Different training results for every fresh train on the same dataset? #21
Description
Does anyone know why I get different training results for every fresh train on the same dataset?
I have a single pair in the whitelist, I use the same timeframe and same dataset, same weights etc. But every time I train the model and backtest I see drastically different results. One result can be a huge profit, then I delete the model, train again with the same parameters and I get a loss.
Does anyone else notice this? Should I be looking for a problem somewhere, or PyTorch will always train randomly?