This is the IndicBART model. For detailed documentation look here: https://indicnlp.ai4bharat.org/indic-bart/ and https://github.com/AI4Bharat/indic-bart/ Usage: ``` from transformers import MBartForConditionalGeneration, AutoModelForSeq2SeqLM from transformers import AlbertTokenizer, AutoTokenizer tokenizer = AutoTokenizer.from_pretrained("prajdabre/CreoleM2M", do_lower_case=False, use_fast=False, keep_accents=True) # Or use tokenizer = AlbertTokenizer.from_pretrained("prajdabre/CreoleM2M", do_lower_case=False, use_fast=False, keep_accents=True) model = AutoModelForSeq2SeqLM.from_pretrained("prajdabre/CreoleM2M") # Or use model = MBartForConditionalGeneration.from_pretrained("prajdabre/CreoleM2M") # Some initial mapping bos_id = tokenizer._convert_token_to_id_with_added_voc("") eos_id = tokenizer._convert_token_to_id_with_added_voc("") pad_id = tokenizer._convert_token_to_id_with_added_voc("") # To get lang_id use any of ["", "", "<2acf>", "<2eng>", "<2bis>", "<2bzj>", "<2cbk>", "<2crs>", "<2djk>", "<2gul>", "<2hat>", "<2hwc>", "<2icr>", "<2jam>", "<2kri>", "<2ktu>", "<2mbf>", "<2mfe>", "<2mkn>", "<2pap>", "<2pcm>", "<2pis>", "<2rop>", "<2sag>", "<2srm>", "<2srn>", "<2tcs>", "<2tdt>", "<2tpi>"] # First tokenize the input and outputs. The format below is how CreoleM2M was trained so the input should be "Sentence <2xxx>" where xxx is the language code. Similarly, the output should be "<2yyy> Sentence ". inp = tokenizer('Wen dey wen stretch him out fo whip him real hard , Paul wen tell da captain dat stay dea , “ Dis okay in da rules fo da Rome peopo ? fo you fo whip one guy dat get da same rights jalike da Rome peopo ? even one guy dat neva do notting wrong ? ' <2hwc>", add_special_tokens=False, return_tensors="pt", padding=True).input_ids model.eval() # Set dropouts to zero model_output=model.generate(inp, use_cache=True, num_beams=4, max_length=60, min_length=1, early_stopping=True, pad_token_id=pad_id, bos_token_id=bos_id, eos_token_id=eos_id, decoder_start_token_id=tokenizer._convert_token_to_id_with_added_voc("")) # Decode to get output strings decoded_output=tokenizer.decode(model_output[0], skip_special_tokens=True, clean_up_tokenization_spaces=False) print(decoded_output) Notes: 1. This is compatible with the latest version of transformers but was developed with version 4.3.2 so consider using 4.3.2 if possible. 2. While I have only shown how to let logits and loss and how to generate outputs, you can do pretty much everything the MBartForConditionalGeneration class can do as in https://huggingface.co/docs/transformers/model_doc/mbart#transformers.MBartForConditionalGeneration 3. Note that the tokenizer I have used is based on sentencepiece and not BPE. Therefore I use the AlbertTokenizer class and not the MBartTokenizer class.