borrepp
commented
6 months ago If I changed the conv_engine to "numpy", I go the following error:
ValueError Traceback (most recent call last) Cell In[47], line 18 13 peaks = np.load(os.path.join(motion_folder, 'peaks.npy')) 15 peak_localization = np.load(os.path.join(motion_folder, 'peak_locations.npy')) ---> 18 motion, temporal_bins, spatial_bins = estimate_motion(recording=rec_filter_file, 19 peaks=peaks, 20 peak_locations=peak_localization, 21 method="decentralized", 22 pairwise_displacement_method = 'conv', 23 conv_engine="numpy", 24 batch_size = 1, 25 progress_bar = True, 26 verbose = True) 28 print(motion, temporal_bins, spatial_bins)
File m:\Monkey_Python\SIenv\lib\site-packages\spikeinterface\sortingcomponents\motion_estimation.py:152, in estimate_motion(recording, peaks, peak_locations, direction, bin_duration_s, bin_um, margin_um, rigid, win_shape, win_step_um, win_sigma_um, post_clean, speed_threshold, sigma_smooth_s, method, output_extra_check, progress_bar, upsample_to_histogram_bin, verbose, method_kwargs) 150 # run method 151 method_class = estimate_motion_methods[method] --> 152 motion, temporal_bins = method_class.run( 153 recording, 154 peaks, 155 peak_locations, 156 direction, 157 bin_duration_s, 158 bin_um, 159 spatial_bin_edges, 160 non_rigid_windows, 161 verbose, 162 progress_bar, 163 extra_check, 164 method_kwargs, 165 ) 167 # replace nan by zeros 168 motion[np.isnan(motion)] = 0
File m:\Monkey_Python\SIenv\lib\site-packages\spikeinterface\sortingcomponents\motion_estimation.py:363, in DecentralizedRegistration.run(cls, recording, peaks, peak_locations, direction, bin_duration_s, bin_um, spatial_bin_edges, non_rigid_windows, verbose, progress_bar, extra_check, histogram_depth_smooth_um, histogram_time_smooth_s, pairwise_displacement_method, max_displacement_um, weight_scale, error_sigma, conv_engine, torch_device, batch_size, corr_threshold, time_horizon_s, convergence_method, soft_weights, normalized_xcorr, centered_xcorr, temporal_prior, spatial_prior, force_spatial_median_continuity, reference_displacement, reference_displacement_time_s, robust_regression_sigma, lsqr_robust_n_iter, weight_with_amplitude) 360 if verbose: 361 print(f"Computing pairwise displacement: {i + 1} / {len(non_rigid_windows)}") --> 363 pairwise_displacement, pairwise_displacement_weight = compute_pairwise_displacement( 364 motion_histogram[:, window_slice], 365 bin_um, 366 window=win[window_slice], 367 method=pairwise_displacement_method, 368 weight_scale=weight_scale, 369 error_sigma=error_sigma, 370 conv_engine=conv_engine, 371 torch_device=torch_device, 372 batch_size=batch_size, 373 max_displacement_um=max_displacement_um, 374 normalized_xcorr=normalized_xcorr, 375 centered_xcorr=centered_xcorr, 376 corr_threshold=corr_threshold, 377 time_horizon_s=time_horizon_s, 378 bin_duration_s=bin_duration_s, 379 progress_bar=False, 380 ) 382 if spatial_prior: 383 all_pairwise_displacements[i] = pairwise_displacement
File m:\Monkey_Python\SIenv\lib\site-packages\spikeinterface\sortingcomponents\motion_estimation.py:865, in compute_pairwise_displacement(motion_hist, bin_um, method, weight_scale, error_sigma, conv_engine, torch_device, batch_size, max_displacement_um, corr_threshold, time_horizon_s, normalized_xcorr, centered_xcorr, bin_duration_s, progress_bar, window) 862 correlation = np.empty((size, size), dtype=motion_hist.dtype) 864 for i in xrange(0, size, batch_size): --> 865 corr = normxcorr1d( 866 motion_hist_engine, 867 motion_hist_engine[i : i + batch_size], 868 weights=window_engine, 869 padding=possible_displacement.size // 2, 870 conv_engine=conv_engine, 871 normalized=normalized_xcorr, 872 centered=centered_xcorr, 873 ) 874 if conv_engine == "torch": 875 max_corr, best_disp_inds = torch.max(corr, dim=2)
File m:\Monkey_Python\SIenv\lib\site-packages\spikeinterface\sortingcomponents\motion_estimation.py:1415, in normxcorr1d(template, x, weights, centered, normalized, padding, conv_engine) 1407 wt = template[:, None, :] * weights 1409 # conv1d valid rule: 1410 # (B,1,L),(O,1,L)->(B,O,L) 1411 1412 # compute expectations 1413 # how many points in each window? seems necessary to normalize 1414 # for numerical stability. -> 1415 N = conv1d(ones, weights, padding=padding) 1416 if centered: 1417 Et = conv1d(ones, wt, padding=padding)
File m:\Monkey_Python\SIenv\lib\site-packages\spikeinterface\sortingcomponents\motion_estimation.py:1476, in scipy_conv1d(input, weights, padding) 1474 for m in range(n): 1475 for c in range(c_out): -> 1476 output[m, c] = correlate(input[m, 0], weights[c, 0], mode=mode) 1478 return output
File m:\Monkey_Python\SIenv\lib\site-packages\scipy\signal_signaltools.py:241, in correlate(in1, in2, mode, method) 239 # this either calls fftconvolve or this function with method=='direct' 240 if method in ('fft', 'auto'): --> 241 return convolve(in1, _reverse_and_conj(in2), mode, method) 243 elif method == 'direct': 244 # fastpath to faster numpy.correlate for 1d inputs when possible 245 if _np_conv_ok(in1, in2, mode):
File m:\Monkey_Python\SIenv\lib\site-packages\scipy\signal_signaltools.py:1428, in convolve(in1, in2, mode, method) 1425 elif method == 'direct': 1426 # fastpath to faster numpy.convolve for 1d inputs when possible 1427 if _np_conv_ok(volume, kernel, mode): -> 1428 return np.convolve(volume, kernel, mode) 1430 return correlate(volume, _reverse_and_conj(kernel), mode, 'direct') 1431 else:
File <__array_function__ internals>:200, in convolve(*args, **kwargs)
File m:\Monkey_Python\SIenv\lib\site-packages\numpy\core\numeric.py:848, in convolve(a, v, mode) 846 a, v = v, a 847 if len(a) == 0: --> 848 raise ValueError('a cannot be empty') 849 if len(v) == 0: 850 raise ValueError('v cannot be empty')
ValueError: a cannot be empty
Hello, I have a 32-channel recording with the following specs:
I have saved the filtered recording and now I'm trying to run "correc_motion" with the preset="nonrigid_accurate", but I got the following error:
I'm running on Windows 11, with spikeInterface version = 0.100.0