-
Notifications
You must be signed in to change notification settings - Fork 25
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Retrained model does not get the same SPL on val unseen as reported in paper #14
Comments
Sorry for the late reply. The original code with reproducible code/results is provided in another issue: #11. Given the results in xlsx, the 2% drop in SPL (46%) is possibly caused by the drop in SR, which is still much higher compared to previous SotA (38%). The reason I currently find is some implementation differences inside the speaker when I cleaned the code (the original, reproducible code is provided in the other issue). Since the beam-search results which only rely on the inference of the speaker also changed. I haven't located which differences cause this issue. All the differences seem not to affect the training/inference process but the predictions are actually changed. Please kindly check the original code before I find it. Best, |
Thanks for your reply! |
please help me ! After I train the model, i use the test environment to evaulate,the success rate result is below, i dont understand why the result is so low? please help me, is there something wrong when i test ? |
Hi @jingquanliang , I didn't see your result, have you fixed it? Or you can try this script. name=agent_bt CUDA_VISIBLE_DEVICES=$1 python r2r_src/train.py $flag --name $name |
Hi,
We are trying to retrain the EnvDrop model based on this repo, but the results are not same as reported in paper, we have tried different PyTorch versions, our best result with PyTorch 0.4.1 is 0.46, which is less than the reported 48% on val unseen dataset in terms of SPL, for detailed results you can refer to the attachment below.
Have we missed something important? or can you specify your working environment?
Our retrained model:
retrained_envdrop_results.xlsx
Results in paper:
The text was updated successfully, but these errors were encountered: