I appreciate quite a few people have asked similar questions, but I haven’t found a solution that works for me.
I am running a vanilla sequence classifier on colab and each time I run it I get different results - even if I don’t restart my runtime.
I have specified seed and data_seed in my training arguments (though it looks like they already have default parameters) and I have tried the suggestion here: Different BERT results. However neither of these work and my training run is different each time.
My data is created (and split into train, dev, test) so there are no changes in that each time.
Any suggestions would be super helpful!