| 433 | return hidden_states |
| 434 | |
| 435 | class CLIPTextEmbeddings: |
| 436 | def __init__(self): |
| 437 | self.token_embedding = Embedding(49408, 768) |
| 438 | self.position_embedding = Embedding(77, 768) |
| 439 | |
| 440 | def __call__(self, input_ids, position_ids): |
| 441 | return self.token_embedding(input_ids) + self.position_embedding(position_ids) |
| 442 | |
| 443 | class CLIPTextTransformer: |
| 444 | def __init__(self): |