can Rapidminer use GPU for student license?
Given that the corresponding CUDA version is installed and working and tested in KERAS within Anaconda, will Rapidminer use GPU for student license? Is there a version of CUDA that must be used? I'm using RapidMiner 9.1 Educational Edition.
The Error message is:
"There was an error while switching to the GPU backend."
followed by:
"Error while switching to GPU backend. Either CUDA 9.0 is not installed or you have a free license. Check the log for more information."
I'm new to setting up KERAS for Deep Learning in RapidMiner.
Thanks!
The Error message is:
"There was an error while switching to the GPU backend."
followed by:
"Error while switching to GPU backend. Either CUDA 9.0 is not installed or you have a free license. Check the log for more information."
I'm new to setting up KERAS for Deep Learning in RapidMiner.
Thanks!
Find more posts tagged with
Sort by:
1 - 8 of
81
Hi Varun,
Thank you for at least letting me know that is will work with CUDA 9. So I did not mention that I am using RTX I believe. For the new cards that can utilise Tennsor units, it can use CUDA 9 (Volta and Tensor core support). The graphics driver version that supports it, however, is installing CUDA 10. There were ways to install only the cuda 9 toolkit and some users have done that. Installing the Tensorflow and Keras in Anaconda was successful and automatically install the CUDA 9.0 toolkit and I was able to run the samples from Keras. Still I think that was a problem when using it with Rapidminer. Currently the updates may may broken the version compatability....
Before RapidMiner ask for CUDA 9.0, it asked for CUDA 9.1, Does that mean something?
Thanks,
Ken
Thank you for at least letting me know that is will work with CUDA 9. So I did not mention that I am using RTX I believe. For the new cards that can utilise Tennsor units, it can use CUDA 9 (Volta and Tensor core support). The graphics driver version that supports it, however, is installing CUDA 10. There were ways to install only the cuda 9 toolkit and some users have done that. Installing the Tensorflow and Keras in Anaconda was successful and automatically install the CUDA 9.0 toolkit and I was able to run the samples from Keras. Still I think that was a problem when using it with Rapidminer. Currently the updates may may broken the version compatability....
Before RapidMiner ask for CUDA 9.0, it asked for CUDA 9.1, Does that mean something?
Thanks,
Ken
Hi all,
yes the dependency on CUDA 9.0 is correct.We changed that for compatibility reasons with the DL4J back end.
But having several CUDA toolkit instances installed should not be a problem as long as each application can find its required version.
Best,
David
yes the dependency on CUDA 9.0 is correct.We changed that for compatibility reasons with the DL4J back end.
But having several CUDA toolkit instances installed should not be a problem as long as each application can find its required version.
Best,
David
Hi @18a637y
As David mentioned you can have multiple versions. I used keras in rapidminer before but there are lot of issues with it. The same configuration used to work in python code(keras & TF) but not in RM due to issues with the extension. I got to know from RM team about updated Deep learning extension which uses DL4J instead of tensorflow, this extensions works well.
You can look at our discussion here.
https://community.rapidminer.com/discussion/comment/55180#Comment_55180
Regards,
Varun
As David mentioned you can have multiple versions. I used keras in rapidminer before but there are lot of issues with it. The same configuration used to work in python code(keras & TF) but not in RM due to issues with the extension. I got to know from RM team about updated Deep learning extension which uses DL4J instead of tensorflow, this extensions works well.
You can look at our discussion here.
https://community.rapidminer.com/discussion/comment/55180#Comment_55180
Regards,
Varun
I have to dig out this old thread again.
Using an older CUDA version with an updated driver should also work? I'm having trouble to switch to GPU with a GTX 970, driver 430.00 which comes with CUDA driver 10.1.120.
The CUDA 9 installer comes with driver 385.xx. I have all CUDA 9 updates installed also, except update 1 which doesn't install at all.
Same as the OP I'm on an Educational License as well.
Using an older CUDA version with an updated driver should also work? I'm having trouble to switch to GPU with a GTX 970, driver 430.00 which comes with CUDA driver 10.1.120.
The CUDA 9 installer comes with driver 385.xx. I have all CUDA 9 updates installed also, except update 1 which doesn't install at all.
Same as the OP I'm on an Educational License as well.
Hi @brandjoe
CUDA 9.0 works. You can see my cuda version below. My driver is 417.35. NVIDIA says anything above 384.81 should work with 9.0. 1080Ti
CUDA 9.0 works. You can see my cuda version below. My driver is 417.35. NVIDIA says anything above 384.81 should work with 9.0. 1080Ti
I installed NVIDIA CUDA 10.1, but when going to Rapidminer Studio-Settings Preferences-General-Deep Learning Back End I set it GPU but it says I need CUDA 9.0 environment. Is 10.1 no backwards compatible with CUDA 9.0? I am running a Nvidia 2080 RTX Max-Q. Thank you.
Hello @mortiz
Nope, you should uninstall and install the 9.0 version of CUDA. But from what I heard CUDA 9.0 is not compatible with RTX 2080.
So, I guess it might not be possible, but give 9.0 a try on RTX.
Nope, you should uninstall and install the 9.0 version of CUDA. But from what I heard CUDA 9.0 is not compatible with RTX 2080.
So, I guess it might not be possible, but give 9.0 a try on RTX.
Sort by:
1 - 3 of
31
Hi Varun,
Thank you for at least letting me know that is will work with CUDA 9. So I did not mention that I am using RTX I believe. For the new cards that can utilise Tennsor units, it can use CUDA 9 (Volta and Tensor core support). The graphics driver version that supports it, however, is installing CUDA 10. There were ways to install only the cuda 9 toolkit and some users have done that. Installing the Tensorflow and Keras in Anaconda was successful and automatically install the CUDA 9.0 toolkit and I was able to run the samples from Keras. Still I think that was a problem when using it with Rapidminer. Currently the updates may may broken the version compatability....
Before RapidMiner ask for CUDA 9.0, it asked for CUDA 9.1, Does that mean something?
Thanks,
Ken
Thank you for at least letting me know that is will work with CUDA 9. So I did not mention that I am using RTX I believe. For the new cards that can utilise Tennsor units, it can use CUDA 9 (Volta and Tensor core support). The graphics driver version that supports it, however, is installing CUDA 10. There were ways to install only the cuda 9 toolkit and some users have done that. Installing the Tensorflow and Keras in Anaconda was successful and automatically install the CUDA 9.0 toolkit and I was able to run the samples from Keras. Still I think that was a problem when using it with Rapidminer. Currently the updates may may broken the version compatability....
Before RapidMiner ask for CUDA 9.0, it asked for CUDA 9.1, Does that mean something?
Thanks,
Ken
Hi all,
yes the dependency on CUDA 9.0 is correct.We changed that for compatibility reasons with the DL4J back end.
But having several CUDA toolkit instances installed should not be a problem as long as each application can find its required version.
Best,
David
yes the dependency on CUDA 9.0 is correct.We changed that for compatibility reasons with the DL4J back end.
But having several CUDA toolkit instances installed should not be a problem as long as each application can find its required version.
Best,
David
Hi @18a637y
As David mentioned you can have multiple versions. I used keras in rapidminer before but there are lot of issues with it. The same configuration used to work in python code(keras & TF) but not in RM due to issues with the extension. I got to know from RM team about updated Deep learning extension which uses DL4J instead of tensorflow, this extensions works well.
You can look at our discussion here.
https://community.rapidminer.com/discussion/comment/55180#Comment_55180
Regards,
Varun
As David mentioned you can have multiple versions. I used keras in rapidminer before but there are lot of issues with it. The same configuration used to work in python code(keras & TF) but not in RM due to issues with the extension. I got to know from RM team about updated Deep learning extension which uses DL4J instead of tensorflow, this extensions works well.
You can look at our discussion here.
https://community.rapidminer.com/discussion/comment/55180#Comment_55180
Regards,
Varun
I am using with Student License and it works fine. I have CUDA 9.0 on my pc for 1080Ti. Check if your software is registered successfully with student edition by going to Settings --> Manage Licenses. It will show below screen. If not, you need to click on "sync licenses from your account option" in the below window and login to sync.
Also, install CUDA 9.0. You can use deep learning extension from Marketplace instead of Keras in Rapid Miner. This is due to the instability of extension and issue with input shapes. Samples for deep learning extensions are provided in RM once you install it. I am also attaching a link for the Deep learning thread below.
https://community.rapidminer.com/discussion/53800/a-more-native-deep-learning-solution
Regards,
Varun