputs = self.encoder( |
| 107 | hidden_states, |
| 108 | attention_mask=attention_mask, |
| 109 | output_attentions=output_attentions, |
| 110 | output_hidden_states=output_hidden_states, |
| 111 | return_dict=return_dict, |
| 112 | ) |
| 113 | |
| 114 | hidden_states = encoder_outputs[0] |
| 115 | |
| 116 | if self.adapter is not None: |
| 117 | hidden_states = self.adapter(hidden_states) |
| 118 | |
| 119 | if not return_dict: |
| 120 | return (hidden_states, ) + encoder_outputs[1:] |
| 121 | return BaseModelOutput( |
| 122 | last_hidden_state=hidden_states, |
| 123 | hidden_states=encoder_outputs.hidden_states, |
| 124 | attentions=encoder_outputs.attentions, |
| 125 | ) |
nothing calls this directly
no outgoing calls
no test coverage detected