It seems newer versions of the
transformers library expect a
tokenizer_config.json file. I wonder if the model https://huggingface.co/KBLab/wav2vec2-large-voxrex-swedish/ uses the same
tokenizer_config.json, and if so, if the file there could be copied here.
Just realizing this is a pure acoustic model without any CTC on top, so it does not make sense to have a tokenizer I guess.
Correct, this model (
KBLab/wav2vec2-large-voxrex) is provided so that people can continue pretraining on the acoustic model, or alternatively do their own finetuning on any downstream task they might be interested in.