Questions about reproducing the process of the experiment

[Nora-jsu]

Hello, I am reproducing your experiment recently, but I can't reproduce your effect in the report, the highest F-score is only 9, the following is my parameter configuration, please ask what needs to be changed in the parameter configuration of the large model, I look forward to your reply. Here is my parameter interface:

[axeber01]

Can you check if the model weights are loaded correctly? Try going to this line and use 'strict=True' instead https://github.com/axeber01/ngcc-seld/blob/64aa259165b7d57a773b1e2c5e528258bf9f28ea/train_seldnet.py#L734

Then set the learning rate to 0 (so that the weights are not changed after loading them) and the number of epochs to 1 (so that the evaluation is done faster). Then you should get the expected results.