Join us at the Everyday AI Conference in London, New York & Bengaluru! REGISTER NOW

unable to create code env for forecast plugin

atulchavan
Level 2
Level 2
unable to create code env for forecast plugin
 
 
 

Forecast_plugin.PNG

0 Kudos
9 Replies
sergeyd
Dataiker
Dataiker

Hi @atulchavan 

This plugin requires R binary present on the DSS system. Have you run install-r-integration:

https://doc.dataiku.com/dss/latest/installation/r.html?

0 Kudos
atulchavan
Level 2
Level 2
Author

yes i have installed the R, 

please see attached

0 Kudos
sergeyd
Dataiker
Dataiker

Thanks, please try to create the code environment for this plugin (reproduce the issue) and provide backend.log (Administration->Maintenance->Log files) for review. 

0 Kudos
atulchavan
Level 2
Level 2
Author

this is resolved i am able to use forecast plugin..

it would be great if you help with flow , with example to use forecast plugin

0 Kudos
Ignacio_Toledo
0 Kudos
atulchavan
Level 2
Level 2
Author

Thanks it is helpful..
however i  wanted to forecast multiple time series which needs data to be partitioned and this function is not available for trail version

0 Kudos
Reynholds
Level 2

Hi everyone,

I guess I have the same issues, when I'm trying to follow the "Hands-On Tutorial: Forecasting Time Series (Plugin)"

When I'm trying to run the recipe : Train and evaluate forecasting models, i've got the error : 

 

Job failed: Error in Python process: At line 2: <class 'gluonts_forecasts.mxnet_utils.GPUError'>: Error when importing mxnet, please check that you have CUDA {CUDA_VERSION} installed. Detailed error: libquadmath.so.0: cannot open shared object file: No such file or directory

 

Do you have any idea about what to do ?

Thanks by advance 🙂

0 Kudos
tgb417
Neuron
Neuron

@Reynholds 

Looks like you are having cuda install problems. From my point of view this can be challenging. 

When installing Cuda you have to pick the appropriate version depending on if you have a Cuda supported GPU available in the computer running DSS or not.  These are generally NVidea GPUs. If not you have to pick the slower CPU version of the code.    As a test you might try a CPU code environment see if you can get things working.  And then work on the harder to install GPU version of Cuda.

Forcast CPU VS GPU.png

 

You might want to look at these instructions for some help. https://doc.dataiku.com/dss/latest/machine-learning/deep-learning/runtime-gpu.html 

Let us know how you are getting along with your project.

--Tom
sergeyd
Dataiker
Dataiker

Hi @Reynholds 

Looks like you are missing libquadmath (gcc shared support library) OS package so you will need to install it.