(self,
builder,
feature_column,
embedding_table,
get_seq_len=False)
| 229 | return logits |
| 230 | |
| 231 | def _get_embedding_input(self, |
| 232 | builder, |
| 233 | feature_column, |
| 234 | embedding_table, |
| 235 | get_seq_len=False): |
| 236 | sparse_tensors = feature_column._get_sparse_tensors(builder) |
| 237 | sparse_tensors_ids = sparse_tensors.id_tensor |
| 238 | sparse_tensors_weights = sparse_tensors.weight_tensor |
| 239 | |
| 240 | if self._emb_fusion and not self.tf: |
| 241 | from tensorflow.python.ops.embedding_ops import fused_safe_embedding_lookup_sparse |
| 242 | embedding = fused_safe_embedding_lookup_sparse( |
| 243 | embedding_weights=embedding_table, |
| 244 | sparse_ids=sparse_tensors_ids, |
| 245 | sparse_weights=sparse_tensors_weights) |
| 246 | else: |
| 247 | embedding = tf.nn.safe_embedding_lookup_sparse( |
| 248 | embedding_weights=embedding_table, |
| 249 | sparse_ids=sparse_tensors_ids, |
| 250 | sparse_weights=sparse_tensors_weights) |
| 251 | |
| 252 | if get_seq_len: |
| 253 | sequence_length = fc_utils.sequence_length_from_sparse_tensor( |
| 254 | sparse_tensors_ids) |
| 255 | return embedding, sequence_length |
| 256 | else: |
| 257 | return embedding |
| 258 | |
| 259 | def _embedding_input_layer(self): |
| 260 | for key in SEQ_COLUMNS: |
no test coverage detected