Yuhao Lin

Results 3 comments of Yuhao Lin

you need to convert the dataset into LMDB file first. trainroot is the location of this lmdb file. See the create_dataset.py to build the lmdb file

can you share your repo or maybe the hyper-param settings ? I tried BERT+FC and BERT+CNN, both ended up about 84% on the test set.

> Actually I found out that using the hyperparameters provided by the doc can get 93% acc. I didn't find such doc. Can you share with me the link ?...