Mateusz Pieniak

@BigBadBurrow What batch size & float precision did you end up on Titan X (12GB)?

``` from transformers import BertTokenizer, BertForTokenClassification tokenizer = BertTokenizer.from_pretrained('allenai/scibert_scivocab_uncased') model = BertForTokenClassification.from_pretrained('allenai/scibert_scivocab_uncased') nlp = pipeline('ner', model=model, tokenizer=tokenizer) text = 'Clinical features of culture-proven Mycoplasma pneumoniae infections at King Abdulaziz University...

Thank you for the workaround!

About the multi-label approach, equation 26 already gives you a scalar, so you don't need to pass it to any NLL function. Just remember to scale your `q` (multi-label target)...

About the second bug from cost calculator, it is because `prompt_tokens` in the response from Voyage API is `None`. > CreateEmbeddingResponse(data=[Embedding(embedding=[], index=0, object='embedding')], model='voyage-code-2', object='list', usage=Usage(prompt_tokens=None, total_tokens=7))