Get errors "NVTX functions not installed. Are you sure you have a CUDA build?" when run RF2NA on CPU. #36
Comments
|
I am having the same issue. Any updates on this? |
|
The environment file is missing a dependency for pytorch to see cuda. You have to add/modify lines in the environment file: Here is my modified conda environment file: Cheers, Kamil |
|
Thanks @kcygan! The NVTX errors disappeared. But I got a new DGL error shown below. The details are here: The current package versions: |
|
Hiii I have reported same issue I have updated the software as you mentioned but the problem still there can't figure out what to do.. This problem I have faced in HPC installation.... Help me out what to do... |
|
I resolved this issue by uninstalling Pytorch and reinstalling Pytorch with cuda support |
|
run |
|
Dear authors, I have found the same problem, tried caseyh9438 and v-shaoningli suggestions but still having exactly the same error: I paste the full output just in case somebody can advise, (RF2NA) rosetta@amber3:~/RoseTTAFold2NA/example$ ../run_RF2NA.sh rna_pred rna_binding_protein.fa R:RNA.fa |
|
Hello, I have solved them by using a modified conda environment with CUDA 11.6 and cudatoolkit from nvidia instead of conda-forge
Also, I have to systematically run Hope it helps. |
|
Dear Jessica
Thanks a lot for your kind answer,
Could you please detail a little how exactly did you proceed?
Regards,
Fabian
…--------------------------
Fabian Glaser, PhD
Structural and Computational Biology Unit
The Lorry I. Lokey Center for Life Sciences and Engineering
Technion - Israel Institute of Technology, Haifa, Israel
On 2 Dec 2023, at 15:58, Jessica Andreani ***@***.***> wrote:
Hello,
Not sure if it will work for you but I have experienced issues with NVTX and GPU too.
I have solved them by using a modified conda environment with CUDA 11.6 and cudatoolkit from nvidia instead of conda-forge
nvidia::cudatoolkit=11.6
dglteam::dgl-cuda11.6
Also, I have to systematically run
export CUDA_VISIBLE_DEVICES=0
once I have activated the conda environment and before running the RF2NA program.
Hope it helps.
—
Reply to this email directly, view it on GitHub <#36 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACSBVSC5CPVWIVNXC3QHHELYHMXX7AVCNFSM6AAAAAAS7HETVKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMZXGE2TIOJXG4>.
You are receiving this because you commented.
|
|
Dear Fabian,
This solved my library errors, but I haven't tried with the new yml file yet. Also, if I did not run the command indicated before (export CUDA_VISIBLE_DEVICES=0) in the terminal before running RF2NA, my GPU card was not detected. Best regards |
|
Jessica, maybe I misunderstand something here, but you say that you have to run the command you indicated, unless your GPU card was not detected. Isn't the whole point of this thread to get the program run without the need for GPU? I find it a huge drawback that you still need a specific GPU card for RF2NA to run. Also, there are no real alternatives to running the program locally, no Colab notebook, nothing. |
|
Hello, sorry if I wasn't clear. The program runs on CPU by default. We had trouble getting it to run on GPU, because even though we had an available GPU, it was not detecting it until I found that I need to specify this "export CUDA_VISIBLE_DEVICES=0". |
|
No problem. So I should be able to run it on CPU while not having an NVIDIA GPU, nor cuda? |
|
For those still struggling with the newest yml file, I think I stumbled upon a solution that may work universally. Before doing anything, run this:
Based on https://pytorch.org/blog/deprecation-cuda-python-support/, I substituted pytorch for pytorch::pytorch=2.0 in the yml file, because just pytorch installs pytorch 2.1.*. So my yml file looks like this: Then go about with Finally, probably unnecessary to say, but after having run
|
|
@stianale that worked great thank you. After doing your solution I got an error that said
A simple I wonder if pandas should be in the .yml file? Did anyone else have install pandas on their own? |
|
I get the same error with the version released on April, 14th. Going to install pandas as well and see if this fixed it |
|
That fixed the pandas error. I also tried the approach of @stianale. It is still not run on the gpu. Still have issues running the example though. @samhimes92 did you change anything else? |
|
@boesr |
|
For everyone else having the same error. The RNA database was corrupt due to ASCI characters. I exchanged the installation command to That does run now. It also uses the GPU (because of the other errors I never reached the "gpu part" of the test script) |
|
Hello everyone, please see #105 for potential solution. |
|
I also had this issue and want to add some points: However, the new environment will only work if the RF2NA name is used and not a new name like CudaTestEnv. The bash-script must be run out of the base env and then activates the RF2NA env. I added a (for me working) yml file in my fork including comments. |
First of all, thank you very much for this suggestion. After having Do you know how to resolve those invalid residues as well? |
Hi,
I get the following errors when run RoseTTAFold2NA on CPU. I have replaced the "torch.cuda.amp.autocast" with "torch.amp.autocast" in predict.py in order to overcome another "NVTX functions not installed." error when running the run_RF2NA.sh.
Seems there is some part of the codes still calling cuda or searching for GPUs?
Thank you!
My package versions:
The text was updated successfully, but these errors were encountered: