Intel MKL FATAL ERROR: Cannot load libmkl_avx512.so or libmkl_def.so.

[ZarulHanifah]

Hello SemiBin maintainers,

I am running SemiBin coassembly mode as below.

rule semibin_binning_coassembly:
        input:
                assem = rules.medaka_polish_coassembly.output,
                bam = lambda wildcards: expand(rules.map_reads_to_coassembly.output, source = wildcards.source, sample = get_samples_given_source(wildcards.source)),
                semibin_db = config["semibin_db"]
        output:
                directory("results/binning/coassembly/semibin/{source}")
        conda:
                "../envs/semibin.yaml"
        threads: 12
        shell:
                """
                SemiBin single_easy_bin -i {input.assem} -b {input.bam} -r {input.semibin_db} -o {output}
                """

I used conda to install semibin:

name: semibin
channels:
 - bioconda
 - conda-forge
dependencies:
 - python==3.7
 - semibin=0.2=pyh5e36f6f_1

But the job exited prematurely. I am attaching the slurm error and output files.

In the slurm output, it says somethig about libmkl.

Thank you. semibin_binning_coassembly.source=riverine.err.txt semibin_binning_coassembly.source=riverine.out.txt

[psj1997]

Hi, it seems that it has an issue of pytorch. Can you try our latest SemiBin 0.3 to see if this issue still had? You can just use Semibin train command to see if the training process is normal.

Shaojun

[ZarulHanifah]

Hello Shaojun,

Thank you for the reply. I can confirm now that it is working wonderfully. Thank you.