skhcjh231
commented
2 months ago Hello, Thank you for your interest in my paper and for trying out the MATR code. The issue that comes to mind immediately might be due to environment version compatibility or differences in the configuration settings compared to my experimental setup. If you could share the versions of your environment or the configuration settings you've used in your experiments, It would help me assist you more effectively.
maojiaqi2324
Hi skhcjh231, First of all, I want to thank you for your outstanding work on this project. The tools and resources you've provided have been incredibly valuable, and your contributions to the community are greatly appreciated. While I was able to successfully reproduce the results from the paper using the provided checkpoint, I have encountered difficulties when attempting to reproduce the same results through my own training process. Despite following the instructions and settings outlined in the repository, the results I obtain are not matching those reported in the paper. I was wondering if there might be any additional details or tips you could provide that might help address this discrepancy? Any advice on troubleshooting or further refining the training process would be greatly appreciated. Thank you again for your fantastic work and for your time in addressing this issue.
Successfully reproduce the results through the provided trained model:
results through my own training process, achieved a peak result of 41.23 at the 25th epoch, but subsequent results fluctuate around 40.5 without further improvement