Error: 'Unable to synchronously open object (message not aligned)'

[cl05e]

Hi. I was running the predictions normally, then after a few runs, the error in the title appeared.

The output of the run window was "Inference failed or was canceled." and the log shows this:

Running: sleap track --gui --data_path C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social/MO-batch2.slp --video_index 1 --frames 15114 --model_paths C:\Users\sysadmin.AORUS-SYSADMIN\Documents\NATAD\SLEAP\cond-MO-social\models\260604_152552.bottomup.n=423 -o C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social\predictions\041526-MO-cond-B2B-290F-D5.mp4_MO-batch2.slp.260605_103621.predictions.slp --batch_size 4 --max_instances 2
W0605 10:36:26.348000 27808 Lib\site-packages\torch\utils\flop_counter.py:29] triton not found; flop counting will not work for triton kernels
2026-06-05 10:36:29 | Started inference at: 2026-06-05 10:36:29.752329
2026-06-05 10:36:31 | sleap-nn 0.2.0 | Python 3.13.12 | PyTorch 2.11.0+cu128 | CUDA 12.8 | 1 GPU(s)
2026-06-05 10:36:31 | Using device: cuda
2026-06-05 10:36:32 | Finished inference at: 2026-06-05 10:36:32.958327
2026-06-05 10:36:32 | Total runtime: 3.2060041427612305 secs
2026-06-05 10:36:33 | Predictions output path: C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social\predictions\041526-MO-cond-B2B-290F-D5.mp4_MO-batch2.slp.260605_103621.predictions.slp
2026-06-05 10:36:33 | Saved file at: 2026-06-05 10:36:33.720102
Error: 'Unable to synchronously open object (message not aligned)'

Is the predictions file or the project file corrupted? Any ideas how this happened and how to fix this? Thank you!

EDIT: It's now giving more details:

Running: sleap track --gui --data_path C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social/MO-batch2.slp --video_index 1 --frames 15114 --model_paths C:\Users\sysadmin.AORUS-SYSADMIN\Documents\NATAD\SLEAP\cond-MO-social\models\260604_152552.bottomup.n=423 -o C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social\predictions\041526-MO-cond-B2B-290F-D5.mp4_MO-batch2.slp.260605_103621.predictions.slp --batch_size 4 --max_instances 2
W0605 10:36:26.348000 27808 Lib\site-packages\torch\utils\flop_counter.py:29] triton not found; flop counting will not work for triton kernels
2026-06-05 10:36:29 | Started inference at: 2026-06-05 10:36:29.752329
2026-06-05 10:36:31 | sleap-nn 0.2.0 | Python 3.13.12 | PyTorch 2.11.0+cu128 | CUDA 12.8 | 1 GPU(s)
2026-06-05 10:36:31 | Using device: cuda
2026-06-05 10:36:32 | Finished inference at: 2026-06-05 10:36:32.958327
2026-06-05 10:36:32 | Total runtime: 3.2060041427612305 secs
2026-06-05 10:36:33 | Predictions output path: C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social\predictions\041526-MO-cond-B2B-290F-D5.mp4_MO-batch2.slp.260605_103621.predictions.slp
2026-06-05 10:36:33 | Saved file at: 2026-06-05 10:36:33.720102
Traceback (most recent call last):
File "", line 198, in _run_module_as_main
File "", line 88, in run_code
File "C:\Users\sysadmin.AORUS-SYSADMIN.local\bin\sleap.exe_main.py", line 10, in
sys.exit(cli())

File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\rich_click\rich_command.py", line 402, in call
return super().call(*args, **kwargs)
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\click\core.py", line 1524, in call
return self.main(*args, **kwargs)
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\rich_click\rich_command.py", line 216, in main
rv = self.invoke(ctx)
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\click\core.py", line 1912, in invoke
return process_result(sub_ctx.command.invoke(sub_ctx))
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\click\core.py", line 1308, in invoke
return ctx.invoke(self.callback, **ctx.params)
~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\click\core.py", line 877, in invoke
return callback(*args, **kwargs)
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\sleap_nn\cli.py", line 959, in track
return run_inference(**kwargs)
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\sleap_nn\predict.py", line 882, in run_inference
output.save(Path(output_path).as_posix(), restore_original_videos=False)
~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\sleap_io\model\labels.py", line 1365, in save
save_file(
~~~~~~~~~^
self,
^^^^^
...<6 lines>...
**kwargs,
^^^^^^^^^
)
^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\sleap_io\io\main.py", line 1018, in save_file
save_slp(
~~~~~~~~^
labels,
^^^^^^^
...<3 lines>...
**kwargs,
^^^^^^^^^
)
^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\sleap_io\io\main.py", line 100, in save_slp
return slp.write_labels(
~~~~~~~~~~~~~~~~^
filename,
^^^^^^^^^
...<6 lines>...
progress_callback=progress_callback,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\sleap_io\io\slp.py", line 5614, in write_labels
write_metadata(labels_path, labels)
~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\sysadmin.AORUS-SYSADMIN\AppData\Roaming\uv\tools\sleap\Lib\site-packages\sleap_io\io\slp.py", line 1528, in write_metadata
grp.attrs["json"] = np.bytes(json.dumps(md, separators=(",", ":")))
~~~~~~~~~^^^^^^^^
Error: Inference failed with code 1

[yudin-s]

That error is coming from the HDF5/SLP file write path, after inference has already finished. I would treat the generated *.predictions.slp as suspect or partially written, but I would not assume the original project file or model is corrupted from this log alone.

The first isolation I would try is writing the prediction file to a very simple local path, outside any synced folder and with no existing output file:

sleap track ... -o C:/temp/sleap-test.predictions.slp

Also make sure the output file is not open in the GUI while the CLI is writing it, and check free disk space/antivirus/cloud-sync interference. On Windows, HDF5 errors like "message not aligned" can show up around incomplete writes, files being overwritten while open, or failed writes later being read back.

For cleanup, I would:

1. Move the failed .predictions.slp aside instead of reusing it.
2. Rerun one short frame range to a new local output path.
3. If that works, rerun the full prediction to the same kind of clean path.
4. Only then import/merge the prediction file back into the project.

If it fails even at C:/temp with a short frame range, then the useful next detail for maintainers is the exact SLEAP/sleap-io/sleap-nn versions and whether the same command fails on CPU as well as CUDA.

If the clean local-path test works, that is probably the useful answer for future readers.

[cl05e]

hi @gitttt-1234 !

It still won't work. However, the error is different now:

Running: sleap track --gui --data_path C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social/MO-batch2.slp --video_index 0 --frames 18194 --model_paths C:\Users\sysadmin.AORUS-SYSADMIN\Documents\NATAD\SLEAP\cond-MO-social\models\260604_152552.bottomup.n=423 -o C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social\predictions\041526-MO-cond-B2B-290F-D5.mp4_MO-batch2.slp.260624_151131.predictions.slp --max_instances 2
Error: 'charmap' codec can't decode byte 0x90 in position 627: character maps to <undefined>

Have you encountered something like this before? I can upload the files later, I just need to organize them into a .zip.

Thanks again!

[gitttt-1234]

Hi @cl05e !

We have't encountered this before, it would be great if you could upload the files so that we can take a deeper look at it!

Thanks,

Divya

[cl05e]

hi @gitttt-1234 ! I have uploaded the files thru the Google forms link. Let me know if there's anything else you need.

Thank you!

[gitttt-1234]

Hi @cl05e !

Apologies for the delay! Your project's merge history has grown large enough to hit a size limit which is an HDF5 constraint, and inference flags this when we save the predictions to create a slp file. We've fixed this in our I/O library (sleap-io). It'll be a part of the upcoming release, to use it right away, you can install sleap from source!

Thanks,

Divya

[cl05e]

hi @gitttt-1234 , that's great news! Just curious, what can I do right now in the old version? Is there anything to do to reset the merge history of the project?

[gitttt-1234]

Hi @cl05e !

Could you try this trick here to reset the metadata and save it in a separate json for a backup?

Thanks,

Divya

[cl05e]

Thank you so much!! I will try to install the pre-release version and then go from there.

[cl05e]

hi @gitttt-1234 I'm still getting an error:

Running: sleap track --gui --data_path C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social/MO-batch2.slp --video_index 0 --frames 18072 --model_paths C:\Users\sysadmin.AORUS-SYSADMIN\Documents\NATAD\SLEAP\cond-MO-social\models\260604_152552.bottomup.n=423 -o C:/Users/sysadmin.AORUS-SYSADMIN/Documents/NATAD/SLEAP/cond-MO-social\predictions\041526-MO-cond-B2B-290F-D5.mp4_MO-batch2.slp.260702_134104.predictions.slp --max_instances 2
SLEAP - Social LEAP Estimates Animal Poses
Usage: sleap track [OPTIONS]
 Try 'sleap track -h' for help
┌─ Error ─────────────────────────────────────────────────────────────────────┐
│ No such option '--gui'.                                                     │
└─────────────────────────────────────────────────────────────────────────────┘
Error: Inference failed with code 2

I followed the instructions on installing the pre-release. Part of it was installing 'sleap nn' so not sure what's happening here.