patrickvonplaten/wav2vec2-base-100h-13K-steps

Model was trained on librispeech-clean-train.100 with following hyper-parameters: "clean" and "other" The model seems to have overfitting to the "Clean" data . The result on other data is significantly worse - the model seems overfit to the 'clean' data . Results are very similar to the ones of the official model, but the result on 'other' data is much worse . The model was trained using 32-GPUs with a total batch size of ca. ~1500 seconds with linear decaying learning rate with 3000 warmup steps . The data was not used during training at the time of the training and the model was not given a mask,