kosugi11037 / bert-int

46 stars 20 forks source link

Results aren't matching the paper results #6

Closed KartavyaKothari closed 4 years ago

KartavyaKothari commented 4 years ago

Hi,

I trained the model from scratch keeping all parameters same. I am getting hits@1 to be ~0.87. What kind of changes would be required to get the results mentioned in your paper.

KartavyaKothari commented 4 years ago

I misinterpreted the mbert finetuning to this, closing issue