-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrat_robot_kilosort4_iblcbin_hc4.py
216 lines (177 loc) · 9.56 KB
/
rat_robot_kilosort4_iblcbin_hc4.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
# Set logging before the rest as neo (and neo-based imports) needs to be imported after logging has been set
import logging
import os
logging.basicConfig(
format='%(asctime)s %(levelname)-8s %(message)s',
level=logging.INFO,
datefmt='%Y-%m-%d %H:%M:%S')
logger = logging.getLogger('sorting')
logger.setLevel(logging.DEBUG)
from pathlib import Path
import datetime
import json
import jsmin
from jsmin import jsmin
os.environ['CUDA_VISIBLE_DEVICES'] = '0'
os.environ['NUMEXPR_MAX_THREADS'] = '16'
os.environ['OPENBLAS_NUM_THREADS'] = '16'
import spikeinterface.extractors as se
import spikeinterface.preprocessing as spre
import spikeinterface.sorters as ss
import spikeinterface.core as sc
import spikeinterface.curation as scu
import spikeinterface.qualitymetrics as sqm
import spikeinterface.exporters as sexp
import spikeinterface.full as si
from helpers import pipeline_helpers
from spikeinterface.postprocessing import compute_principal_components
import torch
from spikeinterface.qualitymetrics import (compute_snrs, compute_firing_rates,
compute_isi_violations, calculate_pc_metrics,
compute_quality_metrics)
#set limit of 16 threads
from threadpoolctl import threadpool_limits
def spikeglx_preprocessing(recording):
# Preprocessing steps
logger.info(f'preprocessing recording')
# equivalent to what catgt does
recording = spre.phase_shift(recording)
# bandpass filter and common reference can be skipped if using kilosort as it does it internally
# but doesn't change anything to keep it
recording = spre.bandpass_filter(recording, freq_min=300, freq_max=6000)
recording = spre.common_reference(recording, reference='global', operator='median')
return recording
def spikesorting_pipeline(rec_name, params):
# Spikesorting pipeline for a single recording
working_directory = Path(params['working_directory']) / 'tempDir'
recording = se.read_spikeglx(rec_name, stream_id='imec0.ap')
recording = spikeglx_preprocessing(recording)
logger.info(f'running spike sorting')
sorting_output = ss.run_sorters(params['sorter_list'], [recording], working_folder=working_directory,
mode_if_folder_exists='keep',
engine='loop', sorter_params={'pykilosort': {'n_jobs': 19, 'chunk_size': 30000}},
verbose=True)
def spikesorting_postprocessing(params, step_one_complete=False):
jobs_kwargs = params['jobs_kwargs']
if step_one_complete == False:
sorting_output = ss.collect_sorting_outputs(Path(params['working_directory']))
for (rec_name, sorter_name), sorting in sorting_output.items():
logger.info(f'Postprocessing {rec_name} {sorter_name}')
if params['remove_dup_spikes']:
print('remove dup spikes')
logger.info(f'removing duplicate spikes')
sorting = scu.remove_duplicated_spikes(sorting, censored_period_ms=params['remove_dup_spikes_params'][
'censored_period_ms'])
sorting = scu.remove_excess_spikes(sorting, sorting._recording)
logger.info('waveform extraction')
outDir = Path(params['output_folder']) / rec_name / sorter_name
# we = sc.extract_waveforms(sorting._recording, sorting, outDir / 'waveforms_folder2', ms_before=1, ms_after=2., max_spikes_per_unit=300, n_jobs = -1, chunk_size=3000)
we = sc.extract_waveforms(sorting._recording, sorting, outDir / 'waveforms_folder_sparse3',
load_if_exists=True,
# overwrite=False,
ms_before=2,
ms_after=3.,
max_spikes_per_unit=300,
sparse=True,
num_spikes_for_sparsity=100,
method="radius",
radius_um=100,
**jobs_kwargs) # we = sc.load_waveforms(outDir / 'waveforms_sparse_folder')
logger.info(f'Computing quality metrics')
# with PCs
pca = compute_principal_components(we, n_components=3, mode='by_channel_local')
# logger.info('compute lratio')
#changed number of jobs to 1 as was runing out of space from parallel computation
metrics = compute_quality_metrics(we, metric_names=['d_prime'], n_jobs=8)
# metrics = compute_quality_metrics(we)
logger.info('Export report')
print('exporting report')
sexp.export_report(we, outDir / 'report2', format='png', force_computation=False, **jobs_kwargs)
#n_jobs = 8, chunk_size=3000
# logger.info(f'Exporting to phy')
# sexp.export_to_phy(we, outDir / 'phy5_folder', remove_if_exists=True,
# verbose=True,
# compute_pc_features=False,
# **jobs_kwargs)
# try:
# logger.info('Export report')
# sexp.export_report(we, outDir / 'report', padded_amplitudes_array, format='png', force_computation=False, use_padded_amplitudes=False, **jobs_kwargs)
# except Exception as e:
# logger.warning(f'Export report failed: {e}')
def main():
# parser = argparse.ArgumentParser()
##checking if torch device is available
with threadpool_limits(limits=16, user_api='blas'):
logger.info('loading torch')
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
logger.info(device)
params_file = Path('/nfs/nhome/live/carlag/neuropixels_sort/params/rat_params_hc4.json') # 'params/params.json
# parser.add_argument("params_file", help="path to the json file containing the parameters")
# args.params_file = params_file
# args = parser.parse_args()
with open(params_file) as json_file:
minified = jsmin(json_file.read()) # Parses out comments.
params = json.loads(minified)
logpath = Path(params['logpath'])
now = datetime.datetime.now().strftime('%d-%m-%Y_%H_%M_%S')
fh = logging.FileHandler(logpath / f'neuropixels_sorting_logs_{now}.log')
fh.setLevel(logging.DEBUG)
logger.addHandler(fh)
logger.info('Starting')
sorter_list = params['sorter_list'] # ['klusta'] #'kilosort2']
logger.info(f'sorter list: {sorter_list}')
if 'kilosort2' in sorter_list:
ss.Kilosort2Sorter.set_kilosort2_path(params['sorter_paths']['kilosort2_path'])
if 'waveclus' in sorter_list:
ss.WaveClusSorter.set_waveclus_path(params['sorter_paths']['waveclus_path'])
if 'kilosort3' in sorter_list:
ss.Kilosort3Sorter.set_kilosort3_path(params['sorter_paths']['kilosort3_path'])
datadir = Path(params['datadir'])
day_folder = datadir.name
rat_folder = datadir.parent.name
output_folder = Path(params['output_folder'])
#concatenate the rat_folder and day_folder to the output_folder
output_folder = str(output_folder)+ f'_{rat_folder}'+ f'_{day_folder}'
output_folder = Path(output_folder)
#check if output_folder exits
if output_folder.exists() == False:
# working_directory = Path(params['working_directory'])
logger.info('Start loading recordings')
sessions = [f.name for f in datadir.iterdir() if f.is_dir()]
print('sessions are:')
print(sessions)
recordings_list = []
# /!\ This assumes that all the recordings must have same mapping, pretty sure I was reading using the IBL cbin function after compressing the data
for session in sessions:
# Extract sync onsets and save as catgt would
# get_npix_sync(datadir / session, sync_trial_chan=[5])
logger.info(session)
print(session)
imec0_file = session + '_imec0'
try:
recording = se.read_spikeglx(datadir / session, stream_id='imec0.ap')
print(datadir / session/ imec0_file)
except Exception as e:
print(f'error loading recording: {e}, going to ibl file')
recording = se.read_cbin_ibl(datadir / session/ imec0_file)
recording = spikeglx_preprocessing(recording)
recordings_list.append(recording)
# except:
# print('issue preprocessing:'+ session)
multirecordings = sc.concatenate_recordings(recordings_list)
multirecordings = multirecordings.set_probe(recordings_list[0].get_probe())
#save the multirecordings
# logger.info('saving multirecordings')
# job_kwargs = dict(n_jobs=-1, chunk_duration="1s", progress_bar=True)
# multirecordings.save(folder = output_folder, **job_kwargs)
logger.info('sorting now')
sorting = ss.run_sorter(sorter_name="kilosort4", recording=multirecordings, output_folder=output_folder, batch_size = 60000, verbose=True)
pipeline_helpers.spikesorting_postprocessing(sorting, output_folder, datadir)
logger.info('Postprocessing done')
else:
logger.info('Output folder already exists, skipping sorting and trying postprocessing')
sorting = si.read_sorter_folder(output_folder)
pipeline_helpers.spikesorting_postprocessing(sorting, output_folder, datadir)
logger.info('Postprocessing done')
if __name__ == '__main__':
main()