huangqingyi-code

the file of the tpot.fitted_pipeline_ is very large

1

Hi tpot team, Now,I can save the tpot.fitted_pipeline_ and load it,but the file of the tpot.fitted_pipeline_ is very large.it is relate to training data,the more data, the bigger the tpot.fitted_pipeline_.My...

你好，tokenizer.tokenize("")时为什么会多出来一个'▁'符号。tokenizer.encode导致时多了一个31106id

markdown格式的数据预训练

3

你好，我看你们v1.5版本的模型加入markdown格式的数据，在做预训练的时候，比如表格markdown的数据是不是在计算损失的时候是不是应该mask掉