Questions About Your Training Dataset and Model Performance
#2
by nmcuong - opened
Thank you for sharing. May I ask if you trained on a large amount of data? What is the quality of your dataset like (the minimum and maximum duration of audio in the dataset)? How long did you train the model? Do you find that the 7B model performs significantly better than the 1.5B model? Does it have issues such as adding extra words, repeating phrases, or generating music or strange sounds? Thank you.