bollossom / gac Goto Github PK

是否做过相关实验，将GAC添加到23年的spikformer中，

Hello author, thank you for your work. How do you count the number of spikes and neurons? Can you provide relevant code?

Hello author, very interested in your work, but I am tuning the code is the following error occurs, looking forward to your reply.
Files already downloaded and verified
Files already downloaded and verified
Traceback (most recent call last):
File "main.py", line 151, in
parallel_model = msresnet18(num_classes=100,T=6)
File "/home/GAC-main/CODE/CIFAR/models/MS_ResNet.py", line 167, in msresnet18
return _resnet('resnet18', BasicBlock_MS, [3, 3, 2],
File "/home/GAC-main/CODE/CIFAR/models/MS_ResNet.py", line 164, in _resnet
model = ResNet(block, layers, **kwargs)
TypeError: init() got an unexpected keyword argument 'T'

Hello authors, I'm really interested in your work and thank you for sharing your codes.
But I got some questions on reproducing results on Cifar10/100.

1. when I trained MsResNet18 model based on your instructions on github, the results on Cifar10/100 didn't even get close to the results on the AAAI24 paper.
2. And using GAC encoder didn't seem to have much improvement on the performance (setting self.GAC=False on Ms_ResNet.py didn't show much accuracy drop on Cifar10).
3. When I changed hyperparameter settings on the github code based on the AAAI24 paper appendix, LIF param (threshold, tau) = (1, 0.5) -> (0.5, 0.25), accuracy showed significant drop.
   Could you tell me what am I missing in this experiment? Thank you in advance!

Hello author, very interested in your work, but I am tuning the imagenet code is the following error occurs, looking forward to your reply.
Traceback (most recent call last):
File "train_amp.py", line 212, in
Traceback (most recent call last):
File "train_amp.py", line 212, in
criterion = TET_loss().cuda()
TypeError: TET_loss() missing 5 required positional arguments: 'outputs', 'labels', 'criterion', 'means', and 'lamb'
criterion = TET_loss().cuda()
TypeError: TET_loss() missing 5 required positional arguments: 'outputs', 'labels', 'criterion', 'means', and 'lamb'

用百度网盘里面给的权重跑test.py精度只有1%