The neuro-comma from marlon-br

Colab to prepare data for any language: https://colab.research.google.com/drive/1-EUY-Do3vR_h3htyjjAMR-zVQyUsaH-t?usp=sharing
Just set right language in LANG1 and LANG2 parameters. Also you have to set right UTF-8 characters for you language: https://www.periodni.com/unicode_utf-8_encoding.html.

Training Datasets in neuro-comma format:

Dataset language	Google Drive Link	Zipped Size
Portuguese	https://drive.google.com/file/d/1-0t7sskF3bKGeXiCOx4nL9R53-Z0f7ak/view?usp=sharing	358.2 MB
Spanish	https://drive.google.com/file/d/1JHATFSu4amgz-cHcKXsSUyUs9T70kPa7/view?usp=sharing	476.7 MB
German	https://drive.google.com/file/d/1Us6mMtPjOxUi1Mu_M1PHWTB9pxuwyod7/view?usp=sharing	721.9 MB
Russian	https://drive.google.com/file/d/1-0MBsnGfId6AGxgPhnkO2gAxaDn1DWGQ/view?usp=sharing	584.7 MB
French	https://drive.google.com/file/d/1--ocXMVscsIKghB9xzEUUeIVvJqY0qUX/view?usp=sharing	971.8 MB

Neuro-comma (puntuation-restoration) format: Oh, nun ja, dann wird sie es wohl gewesen sein. Wo sind die Karten geblieben?

turns into:
oh COMMA
nun O
ja COMMA
dann O
wird O
sie O
es O
wohl O
gewesen O
sein PERIOD
wo O
sind O
die O
karten O
geblieben QUESTION

Pretrained models:

Model language	Base model	Google Drive Link
Portuguese (quantized)	xlm-roberta-large	https://drive.google.com/file/d/174kc3436Vck9jBYbkfr5gwY6bn5wki14/view?usp=sharing
Spanish (quantized)	xlm-roberta-large	https://drive.google.com/file/d/1181DFuVoYIEiTqaAnNAVSQKe9n69xB49/view?usp=sharing
German (qunatized	xlm-roberta-large	https://drive.google.com/file/d/1Tw4ISUVMPqMXIwEU0IFy5ldSLa9z7Zjt/view?usp=sharing
Russian (quantized)	xlm-roberta-large	https://drive.google.com/file/d/1VgmF4wHbpftyWxZQrY7BJHOYdKYjZSBH/view?usp=sharing
French	xlm-roberta-large	https://drive.google.com/file/d/1wg6GeRi5EKE_60bni3PWsDQM_IdpXIn_/view?usp=sharing
English (from punctuation-restoration)	roberta-large	https://drive.google.com/file/d/17BPcnHVhpQlsOTC8LEayIFFJ7WkL00cr/view?usp=sharing

How it works