I am training Informer on a private dataset based on the pipeline MAZiqing/FEDformer (github.com). I have fixed the random seed and the num_workers is set to 1. The problem is, I get same results on different training using single 3090 or 3090ti gpu, even on different devices(Ubuntu on clound server or Ubuntu(WSL) on local PC). But I get different results on different training using single 4090 gpu.
We conduct the training both on Ubuntu clound server and local PC(Ubuntu, WSL). All the code in training is same.
Is there any solutions to fixed this problem?
Did you set all needed variables as explained in the Reproducibility docs? If so, could you post a minimal and executable code snippet which would show this behavior, please?