Finished without creating output_bins

[Sung-hub]

I am running SemiBin2 single_easy_bin and experiencing premature finishing without output_bins. There was no error message. I used large size metagenome assemblies, and the assemblies were made from pacbio hifi long-read sequences. Some of runs successfully finished, but most of them could not create output_bins. Even though I put -t 30, it generally run on single thread. I could not see any memory issue on our server.

This command line I put in:
$ nohup singularity exec https://depot.galaxyproject.org/singularity/semibin:2.1.0--pyhdfd78af_0 SemiBin2 single_easy_bin -i /home/sung.shin/pool_2_metaMDBG/contigs.fasta -b pool2metaMDBG_aln.sort.bam -o pool2_metaMDBG_sb2_binning/ --self-supervised --sequencing-type=long_reads -t 30 --random-seed 123 2>semibin.log &

log messages I got:
$ cat semibin.log
nohup: ignoring input and appending output to 'nohup.out'
2025-01-03 11:22:08 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Binning for long_read
2025-01-03 11:22:12 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Did not detect GPU, using CPU.
2025-01-03 11:22:38 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Generating training data...
2025-01-03 15:17:52 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Calculating coverage for every sample.
2025-01-03 16:44:37 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Processed: pool2metaMDBG_aln.sort.bam
2025-01-03 16:52:53 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Start training from a single sample.
2025-01-03 16:53:42 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Training model...
100%|██████████| 15/15 [53:05<00:00, 212.34s/it]
2025-01-03 17:46:48 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Training finished.
2025-01-03 17:46:49 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Start binning.
2025-01-03 17:47:59 arsnecla0ap2.marc.usda.gov SemiBin[1485536] INFO Running naive ORF finder
[sung.shin@arsnecla0ap2 pool2_dastool]$ cd pool2_metaMDBG_sb2_binning/

$ cat SemiBinRun.log
[2025-01-03 11:22:08,400] INFO: Binning for long_read
[2025-01-03 11:22:12,494] INFO: Did not detect GPU, using CPU.
[2025-01-03 11:22:38,866] INFO: Generating training data...
[2025-01-03 15:17:52,924] INFO: Calculating coverage for every sample.
[2025-01-03 16:44:37,329] INFO: Processed: pool2metaMDBG_aln.sort.bam
[2025-01-03 16:52:53,075] INFO: Start training from a single sample.
[2025-01-03 16:53:42,830] INFO: Training model...
[2025-01-03 17:46:48,649] INFO: Training finished.
[2025-01-03 17:46:49,343] INFO: Start binning.
[2025-01-03 17:47:59,090] INFO: Running naive ORF finder

resulting files:
$ ls
data.csv data_split.csv markers.hmmout model.h5 pool2metaMDBG_aln.sort.bam_0_data_cov.csv SemiBinRun.log

Could you give me some advice?

[Sung-hub]

Is this issue can be possibly fixed by running binning step separately, when run was finished without output_bins?
I found "Advanced single-sample binning workflows" on https://semibin.readthedocs.io/en/latest/usage/.
What is difference between bin_short and bin_long?
SemiBin2 bin_short
-i S1.fa
--environment human_gut
--data S1_output/data.csv
-o S1_output
SemiBin2 bin_long
-i S1.fa
--environment human_gut
--data S1_output/data.csv
-o S1_output