I'm trying to run pySYD on a few Kepler targets, but when running with the --mc 200 option the sampler appears to get stuck at during the sampling, while for other stars it runs flawlessly. It doesn't crash, just hangs indefinitely.
The initial parts of the run seem to work fine and output is produced, it only hangs when it gets to the sampling.
Running on the same target multiple times, the sampler appears to get stuck at different times.
It's not entirely clear to me how the sampling is done, but I know in EMCEE something like this might occur if the sampler gets stuck beyond the bounds of a fit. Perhaps something like that is happening here?
It doesn't appear to be a memory issue, since the ones that get stuck appear to be shorter time series.
I used the pip installable version of pySYD.
Examples:
pysyd run --star KIC4646780 --info info/star_info.csv --mc 200 -vto
Target: KIC4646780
LIGHT CURVE: 43076 lines of data read
Time series cadence: 59 seconds
POWER SPECTRUM: 21677 lines of data read
PS is critically-sampled
PS resolution: 0.391961 muHz
Estimating numax:
PS binned to 195 datapoints
Numax estimate 1: 1040.94 +/- 18.96
S/N: 2.42
Numax estimate 2: 1085.80 +/- 35.76
S/N: 2.39
Numax estimate 3: 1179.34 +/- 61.26
S/N: 1.50
Selecting model 1
Determining background model:
PS binned to 424 data points
Comparing 6 different models:
Model 0: 0 Harvey-like component(s) + white noise fixed
Model 1: 0 Harvey-like component(s) + white noise term
Model 2: 1 Harvey-like component(s) + white noise fixed
Model 3: 1 Harvey-like component(s) + white noise term
Model 4: 2 Harvey-like component(s) + white noise fixed
Model 5: 2 Harvey-like component(s) + white noise term
Based on BIC statistic: model 4
background-corrected PS saved
Running sampling routine:
12%|█▎ | 24/200 [00:02<00:19, 8.95it/s]
18%|█▊ | 36/200 [00:04<00:19, 8.50it/s]
pysyd run --star KIC1435467 --info info/star_info.csv --mc 200 -vto
Target: KIC1435467
LIGHT CURVE: 486239 lines of data read
Time series cadence: 59 seconds
POWER SPECTRUM: 247854 lines of data read
PS is critically-sampled
PS resolution: 0.034280 muHz
Estimating numax:
PS binned to 195 datapoints
Numax estimate 1: 1445.36 +/- 85.66
S/N: 8.81
Numax estimate 2: 1455.28 +/- 93.08
S/N: 30.95
Numax estimate 3: 1481.68 +/- 103.31
S/N: 100.00
Selecting model 3
Determining background model:
PS binned to 424 data points
Comparing 6 different models:
Model 0: 0 Harvey-like component(s) + white noise fixed
Model 1: 0 Harvey-like component(s) + white noise term
Model 2: 1 Harvey-like component(s) + white noise fixed
Model 3: 1 Harvey-like component(s) + white noise term
Model 4: 2 Harvey-like component(s) + white noise fixed
Model 5: 2 Harvey-like component(s) + white noise term
Based on BIC statistic: model 4
background-corrected PS saved
Running sampling routine:
11%|█ | 22/200 [00:16<02:26, 1.22it/s]
21%|██ | 42/200 [00:33<02:10, 1.21it/s]
31%|███ | 62/200 [00:49<01:55, 1.21it/s]
40%|████ | 81/200 [01:05<01:39, 1.20it/s]
50%|████▉ | 99/200 [01:21<01:23, 1.22it/s]
58%|█████▊ | 116/200 [01:35<01:09, 1.21it/s]
67%|██████▋ | 133/200 [01:49<00:55, 1.21it/s]
75%|███████▌ | 150/200 [02:03<00:41, 1.20it/s]
84%|█████�██▎ | 166/200 [02:16<00:28, 1.21it/s]
92%|████��
99%|█████████▉| 198/200 [02:42<00:02, 1.21it/s]
100%|██████████| 200/200 [02:44<00:00, 1.21it/s]
Output parameters:
tau_1: 322.37 +/- 22.77 s
sigma_1: 53.03 +/- 2.68 ppm
tau_2: 112.21 +/- 12.13 s
sigma_2: 45.31 +/- 3.47 ppm
numax_smooth: 1388.91 +/- 11.23 muHz
A_smooth: 0.75 +/- 0.04 ppm^2/muHz
numax_gauss: 1425.56 +/- 11.67 muHz
A_gauss: 0.68 +/- 0.03 ppm^2/muHz
FWHM: 250.91 +/- 13.64 muHz
dnu: 70.48 +/- 0.11 muHz
TESTING INFORMATION:
362/424 points are being used for background fit
----------------------------------------------
------------- MODEL COMPARISONS --------------
----------------------------------------------
Model 0: 0 Harvey-like component(s) + white noise fixed
BIC = 408540.44 | AIC = 1128.56
Model 1: 0 Harvey-like component(s) + white noise term
BIC = 408502.55 | AIC = 1128.45
Model 2: 1 Harvey-like component(s) + white noise fixed
BIC = 402832.79 | AIC = 1112.78
Model 3: 1 Harvey-like component(s) + white noise term
BIC = 402819.23 | AIC = 1112.73
Model 4: 2 Harvey-like component(s) + white noise fixed
BIC = 402375.08 | AIC = 1111.49
Model 5: 2 Harvey-like component(s) + white noise term
BIC = 402388.49 | AIC = 1111.52
The power spectrum mask includes 36370 data points
i.e. the power excess region ~[858.33,2105.04]
D method to estimate dnu ~= 70.34
Median of ED: 51.64
Clip value of ED: 206.58
M method to estimate dnu ~= 70.31
A method to estimate dnu ~= 70.34
D method to estimate dnu ~= 70.34
- combining results into single csv file