Problem with Zhou2016

[toncho11]

I am using MOABB from git and MOABB benchmark function.

Does anyone has any problems with Zhou2016? Is it possible that the last commit to Zhou2016 made it unusable?

I am getting this error:

Zhou2016-WithinSession:   0%|          | 0/4 [00:00<?, ?it/s]
---------------------------------------------------------------------------
OverflowError                             Traceback (most recent call last)
File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\mne\io\cnt\cnt.py:545, in RawCNT.__init__(self, input_fname, eog, misc, ecg, emg, data_format, date_format, header, preload, verbose)
    544 try:
--> 545     info, cnt_info = _get_cnt_info(
    546         input_fname, eog, ecg, emg, misc, data_format, _date_format, header
    547     )
    548 except Exception:

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\mne\io\cnt\cnt.py:346, in _get_cnt_info(input_fname, eog, ecg, emg, misc, data_format, date_format, header)
    345 if data_format == "auto":
--> 346     if n_samples == 0 or data_size // (n_samples * n_channels) not in [2, 4]:
    347         warn(
    348             "Could not define the number of bytes automatically. "
    349             "Defaulting to 2."
    350         )

OverflowError: Python int too large to convert to C long

During handling of the above exception, another exception occurred:

RuntimeError                              Traceback (most recent call last)
File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\spyder_kernels\customize\utils.py:209, in exec_encapsulate_locals(code_ast, globals, locals, exec_fun, filename)
    207     if filename is None:
    208         filename = "<stdin>"
--> 209     exec_fun(compile(code_ast, filename, "exec"), globals, None)
    210 finally:
    211     if use_locals_hack:
    212         # Cleanup code

File c:\work\pythoncode\ml_examples\eeg\mdm-mf\article_comparison_moabb.py:81
     77     pipeline_folder = "C:\\Work\\PythonCode\\ML_examples\\EEG\\MDM-MF\\pipelines9\\"
     79 #The EN_grid.yml is too slow!!!!!!!! and memory consuming
---> 81 results = benchmark(
     82     pipelines=pipeline_folder,
     83     evaluations=["WithinSession"],
     84     paradigms=["LeftRightImagery"],
     85     #include_datasets=["BNCI2014_001"],
     86     #include_datasets=["Stieger2021"],
     87     
     88     exclude_datasets=["Stieger2021","Liu2024"],
     89     results="./results/",
     90     overwrite=False,
     91     plot=False,
     92     n_jobs=9, #4 otherwise memory is not enough
     93     output="./benchmark/",
     94 )
     96 print("Results:")
     97 print(results)

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\benchmark.py:214, in benchmark(pipelines, evaluations, paradigms, results, overwrite, output, n_jobs, plot, contexts, include_datasets, exclude_datasets, n_splits, cache_config, optuna)
    202 if len(ppl_with_array) > 0:
    203     context = eval_type[evaluation](
    204         paradigm=p,
    205         datasets=d,
   (...)
    212         optuna=optuna,
    213     )
--> 214     paradigm_results = context.process(
    215         pipelines=ppl_with_array, param_grid=param_grid
    216     )
    217     paradigm_results["paradigm"] = f"{paradigm}"
    218     paradigm_results["evaluation"] = f"{evaluation}"

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\evaluations\base.py:237, in BaseEvaluation.process(self, pipelines, param_grid, postprocess_pipeline)
    228 # (we only keep the pipeline for the first frequency band, better ideas?)
    230 results = self.evaluate(
    231     dataset,
    232     pipelines,
   (...)
    235     postprocess_pipeline=postprocess_pipeline,
    236 )
--> 237 for res in results:
    238     self.push_result(res, pipelines, process_pipeline)
    239 res_per_db.append(
    240     self.results.to_dataframe(
    241         pipelines=pipelines, process_pipeline=process_pipeline
    242     )
    243 )

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\evaluations\evaluations.py:427, in WithinSessionEvaluation.evaluate(self, dataset, pipelines, param_grid, process_pipeline, postprocess_pipeline)
    423     yield from self._evaluate_learning_curve(
    424         dataset, pipelines, process_pipeline, postprocess_pipeline
    425     )
    426 else:
--> 427     yield from self._evaluate(
    428         dataset, pipelines, param_grid, process_pipeline, postprocess_pipeline
    429     )

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\evaluations\evaluations.py:157, in WithinSessionEvaluation._evaluate(self, dataset, pipelines, param_grid, process_pipeline, postprocess_pipeline)
    154     return []
    156 # get the data
--> 157 X, y, metadata = self.paradigm.get_data(
    158     dataset=dataset,
    159     subjects=[subject],
    160     return_epochs=self.return_epochs,
    161     return_raws=self.return_raws,
    162     cache_config=self.cache_config,
    163     postprocess_pipeline=postprocess_pipeline,
    164 )
    165 # iterate over sessions
    166 for session in np.unique(metadata.session):

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\paradigms\base.py:274, in BaseProcessing.get_data(self, dataset, subjects, return_epochs, return_raws, cache_config, postprocess_pipeline)
    269 process_pipelines = self.make_process_pipelines(
    270     dataset, return_epochs, return_raws, postprocess_pipeline
    271 )
    272 labels_pipeline = self.make_labels_pipeline(dataset, return_epochs, return_raws)
--> 274 data = [
    275     dataset.get_data(
    276         subjects=subjects,
    277         cache_config=cache_config,
    278         process_pipeline=process_pipeline,
    279     )
    280     for process_pipeline in process_pipelines
    281 ]
    283 X = []
    284 labels = []

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\paradigms\base.py:275, in <listcomp>(.0)
    269 process_pipelines = self.make_process_pipelines(
    270     dataset, return_epochs, return_raws, postprocess_pipeline
    271 )
    272 labels_pipeline = self.make_labels_pipeline(dataset, return_epochs, return_raws)
    274 data = [
--> 275     dataset.get_data(
    276         subjects=subjects,
    277         cache_config=cache_config,
    278         process_pipeline=process_pipeline,
    279     )
    280     for process_pipeline in process_pipelines
    281 ]
    283 X = []
    284 labels = []

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\datasets\base.py:433, in BaseDataset.get_data(self, subjects, cache_config, process_pipeline)
    431     if subject not in self.subject_list:
    432         raise ValueError("Invalid subject {:d} given".format(subject))
--> 433     data[subject] = self._get_single_subject_data_using_cache(
    434         subject,
    435         cache_config,
    436         process_pipeline,
    437     )
    438 check_subject_names(data)
    439 check_session_names(data)

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\datasets\base.py:527, in BaseDataset._get_single_subject_data_using_cache(self, subject, cache_config, process_pipeline)
    525 # Load and eventually overwrite:
    526 if len(cached_steps) == 0:  # last option: we don't use cache
--> 527     sessions_data = self._get_single_subject_data(subject)
    528     assert sessions_data is not None  # should not happen
    529 else:

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\moabb\datasets\Zhou2016.py:104, in Zhou2016._get_single_subject_data(self, subject)
    102 for run_ind, fname in enumerate(runlist):
    103     run_key = str(run_ind)
--> 104     raw = read_raw_cnt(fname, preload=True, eog=["VEOU", "VEOL"])
    105     stim = raw.annotations.description.astype(np.dtype("<10U"))
    106     stim[stim == "1"] = "left_hand"

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\mne\io\cnt\cnt.py:263, in read_raw_cnt(input_fname, eog, misc, ecg, emg, data_format, date_format, header, preload, verbose)
    178 @fill_doc
    179 def read_raw_cnt(
    180     input_fname,
   (...)
    190     verbose=None,
    191 ) -> "RawCNT":
    192     """Read CNT data as raw object.
    193 
    194     .. Note::
   (...)
    261     .. versionadded:: 0.12
    262     """
--> 263     return RawCNT(
    264         input_fname,
    265         eog=eog,
    266         misc=misc,
    267         ecg=ecg,
    268         emg=emg,
    269         data_format=data_format,
    270         date_format=date_format,
    271         header=header,
    272         preload=preload,
    273         verbose=verbose,
    274     )

File C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\mne\io\cnt\cnt.py:549, in RawCNT.__init__(self, input_fname, eog, misc, ecg, emg, data_format, date_format, header, preload, verbose)
    545     info, cnt_info = _get_cnt_info(
    546         input_fname, eog, ecg, emg, misc, data_format, _date_format, header
    547     )
    548 except Exception:
--> 549     raise RuntimeError(
    550         f"{_explain_exception()}\n"
    551         "WARNING: mne.io.read_raw_cnt "
    552         "supports Neuroscan CNT files only. If this file is an ANT Neuro CNT, "
    553         "please use mne.io.read_raw_ant instead."
    554     )
    555 last_samps = [cnt_info["n_samples"] - 1]
    556 super().__init__(
    557     info,
    558     preload,
   (...)
    563     verbose=verbose,
    564 )

RuntimeError: :
>   File "C:\Work\PythonCode\envs\eeg4_pr_moabb_git\lib\site-packages\mne\io\cnt\cnt.py", line 346, in _get_cnt_info
>     if n_samples == 0 or data_size // (n_samples * n_channels) not in [2, 4]:
> 
> OverflowError: Python int too large to convert to C long

WARNING: mne.io.read_raw_cnt supports Neuroscan CNT files only. If this file is an ANT Neuro CNT, please use mne.io.read_raw_ant instead.

[toncho11]

@bruAristimunha can you have a look please?

[toncho11]

I am using MNE 1.9.0 and pyRIemann 0.8.dev0

[toncho11]

So the problem is in the version of numpy.
Using numpy 1.26.4 corrects the problem: pip install numpy==1.26.4
The problem is that when you install pyRiemann from git it upgrates your version of numpy to 2.2 which causes the problem later.
So it is in this order:

• install MOABB
• upgrade to newer version of pyRiemann (if you need to)
• then downgrade numpy to 1.26.4

The best solution is to check why MOABB gives an error with numpy 2.2 and correct it.

[bruAristimunha]

Looks something which is the responsibility of the Mne @toncho11, no?

I would love to solve it, but the code breakage happens with a Mne function. What do you think?