Is there a way to use parglm for prediction?

[SantiagoEsteban]

I'm using parglm on a >40M dataset using 10 cores. It trains perfectly but when using it to predict on a nwe dataset and type = 'response', I get the following error: Caused by error in qr.R():! argument is not a QR decomposition

I understand this is because the parglm package stores the model components differently than standard glm. Is there a way to use it for prediction?

Thanks