Closed moniquebm closed 3 years ago
In general, BERT-based models have a sequence maximum length = 512, but when I check:
tokenizer = AutoTokenizer.from_pretrained('neuralmind/bert-base-portuguese-cased') tokenizer.max_len is equal to 1000000000000000019884624838656
It seems you don't have maximum sequence limits. So, what's the right value to consider for max_len?
Hi @moniquebm You're right, it should be 512.
In general, BERT-based models have a sequence maximum length = 512, but when I check:
tokenizer = AutoTokenizer.from_pretrained('neuralmind/bert-base-portuguese-cased') tokenizer.max_len is equal to 1000000000000000019884624838656
It seems you don't have maximum sequence limits. So, what's the right value to consider for max_len?