Parameter 'nt0' causes problems

[DradeAW]

Hi,

I'm using pykilosort on extracellular cerebellar recordings, and I'm interested in complex spikes (which are spikes that can last for 3ms and sometimes more).
So I tried upping the nt0 parameter, and it worked for 61 (default) and 64, but failed for 70, 80 and 81:

13:10:24.531 [I] utils:334            Starting step learn.
Extracting templates: 100%|███████████████████████████████████████████████████████████████████████████████| 181/181 [00:08<00:00, 20.67it/s]
Optimizing templates:   0%|                                                                                        | 0/1810 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "test.py", line 36, in <module>
    kilosort.run(dat_path, probe=probe, dir_path=dir_path, params=params, n_channels=64, dtype=np.int16, sample_rate=30000)
  File "/users/nsr/wyngaard/.conda/envs/kilosort/lib/python3.7/site-packages/pykilosort-2.0.0a0-py3.7.egg/pykilosort/main.py", line 188, in run
    out = learnAndSolve8b(ctx)
  File "/users/nsr/wyngaard/.conda/envs/kilosort/lib/python3.7/site-packages/pykilosort-2.0.0a0-py3.7.egg/pykilosort/learn.py", line 737, in learnAndSolve8b
    dWU, cmap = mexGetSpikes2(Params, dataRAW, wTEMP, iC)
  File "/users/nsr/wyngaard/.conda/envs/kilosort/lib/python3.7/site-packages/pykilosort-2.0.0a0-py3.7.egg/pykilosort/learn.py", line 269, in mexGetSpikes2
    extract_snips((Nchan,), tpS, (d_Params, d_st1, d_id1, d_counter, d_data, d_WU))
  File "cupy/core/raw.pyx", line 52, in cupy.core.raw.RawKernel.__call__
  File "cupy/cuda/function.pyx", line 149, in cupy.cuda.function.Function.__call__
  File "cupy/cuda/function.pyx", line 131, in cupy.cuda.function._launch
  File "cupy/cuda/driver.pyx", line 229, in cupy.cuda.driver.launchKernel
  File "cupy/cuda/driver.pyx", line 82, in cupy.cuda.driver.check_status
cupy.cuda.driver.CUDADriverError: CUDA_ERROR_INVALID_VALUE: invalid argument

Now I know that pykilosort is still in development, and that the nt0 parameter might not be a priority.
I just thought I'd point it out in case you didn't catch that bug.

Also, the matlab version says that this parameter cannot be greater than 81. Will it be possible for the python version to go above that in the future?

Thank you,
Hope this helps