fangwei123456 / spikingjelly

SpikingJelly is an open-source deep learning framework for Spiking Neural Network (SNN) based on PyTorch.
https://spikingjelly.readthedocs.io
Other
1.4k stars 246 forks source link

Problem with the evaluation mode #559

Open gwgknudayanga opened 5 months ago

gwgknudayanga commented 5 months ago

Hi,

In the training mode my LIF based network is learning properly. But when i set the net.eval() , the acccuracy is very low. After finishing the training, I checked the trained model for the evaluation set without setting net.eval(), In that case accuracy was good. But when i evaluated after setting net.eval() as the accuracy it gives a noisy very very low value. Do you have any suggestions or any idea about what is happening ? (i am running in multistep mode. i am doing the functional.reset_net() for each batch also)

Thank you.

Thanks and Rgds, Udayanga