Was not able to exactly reproduce the numbers from the paper

[kamabdi]

Hello Kuang-Huei Lee, thank you for realizing the code. I think your method is quite interesting and I want to use it to propose some additional improvements. I ran the code with specified parameters, however was not able to reproduce the exact numbers from the paper. I would really appreciate if you can realize the per-trained models along with the code.

[kuanghuei]

Hi, many people were able to reproduce similar results, either in their published works, feedback in this repo, or privately informed me. I myself also cloned the code from this repo and run it multiple times, and were also able to reproduce the results.

I think either your training or evaluation setting may has problem. I would need to see the results of your experiments, commands, and data to help you debug.