Conceptually, why can't the model be trained on a new language, assuming that the language has a total of 100 sentences, and we have the sentences of the corresponding sounds, and we want this model to learn these words that are used in these sentences and the model has learned them. Able to read correctly based on the reference sentences that he has been trained to read and produce appropriate output
First, tell me, did you understand what I meant?
I did this with less than 0.08 loss and the result was horribly bad
I do not understand why?
Suppose a language has 3 letters and 3 words
10 sentences containing these words (we have recorded their voices) and we give them to the model for training and it is trained and finally it produces awful sounds.
Conceptually, why can't the model be trained on a new language, assuming that the language has a total of 100 sentences, and we have the sentences of the corresponding sounds, and we want this model to learn these words that are used in these sentences and the model has learned them. Able to read correctly based on the reference sentences that he has been trained to read and produce appropriate output First, tell me, did you understand what I meant? I did this with less than 0.08 loss and the result was horribly bad I do not understand why?
Suppose a language has 3 letters and 3 words 10 sentences containing these words (we have recorded their voices) and we give them to the model for training and it is trained and finally it produces awful sounds.
guide my dear friend