I have been able to use speechbrain to train a couple of models with no problems on Colab using GPU for a few months until today… Usually, when my Colab runtime reaches the time limit, I stop the training and restart the runtime. I did not make any changes to the code. But for some reason, I am getting the following error today:
OSError: libcudart.so.10.2: cannot open shared object file: No such file or directory
Looks like it has something to do with cuda. On Colab, every time I initiate a new run time, I am re-installing speechbrain. Is it possible that there is some mismatch between some updates in speechbrain and the GPU I get assigned on Colab? The full error message is also attached here. Thank you!