search

zhouzhouyang520 / IAMM

4 stars 0 forks source link

Test results and assessment questions #2

Closed yanlong1211 closed 1 month ago

yanlong1211 commented 2 months ago

Dear Author I am in the process of reproducing my paper and the results obtained from model testing as well as model evaluation are quite different from the results in your paper.

I look forward to your response and hope to have the opportunity to learn more from your work.

zhouzhouyang520 commented 2 months ago

Hello, to facilitate model replication, we've created a new branch containing our trained model. You can use this to reproduce the paper's results. Instructions:

1.Switch to the IAMM_bak branch and download the code. 2.Download the trained model ( Baidu cloud link with Code: pm7b or Google cloud link) and place it in the save/test/ folder. 3.Run the code: sh run.sh

If you plan to train the model yourself, note these factors affecting performance:

1.Environmental differences: Our model achieved 55.9 Acc on Tesla T4 GPU, but may be -2 points lower on GeForce RTX 3090. 2.Iteration numbers: We used 14,399 iterations, which yielded good results. 3.Parameter settings: Carefully check all parameters as they significantly impact the model.