We can provide parameters during model loading, such as "vocab_only" and "tensor_only". Loading only the vocabulary can be useful in cases where only the tokenizer is needed, while loading only tensors would be suitable for certain visual models.
I'm open to hearing any suggestions or advice regarding whether it would be better to implement a separate TokenizerModel or make modifications within the existing Model. Any suggestions would be greatly appreciated.
We can provide parameters during model loading, such as "vocab_only" and "tensor_only". Loading only the vocabulary can be useful in cases where only the tokenizer is needed, while loading only tensors would be suitable for certain visual models.
I'm open to hearing any suggestions or advice regarding whether it would be better to implement a separate TokenizerModel or make modifications within the existing Model. Any suggestions would be greatly appreciated.