blog
blog copied to clipboard
Unable to recreate Timit dataset results using wav2vec pretrained model
Hi,
I am unable to recreate the results reported in your blogpost https://huggingface.co/blog/fine-tune-wav2vec2-english on the timit dataset. I just ran the notebook as it is on google-colab, but got very different results. I am attaching a screenshot of the training and validation results. The results are completely different from the ones reported in the blog post. Can u please let m know the issue here? and how we can recreate the results?
Thanks
hi i am facing the same issue as @paulista5 . I am doing the same exercise on custom data , still facing the same WER 1.0 can any you @patrickvonplaten please help out here ?
Hey @paulista5,
Many people have tried to run the notebook and it seemed to work fine - did you change any settings? Also to me it looks like your model is heavily overfitting -> maybe you should add some weight decay and reduce the learning rate
Hi, I am also facing the same issue. The WER is fixed at 1 throughout the training.
Also, the timit dataset seems to have identical samples in your notebook.
I think this happened with an old version of datasets
, could you try updating the datasets library?
can anyone help me out with this?