INNER CODE UNIT · Python

start_tokens

lucidrains/x-transformers · train_copy.py:67

        start_tokens = (torch.ones((1, 1)) * 1).long().to(DEVICE)

        sample = model.generate(src, start_tokens, ENC_SEQ_LEN, mask = src_mask)
        incorrects = (src != sample).long().abs().sum()

        print(f"input:  ", src)
        print(f"predicted output:  ", sample)
        print(f"incorrects: {incorrects}")

View source record →