(self, ids, mask=None)
| 243 | self.apply(init_weights) |
| 244 | |
| 245 | def forward(self, ids, mask=None): |
| 246 | x = self.token_embedding(ids) |
| 247 | x = self.dropout(x) |
| 248 | e = self.pos_embedding(x.size(1), |
| 249 | x.size(1)) if self.shared_pos else None |
| 250 | for block in self.blocks: |
| 251 | x = block(x, mask, pos_bias=e) |
| 252 | x = self.norm(x) |
| 253 | x = self.dropout(x) |
| 254 | return x |
| 255 | |
| 256 | @staticmethod |
| 257 | def state_dict_converter(): |