No file format adaptor could read this file.

[daskandalis]

I am trying to run sleap on a cluster, which seems to run correctly until training is initialised.

INFO:sleap.nn.training:Loading training labels from: sleap/RL_40-pt_9-cam.pkg.slp
sleap/RL_40-pt_9-cam.pkg.slp doesn't match ext for json or json.zip
Traceback (most recent call last):
  File "/home/usr/.conda/envs/sleap-1.3.3/bin/sleap-train", line 8, in <module>
    sys.exit(main())
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/nn/training.py", line 2013, in main
    trainer = create_trainer_using_cli(args=args)
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/nn/training.py", line 2000, in create_trainer_using_cli
    trainer = Trainer.from_config(
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/nn/training.py", line 666, in from_config
    data_readers = DataReaders.from_config(
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/nn/training.py", line 143, in from_config
    return cls.from_labels(
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/nn/training.py", line 167, in from_labels
    training: Labels = sleap.load_file(
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/io/dataset.py", line 2774, in load_file
    return Labels.load_file(filename, search_paths, match_to=match_to)
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/io/dataset.py", line 1968, in load_file
    return read(
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/io/format/main.py", line 113, in read
    return disp.read(filename, *args, **kwargs)
  File "/home/usr/.conda/envs/sleap-1.3.3/lib/python3.10/site-packages/sleap/io/format/dispatch.py", line 59, in read
    raise TypeError("No file format adaptor could read this file.")
TypeError: No file format adaptor could read this file.

The error might come from the custom install script, if a dependency is missing:

conda create -n sleap-1.3.3 python=3.10 -y 
conda activate sleap-1.3.3

wget https://github.com/talmolab/sleap/archive/refs/tags/v1.3.3.tar.gz

tar -zxvf v1.3.3.tar.gz
cd sleap-1.3.3
pip install -r pypi_requirements.txt 
pip install sleap
pip install imgstore==0.2.9 python-dateutil==2.8.2 ruamel.yaml==0.16.13  
pip install pynwb==2.3.3
conda install -n sleap-1.3.3 pyqt -y
pip install segmentation_models

python -c "import sleap; sleap.system_summary()"

Any insight would be appreciated. The training package has been built on both Linux and Windows, using settings that work fine locally.

[eberrigan]

Hi @daskandalis,

Please include the command you used.

Thanks!

Elizabeth

[daskandalis]

Just the regular command:

(sleap-1.3.3) [usr@gpu10 USER]$ sleap-train sleap/single_instance.json sleap/RL_40-pt_9-cam.pkg.slp

Thanks!

[daskandalis]

No obvious cause for this, and it's apparently resolved by repeatedly reinstalling until it works.

[roomrys]

Hi @daskandalis,

I am just doing a bit of journaled thinking below, but if you can, would you mind sharing the pkg.slp with us for troubleshooting? If you feel comfortable sharing, we have a user-upload form.

Thanks,
Liezl

---

TypeError: No file format adaptor could read this file.

The LabelsV1Adaptor should be able to read .pkg.slp extensions, but either LabelsV1Adaptor.can_read_file is returning False sporatically

sleap/sleap/io/format/hdf5.py

Lines 53 to 62 in 5ab305c

	def can_read_file(self, file: format.filehandle.FileHandle):
	if not self.does_match_ext(file.filename):
	return False
	if not file.is_hdf5:
	return False
	if file.format_id is not None and file.format_id >= 2:
	return False
	if "metadata" not in file.file:
	return False
	return True

Or LabelsV1Adaptor is not being registered in the list of Dispatch._adaptors

sleap/sleap/io/format/dispatch.py

Lines 20 to 37 in 5ab305c

	def register(self, adaptor: Union[Adaptor, type, List[Adaptor]]):
	"""
	Registers the class which reads/writes specific file format.
	"""
	# If given a class, then instantiate it since we want the object
	if hasattr(adaptor, "__iter__"):
	self.register_list(adaptor)
	else:
	if type(adaptor) == type:
	adaptor = adaptor()
	self._adaptors.append(adaptor)
	def register_list(self, adaptor_list: List[Union[Adaptor, type]]):
	"""Convenience function for registering multiple adaptors."""
	for adaptor in adaptor_list:
	self.register(adaptor)

sleap/sleap/io/format/main.py

Lines 105 to 111 in 5ab305c

	if for_object == "labels" or hasattr(for_object, "labeled_frames"):
	if as_format == "*":
	for format_name, adaptor in all_labels_adaptors.items():
	disp.register(adaptor)
	# print(f"[registering format adaptor for {format_name}]")
	else:
	disp.register_list(default_labels_adaptors)

[daskandalis]

Uploaded an example project that fails to open on my end. I successfully opened it to continue training once, but after that it is failing. Maybe renaming cfg.outputs.run_name?

[roomrys]

Hi @daskandalis,

I downloaded your data and ran sleap-train several times without running into the issue - not great feedback from a debug session.

I am wondering if this could be a remote training problem?

Things to try would be to run sleap-train

• with absolute paths instead of relative paths
• in the sleap directory without any slashes

I am making the assumption that you are running training remotely. If so, I am curious about what version of SLEAP you have on you local computer vs the remote training computer?

Thanks,
Liezl