Currently, our experiments log a lot of data, considering the amount of metrics and size of the CICIDS dataset. We most likely won't need such high fidelity results, so we should consider reducing that amount for better performance.
The simplest idea would be making the call to wandb.init once every N steps.
Currently, our experiments log a lot of data, considering the amount of metrics and size of the CICIDS dataset. We most likely won't need such high fidelity results, so we should consider reducing that amount for better performance.
The simplest idea would be making the call to
wandb.initonce every N steps.