HyperMoon

This model is a fine-tuned version of facebook/wav2vec2-base-960h on the asvspoof2019 dataset. It achieves the following results on the evaluation set: - Loss: 0.0009 - Accuracy: 0.9998 The following hyperparameters were used during training: - learningrate: 3e-05 - trainbatchsize: 8 - evalbatchsize: 8 - seed: 42 - gradientaccumulationsteps: 4 - totaltrainbatchsize: 32 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lrschedulertype: linear - lrschedulerwarmupratio: 0.1 - numepochs: 5 | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 0.0027 | 1.0 | 793 | 0.0070 | 0.9990 | | 0.0006 | 2.0 | 1586 | 0.0032 | 0.9995 | | 0.0271 | 3.0 | 2379 | 0.0022 | 0.9995 | | 0.0002 | 4.0 | 3172 | 0.0009 | 0.9998 | | 0.0002 | 5.0 | 3965 | 0.0020 | 0.9998 | - Transformers 4.22.1 - Pytorch 1.12.1+cu113 - Datasets 2.4.0 - Tokenizers 0.12.1