Dear Pyro Team,
I hope this message finds you well. I have recently learned about the benefits of Stochastic Weight Averaging (SWA) in improving the performance of well-tuned models across various practical applications. I noticed that PyTorch 1.6 introduced the torch.optim.swa_utils
module, which provides a convenient way to implement SWA.
I am currently working with Pyro and I am curious to know if there is an equivalent or similar functionality available in Pyro to easily implement SWA. Are there any custom functions or modules specifically designed for implementing SWA in Pyro? I would greatly appreciate any guidance or advice you can provide in this regard.
Thank you for your time and assistance.
Best regards,
Fred