My problem looks like a fastai v2 problem.
I just added the callback GradientAccumulation() in the notebook 10_nlp.ipynb: at the beginning of the epoch, the training loss is 18.978.176 instead of… 4 (bs = 128, n_acc = 256).
My problem looks like a fastai v2 problem.
I just added the callback GradientAccumulation() in the notebook 10_nlp.ipynb: at the beginning of the epoch, the training loss is 18.978.176 instead of… 4 (bs = 128, n_acc = 256).
