Method tokenizer

models/audiosep/models/clap_encoder.py:108–116

(self, text)

        return embed.float()

    def tokenizer(self, text):
        result = self.tokenize(
            text,
            padding="max_length",
            truncation=True,
            max_length=512,
            return_tensors="pt",
        )
        return {k: v.squeeze(0) for k, v in result.items()}