Issues with pretrained model `RawNet3/model.pt`

[SangwonSUH]

Hi,

• I'm trying to reproduce the evaluation EER of the provided pretrained models with the instructions from the README file.
• It was successful with the model ResNetSE34V2, but RawNet3 has some issues

Issues

1. There was no config file named configs/RawNet3.yaml
   • Is it okay to use the RawNet3_AAM.yaml instead?
2. Errors from loadParameters

   Traceback (most recent call last):
     File "./voxceleb_trainer/./trainSpeakerNet.py", line 276, in <module>
       main()
     File "./voxceleb_trainer/./trainSpeakerNet.py", line 272, in main
       main_worker(0, None, args)
     File "./voxceleb_trainer/./trainSpeakerNet.py", line 165, in main_worker
       trainer.loadParameters(args.initial_model)
     File "./voxceleb_trainer/SpeakerNet.py", line 248, in loadParameters
       loaded_state = torch.load(path, map_location="cuda:%d" % self.gpu)
     File "/home/miniconda3/envs/torch/lib/python3.9/site-packages/torch/serialization.py", line 608, in load
       return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
     File "/home/miniconda3/envs/torch/lib/python3.9/site-packages/torch/serialization.py", line 777, in _legacy_load
       magic_number = pickle_module.load(f, **pickle_load_args)
   _pickle.UnpicklingError: invalid load key, 'v'.

Tests

I've tried the trainSpeakerNet.py script with followings

• Python env.

  python                    3.9.9           h543edf9_0_cpython    conda-forge
  pytorch                   1.10.1          py3.9_cuda11.3_cudnn8.2.0_0    pytorch
  torchaudio                0.10.1          py39_cu113    pytorch

• Python cmd

  CUDA_VISIBLE_DEVICES=0 python ./trainSpeakerNet.py --eval --config ./configs/RawNet3_AAM.yaml --initial_model RawNet3/model.pt

---

Thanks for your efforts in maintaining this repository.

[SangwonSUH]

Sorry, it was git-lfs issue Pull RawNet3/model.pt with following cmd

# ./voceleb_trainer/RawNet3
git lfs pull