search

SRKH / S4NN

Temporal backpropagation for spiking neural networks with one spike per neuron, by S. R. Kheradpisheh and T. Masquelier, International Journal of Neural Systems (2020), doi: 10.1142/S0129065720500276
https://www.worldscientific.com/doi/10.1142/S0129065720500276
GNU General Public License v3.0
59 stars 19 forks source link

Not able to load weights_pretrained.npy #3

Open JimBartels opened 2 years ago

JimBartels commented 2 years ago

When I try to load the pretrained weights I get the following error under cupy_cuda11x:

`--------------------------------------------------------------------------- ModuleNotFoundError Traceback (most recent call last) in 109 if loading: 110 if GPU: --> 111 W = np.load(LoadFrom, allow_pickle=True) 112 else: 113 for i in range(len(W)):

1 frames /usr/local/lib/python3.7/dist-packages/numpy/lib/format.py in read_array(fp, allow_pickle, pickle_kwargs) 746 pickle_kwargs = {} 747 try: --> 748 array = pickle.load(fp, **pickle_kwargs) 749 except UnicodeError as err: 750 # Friendlier error message

ModuleNotFoundError: No module named 'cupy.core'

---------------------------------------------------------------------------`

Is there any way that you could upload the weights without requiring cupy? For example by the way that has been done in the BS4NN github, splitting the weights up into W0 and W1.

Thanks, Jim Bartels

hamidehmoghadasi commented 1 year ago

you do not activate GPU, or it is finished