(self, config: T5Config, patch_size)
| 326 | ] |
| 327 | |
| 328 | def __init__(self, config: T5Config, patch_size): |
| 329 | super().__init__(config) |
| 330 | self.model_dim = config.d_model |
| 331 | |
| 332 | self.shared = nn.Embedding(config.vocab_size, config.d_model) |
| 333 | |
| 334 | encoder_config = copy.deepcopy(config) |
| 335 | encoder_config.is_decoder = False |
| 336 | encoder_config.use_cache = False |
| 337 | encoder_config.is_encoder_decoder = False |
| 338 | # self.encoder = T5Stack(encoder_config, self.shared) |
| 339 | self.encoder = JointEncoder(encoder_config, self.shared, patch_size) |
| 340 | decoder_config = copy.deepcopy(config) |
| 341 | decoder_config.is_decoder = True |
| 342 | decoder_config.is_encoder_decoder = False |
| 343 | decoder_config.num_layers = config.num_decoder_layers |
| 344 | self.decoder = T5Stack(decoder_config, self.shared) |
| 345 | |
| 346 | self.lm_head = nn.Linear(config.d_model, config.vocab_size, bias=False) |
| 347 | |
| 348 | # Initialize weights and apply final processing |
| 349 | self.post_init() |
| 350 | |
| 351 | # Model parallel |
| 352 | self.model_parallel = False |
| 353 | self.device_map = None |
| 354 | |
| 355 | def forward( |
| 356 | self, |
nothing calls this directly
no test coverage detected