nanoporetech / medaka

Sequence correction provided by ONT Research
https://nanoporetech.com
Other
391 stars 73 forks source link

failed call to cuInit: CUDA_ERROR_UNKNOWN: unknown error #430

Closed XW26 closed 1 year ago

XW26 commented 1 year ago

This error message in Bold shows up during 'medaka_consensus':

Running medaka consensus [13:19:50 - Predict] Setting tensorflow inter/intra-op threads to 1/1. [13:19:50 - Predict] Processing region(s): contig_1:0-3201506 contig_2:0-113281 [13:19:50 - Predict] Using model: /home/microlab/anaconda3/envs/medaka/lib/python3.8/site-packages/medaka/data/r941_min_sup_g507_model.tar.gz. 2023-04-03 13:19:50.721925: E tensorflow/stream_executor/cuda/cuda_driver.cc:271] failed call to cuInit: CUDA_ERROR_UNKNOWN: unknown error [13:19:50 - Predict] Processing 5 long region(s) with batching. [13:19:51 - MdlStrTF] Model <keras.engine.sequential.Sequential object at 0x7f6002557fd0> [13:19:51 - MdlStrTF] loading weights from /tmp/tmpjfz9ivgp/model/variables/variables [13:19:51 - BAMFile] Creating pool of 16 BAM file sets.

It seems the error does not affect the final result, I still get my consensus.fasta, just wondering why and if it compromise the final quality?

cjw85 commented 1 year ago

Hi @XW26,

Warnings like this usually result from running medaka on a computer without a GPU device (or at least one one that medaka cannot access). If you do not have a GPU and wish to avoid the warnings you can install the medaka-cpu package from PyPI.

XW26 commented 1 year ago

Hi, thank you for the information! However I do have one GPU, and I have installed tensorflow and CUDA toolkit, I can well import tensorflow in python. So probably medaka cannot access GPU, any tips on how to grant access? Many thanks!

On Apr 3, 2023, at 15:09, Chris Wright @.***> wrote:



Hi @XW26https://github.com/XW26,

Warnings like this usually result from running medaka on a computer without a GPU device (or at least one one that medaka cannot access). If you do not have a GPU and wish to avoid the warnings you can install the medaka-cpu package from PyPI.

— Reply to this email directly, view it on GitHubhttps://github.com/nanoporetech/medaka/issues/430#issuecomment-1493795553, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AME66XOCYIPR4SCLXITKRNLW7JZRFANCNFSM6AAAAAAWQ2EJOY. You are receiving this because you were mentioned.Message ID: @.***>

cjw85 commented 1 year ago

I'm afraid I am unable to help address these issues. I would suggest using your favourite search engine to search for information in relation to the error text.