search

piotrkawa / deepfake-whisper-features

Implementation of the paper "Improved DeepFake Detection Using Whisper Features"
MIT License
72 stars 4 forks source link

Error running the code #6

Open andysingal opened 8 months ago

andysingal commented 8 months ago

Greetings, Thanks for sharing the amazing git repo. Few questions: -!python ./deepfake-whisper-features/train_models.py --asv_path /content/deepfake-whisper-features/src/datasets/asvspoof_dataset.py --in_the_wild_path /content/deepfake-whisper-features/src/datasets/in_the_wild_dataset.py --config /content/deepfake-whisper-features/configs/training/whisper_specrnet.yaml --batch_size 8 --epochs 10 --train_amount 100000 --test_amount 25000 gives error:

/usr/local/lib/python3.10/dist-packages/torchaudio/functional/functional.py:584: UserWarning: At least one mel filterbank has all zero values. The value for `n_mels` (128) may be set too high. Or, the value for `n_freqs` (257) may be set too low.
  warnings.warn(
usage: train_models.py [-h] [--asv_path ASV_PATH] [--config CONFIG] [--train_amount TRAIN_AMOUNT]
                       [--test_amount TEST_AMOUNT] [--batch_size BATCH_SIZE] [--epochs EPOCHS]
                       [--ckpt CKPT] [--cpu]
train_models.py: error: unrecognized arguments: --in_the_wild_path /content/deepfake-whisper-features/src/datasets/in_the_wild_dataset.py

further love to see some kaggle examples or huggngface dataset in the example folder. Looking forward to hearing from you. Best, Andy

piotrkawa commented 8 months ago

Hi! That is indeed a mistake in the README file - this parameter is unnecessary in train_model.py as in our paper, we only use the In-The-Wild dataset for testing. Thanks for pointing that out! I'll fix it :)

BTW, the train_and_test.py script contains both training and test steps - using it is much more convenient to run experiments. :)

Best, Piotr