Seeding and reproducibility

The result of gym-dssat are reproducible on the same platform, i.e. same software and hardware layers. You can specify the seed with seed argument at environment’s Initialization. The seed fixes as well the internal stochastic weather generation.

Caution

We cannot guarantee cross-platform reproducibility due to the multiple factors at stake.

If you want to fix or change the seed of an already initialized gym-dssat instance named env, you can use:

seed = 123
env.set_seed(seed)