Johnson yue
Johnson yue
> You should use higher value for harmonization_start_scale. You can play with this parameter until you'll get the desired result. > […](#) > On Fri, Nov 22, 2019, 04:26 Johnson...
> This is the result I get with start_scale=6: >  > Did you change anything within the function? No, I do not change anything , Maybe I should try...
@GLivshits which type of dataset ? human face ?
@oliverguhr Hi, if you have some better result please share here, and Which version of code did you try ? only 0.4.23 ? Do you try with newst version ??...
@oliverguhr good result！！
Hi, I think the code for testing part have some bugs such as : 1. not set **num_step** variable in Line 298 2. not used **one_hot** variable in Line 301...
where is source code for DAN ？ did you reference here ：https://github.com/Wang-Tianwei/Decoupled-attention-network you don‘t have this parameter [self.char_embeddings](https://github.com/Wang-Tianwei/Decoupled-attention-network/blob/b4d8b9e32531246dd74215bd0cde78bd8032c5e2/DAN.py#L180) in training， maybe I have to training again。。。
this bug is not fixed for now ？？？？