CODEWITHSUNDEEP
rh2o
Weiye Deng
Weiye Deng

Reputation: 65

model loading error due to version conflict R

I updated my h2o to the newest version and then tried to load the pre-trained model by typing:

randomforest = h2o.loadModel('randomforest')

However, it shows:

 Error in .h2o.doSafeREST(h2oRestApiVersion = h2oRestApiVersion, urlSuffix = page,  : 
 ERROR MESSAGE:
 Found version 3.10.5.3, but running version 3.14.0.7

Does that mean I need to retrained all the model I built before ? That's extremely inconvenient.

Upvotes: 1

Views: 2392

Answers (2)

MNK
MNK

Reputation: 664

You can simply reinstall the old version. In anaconda, for example,

conda install -c h2oai h2o=3.14.0.7

Worked for me!

Upvotes: 0

Lauren
Lauren

Reputation: 5778

Yes, you will need to re-train the models using the same version of H2O (the version you train a model needs to be the same as the version you load the model with). H2O binary models are not compatible across major versions.

For binary models, this is the standard practice -- you will have the same situation if you use scikit-learn, for example.

If you want to use MOJO/POJO models in production, those are not tied to a particular version of H2O since they are just plain Java code and do not require the H2O cluster to be running.

Upvotes: 5

Related Questions