Tensorfllow: not an sstable (bad magic number)

[ShyGuy509]

Hi,

Thanks for the program. It looks great, but I'm having trouble running it. I've just cloned it and created the environment, and tried to run it with the following command:

mcdevol.py
-a abundance_unzipped.tsv
-c contigs_unzipped.fa
-o output
--abundformat metabat
-n 12 \

The abundance file is the output of metabat2 jgi_summarize_bam_contig_depths, and the contigs are generated with megahit.

However, after trying with a couple of datasets I'm getting the following error output:

McDevol is running
Traceback (most recent call last):
File "/data/users/danross/McDevol/mcdevol/mcdevol.py", line 178, in
main()
File "/data/users/danross/McDevol/mcdevol/mcdevol.py", line 147, in main
numcontigs, contig_length, contig_names = ps.compute_kmerembeddings(outdir, fasta_file, min_length, logger, n_fragments=nfragments)
File "/data/users/danross/McDevol/mcdevol/dataparsing.py", line 136, in compute_kmerembeddings
kmermodel.load_weights(path_weight)
File "/data/users/danross/miniconda3/envs/mcdevol_env/lib/python3.9/site-packages/keras/utils/traceback_utils.py", line 67, in error_handler
raise e.with_traceback(filtered_tb) from None
File "/data/users/danross/miniconda3/envs/mcdevol_env/lib/python3.9/site-packages/tensorflow/python/training/py_checkpoint_reader.py", line 45, in error_translator
raise errors_impl.OpError(None, None, error_message, errors_impl.UNKNOWN)
tensorflow.python.framework.errors_impl.OpError: not an sstable (bad magic number)

Can you help me figure out what's going on? I'm not familiar enough with tensorflow to understand the error.

Thank you so much.
David

[Rasmus-allesoee]

Hi,

I am having the exact same problem. I may have found the root cause of the tensorflow.python.framework.errors_impl.OpError: not an sstable (bad magic number) error.

When I cloned the repository, the model weight files in mcdevol/genomeface_weights/ turned out to be only 134-byte pointer text files instead of the actual TensorFlow checkpoint files.

When I tried pulling them with Git LFS, I got the following error:

Error downloading object: mcdevol/genomeface_weights/general_t2eval.m.data-00000-of-00001 (0584bbd): Smudge error: Error downloading mcdevol/genomeface_weights/general_t2eval.m.data-00000-of-00001 (0584bbd6c4dbf0b29911df12cc462d5eb45a8c5b38a3cdad2401a3f3f37ea761): batch response: Git LFS is disabled for this repository.
Errors logged to '/work/software/binning_tools/McDevol/.git/lfs/logs/20251008T181511.423024989.log'.
Use git lfs logs last to view the log.
error: external filter 'git-lfs filter-process' failed
fatal: mcdevol/genomeface_weights/general_t2eval.m.data-00000-of-00001: smudge filter lfs failed
warning: Clone succeeded, but checkout failed.
You can inspect what was checked out with 'git status'
and retry with 'git restore --source=HEAD :/'

This seems to indicate that the repository is configured to use Git LFS, but the large model files were never uploaded to GitHub’s LFS storage? As a result, TensorFlow is trying to read pointer files instead of real checkpoint data, which causes the “not an sstable” error.

So, I also need help resolving this. Maybe you could either re-enable or upload the missing LFS files, or provide the pretrained models via a release or download link?

Thanks!

[Sangita91]

Hi developers,
I’m trying to run McDevol on my own dataset, but it seems that the pretrained model files are missing or incomplete.
After cloning the repository, my genomeface_weights/ folder contains only very small files:
-rw-rw-r-- 1 user user 129 Oct 28 13:47 general_t2eval.m.index
-rw-rw-r-- 1 user user 134 Oct 28 13:47 model.index

When I run McDevol, I get the following error:

tensorflow.python.framework.errors_impl.OpError: not an sstable (bad magic number)

It appears that the full .data-00000-of-00001 and .index weight files are missing from Git LFS or were not uploaded correctly.

Could you please:

Re-enable LFS for these files, or

Provide a download link / release package for the pretrained model weights?

This would really help those of us trying to use McDevol for genome binning.

Thanks for your time and great work on this project!