Can'not train the model more than 5 minutes of audio in google colab for speech recognition architectures

Question

When i am executing a code for speech recognition in English in google colab can execute more than 100 hours audio, while when i put a dataset for another language it does n