INNER CODE UNIT · Python

encode_text

patrickjohncyh/fashion-clip · fashion_clip/fashion_clip.py:200

    def encode_text(self, text: List[str], batch_size: int):
        dataset = Dataset.from_dict({'text': text})
        dataset = dataset.map(lambda el: self.preprocess(text=el['text'], return_tensors="pt",
                                                         max_length=77, padding="max_length", truncation=True),
                              batched=True,
                              remove_columns=['text'])
        dataset.set_format('torch')
        dataloader = DataLoader(dataset, batch_size=batch_size)
        text_embeddings = []
        pbar = tqdm(total=len(text) // batch_size, position=0)
        with torch.no_grad():
            for batch in dataloader:
                batch = {k: v.to(self.device) for k, v in batch.items()}
                text_embeddings.extend(self.model.get_text_features(**batch).detach().cpu().numpy())
                pbar.update(1)
            pbar.close()
        return np.stack(text_embeddings)

View source record →