BA_Cycle & BA_DDG with the same res

[ChangShaole]

Hi, I have tried to run your examples on the skempi dataset, but I got the same results for both the BA Cycle and BA DDG. Does anyone know what the problem may be?

[MogicianHuang]

I got the same result, too. The pretrain result is better and close to the result reported in the paper after I loaded the ckpt trained by the train code.

[kamzero]

Hi, I have tried to run your examples on the skempi dataset, but I got the same results for both the BA Cycle and BA DDG. Does anyone know what the problem may be?

Have you downloaded and loaded the checkpoint file (ckpt) as instructed?
https://drive.google.com/drive/folders/1We-D7XaixxWgWVJzF7suYpoaeaz1cV9w

[MogicianHuang]

Hi, I have tried to run your examples on the skempi dataset, but I got the same results for both the BA Cycle and BA DDG. Does anyone know what the problem may be?

Have you downloaded and loaded the checkpoint file (ckpt) as instructed? https://drive.google.com/drive/folders/1We-D7XaixxWgWVJzF7suYpoaeaz1cV9w

Yes, I downloaded and put it in the /ckpt as instructed but the result were the same.

[kamzero]

@MogicianHuang @ChangShaole
Thanks for catching that! We uploaded the wrong model by mistake. The updated BA-DDG checkpoint is in the original link.