classify-sklearn memory error

[rcarcamoc (Roberto Cárcamo)]

Hi, I have been trying to train the classifier in the V3-V4 region (16S_341F CCTACGGGNGGCWGCAG 16S_806R GACTACHVGGGTATCTATCC) using the following command:

qiime feature-classifier fit-classifier-naive-bayes --i-reference-reads ref_seqs_16S_V3-V4.qza --i-reference-taxonomy 99_otus_16S_taxonomy.qza --o-classifier classifier_16S_V3-V4.qza --verbose 

However, it always ends with a memory error and the process is killed. I have read through the internet that a possible alternative is to set --p-reads-per-batch argument to a lower value than the one by default, but I think this argument is no longer available. Do you know how could I solve this issue?

My version of qiime is qiime2-2022.11. Thank you in advance for your time and assistance,

Roberto.

[crusher083 (Valentyn Bezshapkin)]

Hello,
Please provide the details about error. In that case it might be either RAM or disk space, and then instructions will differ.

Cheers,
V

[rcarcamoc (Roberto Cárcamo)]

Hi Valentyn,

I get no error messages, it just says killed when the process is stopped. However, I ran in parallel htop and I noticed that qiime crashes when the memory gets to 100% of usage (I think it refers to physical memory)

[crusher083 (Valentyn Bezshapkin)]

Ok, this refers to physical memory indeed. Try to reduce parameter "--p-classify--chunk-size", default is 20,000.
This should help.

Cheers
V

[rcarcamoc (Roberto Cárcamo)]

It worked!! Thank you very much for your help Valentyn!