RNAErnieTokenizer or RNATokenizer ?
In the author's github(https://github.com/CatIIIIIIII/RNAErnie2/blob/main/rnaernie/src/models/rnaernie/tokenization_rnaernie.py), he use 'class RNAErnieTokenizer(RNATokenizer):' as his tokenizer, but in the model card of 'multimolecule/rnaernie', 'from .tokenization_utils import RNATokenizer' is appled to import the tokenizer, which one should i use, RNAErnieTokenizer or RNATokenizer ?
Please use RnaTokenizer in our implementation, or simply calls AutoTokenizer.from_pretrained("multimolecule/rnaernie")
All models in MultiMolecule uses a unified tokenizer, and their weights are processed during weight conversion so that they works with our RnaTokenizer.
Please use
RnaTokenizerin our implementation, or simply callsAutoTokenizer.from_pretrained("multimolecule/rnaernie")All models in MultiMolecule uses a unified tokenizer, and their weights are processed during weight conversion so that they works with our
RnaTokenizer.
Thanks for your repley!
Note that all models (checkpoints) and data provided by MultiMolecule are NOT official release.
While we try our best to ensure our models and data are equivalent to the original implementation, they are not guaranteed.