Batch improvements

[matsvanes]

changes:

• implemented group processing in osl.preproc.run_proc_batch
• add read_dataset osl wrapper
• add skip_save argument to run_proc_chain/batch and write_dataset
• added log statements when saving data in write_dataset
• added Study.refresh() function
• added save_pkl for glm_epochs

[matsvanes]

@cgohil8 I've tested the changes (esp the group thing in osl) using something like this (make sure you adapt the paths if you're using this)

"""
Use this script to preprocess the data.
"""
import os
import osl
import logging
import glmtools as glm

raw_dir = "/Users/matsvanes/Documents/Werk/osl/meguk-debugging/raw"
outdir = "/Users/matsvanes/Documents/Werk/osl/meguk-debugging/output_mve"
study = osl.utils.Study(os.path.join(f"{raw_dir}", "sub-oxf{i_sub}/meg/sub-oxf{i_sub}_task-{task}_meg.fif"))
infiles = sorted(study.get(task="resteyesclosed"))[:2]
overwrite = True
use_dask = False

subjects = ["sub-oxf001_task-resteyesclosed", "sub-oxf002_task-resteyesclosed"]

  
def first_level(dataset, userargs):
    logger = logging.getLogger(__name__)
    dataset['raw'].info['bads'] = []        
    dataset['glm'] = osl.glm.glm_spectrum(dataset['raw'].pick_types(meg=True), fmin=1, fmax=95,
                      nperseg=int(500), noverlap=int(250),
                      mode='magnitude',
                      standardise_data=True)
    return dataset
  
def second_level(dataset, userargs):
    logger = logging.getLogger(__name__)    
    target = userargs.get("target", "glm")
    groupDC = glm.design.DesignConfig()
    for subj_ind in range(len(dataset[target])):
        regressor_name=("Subj{}".format(subj_ind+1))
        groupDC.add_regressor(name=regressor_name, rtype="Categorical", codes=subj_ind)
    groupDC.add_contrast(name="Difference",
                    values={"Subj1": 1, "Subj2": -1},
                   )
    dataset['group_glm'] = osl.glm.group_glm_spectrum(dataset[target], groupDC)
    return dataset


config = """
meta:
  event_codes:
preproc:
  - crop:               {tmin: 30}
  - filter:             {l_freq: 0.5, h_freq: 125, method: iir, iir_params: {order: 5, ftype: butter}}
  - notch_filter:       {freqs: 50 100, notch_widths: 2} 
  - resample:           {sfreq: 250}
  - bad_segments:       {segment_len: 500, picks: grad, significance_level: 0.1}
  - bad_segments:       {segment_len: 500, picks: mag, significance_level: 0.1}
  - bad_segments:       {segment_len: 500, picks: grad, mode: diff, significance_level: 0.1}
  - bad_segments:       {segment_len: 500, picks: mag, mode: diff, significance_level: 0.1}
  - bad_channels:       {picks: meg, significance_level: 0.1}
  - interpolate_bads:   {reset_bads: False}
  - first_level:        {}
group:
  - second_level:       {target: glm}
"""

if use_dask:
  from dask.distributed import Client
  client = Client(n_workers=2, threads_per_worker=1)

goods = osl.preprocessing.run_proc_batch(
    config, 
    infiles, 
    subjects=subjects,
    outdir=outdir,
    extra_funcs=[first_level, second_level],
    overwrite=overwrite,
    dask_client=use_dask,
)

[cgohil8]

This looks good.