Open yuchenweng-private opened 11 months ago
Here is a part of my environment packages list. This is so wired that because I have tried to build MultiStepLIFNode before and it runs good with cupy backend.
Hi, please install SpikingJelly from GitHub (master version) and check if it solves this problem.
Thanks, Dr, Fang. Finally, after I install the master version from github, it works with module activation_based.neuron LIFNode, and it is so wired that then I reinstall the old version, everything is good also with MultiStepLIFneuron.
I have already tried to rebuild the whole environment with torch version 1.13.0 with cuda 11.6, but when I tried to build a MultiStepLIFNode with cupy backend, it still shows that I have not installed cupy.
I had tried everything I can do, but it did not work, my god. @fangwei123456
@QiWang233 Hi, you can set logging level as DEBUG (https://docs.python.org/3/library/logging.html#logging.DEBUG) and check the outputs.
Meanwhile, you can also try to install the latest version from GitHub. Note that the MultiStepLIFNode
is removed in the latest version. You can use LIFNode
with step_mode = 'm'
.
thanks for your reply! After changing the logging.debug and installing the latest version, I get questions below. It can be seen that it is because of the lack of "lava" , so I wonder how can I install the "lava" correctly.
Hi, lava is not easy to be installed... I suggest that you can change the source codes to avoid this problem (just ignore lava). Or you can install the latest version from github (we have fixed this problem).
Thanks for your patience!This is very important to me. But after changing to the lastest version 0.0.0.0.15,still the same problem.
import torch
import logging
logging.basicConfig(level=logging.DEBUG,
format='%(asctime)s - %(filename)s[line:%(lineno)d] - %(levelname)s: %(message)s')
try:
import cupy
except BaseException as e:
cupy = None
assert cupy is not None
print(cupy)
import spikingjelly
print(spikingjelly)
from spikingjelly.activation_based import neuron
lif = neuron.LIFNode(step_mode='m', backend='cupy').to('cuda:3')
x = torch.randn(10, 3, 224, 224).to('cuda:3')
out = lif(x)
print(out.shape)
I do not know why...
Set device as cuda:0
and try again?
Sorry, it does not work...it is so weird.
Thank you! After changing to a device with 3090(4090->3090), it is solved. Hope my experience can help the other! But I still do not know why 4090 have problem, maybe it is related to the "High Version".
I have also encountered the same problem and have been struggling for two days... This is the log: INFO:root:spikingjelly.clock_driven.lava_exchange: No module named 'lava.lib' INFO:root:spikingjelly.clock_driven.neuron: No module named 'tensorboard' INFO:root:spikingjelly.clock_driven.neuron: No module named 'lava.lib' Then I installed Lava and TensorBoard and used them normally, although the logs still showed: INFO:root:spikingjelly.clock_driven.lava_exchange: No module named 'lava.lib' INFO:root:spikingjelly.clock_driven.neuron: No module named 'lava.lib' But it no longer affects its use
Read before creating a new issue
For faster response
You can @ the corresponding developers for your issue. Here is the division:
Yanqi-Chen
Yanqi-Chen
Lyu6PosHao
lucifer2859
AllenYolk
Lyu6PosHao
DingJianhao
Yanqi-Chen
fangwei123456
We are glad to add new developers who are volunteering to help solve issues to the above table.
Issue type
SpikingJelly version
0.0.0.0.14
Description
..I have already tried to rebuild the whole environment with torch version 1.8.0 with cuda 11.1, but when I tried to build a MultiStepLIFNode with cupy backend, it still shows that I have not installed cupy.
Minimal code to reproduce the error/bug