Since the weights are usually element-wise multiply to the observed likelihood, I dont see a straightforward way to do it with GLM object. You can try either:
- Writing a custom distribution to express the weighted likelihood
- Express it as an additional term and use pm.Potential to add it to the model.