1

I am using the h2o4gpu package in R to create a random forest model. When the model is created I cannot for the life of me find a way to save it in order to load it in a future date.

I have tried using the standard way of saving and loading objects in R as presented in the code below.

# Saving the model
save(model, file = MODEL_SAVE_PATH)

# Loading the model for future use
load(file = MODEL_PATH)

When I load the object that is saved and try to predict using that object I get the error message "Cannot predict using a non-function file". Additionally, the very small size of the model saved ( < 10 kb) leads me to believe their is an error as I am expecting a large model file given the size of the dataset.

Frankie Figz
  • 9
  • 4
  • you need to use `h2o.saveModel` and `h2o.loadModel` not the R one http://docs.h2o.ai/h2o/latest-stable/h2o-docs/save-and-load-model.html – s.brunel Jun 21 '19 at 14:37
  • 1
    Hi, this question has been incorrectly marked. In R their is a regular h2o library and then their is the "h2o4gpu" library. The right answer to this question is that you currently CANNOT SAVE A h204gpu model in R. @nicola – Frankie Figz Jul 22 '19 at 01:57
  • I am having the same problem. h2o4gpu object is not h2o object. There is an issue still open on this: https://github.com/h2oai/h2o4gpu/issues/443 – Batuhan Kavlak Apr 20 '20 at 14:29

0 Answers0