search

BrainCog-X / Brain-Cog

Brain-inspired Cognitive Intelligence Engine (BrainCog) is a brain-inspired spiking neural network based platform for Brain-inspired Artificial Intelligence and simulating brains at multiple scales. The long term goal of BrainCog is to provide a comprehensive theory and system to decode the mechanisms and principles of human intelligence and its evolution, and develop artificial brains for brain-inspired conscious living AI in future Human-AI symbiotic Society.
http://www.brain-cog.network/
Apache License 2.0
414 stars 65 forks source link

What typical GPU is used? #145

Open lonnietc opened 11 months ago

lonnietc commented 11 months ago

Hello,

Can you please tell me what is the typical GPU cards that you are using and their specification?

On this machine, I have a RTX 3060 Ti with 8GB memory but it does not seem to be enough to run your Unsupervised STDP-based spiking neural network and I keep running out of memory on the GPU.

https://github.com/BrainCog-X/Brain-Cog/tree/main/examples/Perception_and_Learning/UnsupervisedSTDP

I am working my way through the wonder Brain Cog code and examples. This project show the most promise that I have found for truly developing a full virtual brain simulation that will hopefully be able to learn incrementally and online. I have investigated MANY other solutions and one that had good potential was the SOINN by Hasegawa that uses unsupervised learning to incrementally learn.

Also, I am wondering if you have considered adding Liquid time-constant Networks (LTCs) which are supposed to be a significant improvement over classic RNN's in that they evolve and self-adjust to optimal solutions.

https://github.com/raminmh/liquid_time_constant_networks

PAPER: https://arxiv.org/abs/2006.04439

Thanks and have a great day

sunbaby01 commented 11 months ago

Thank you for your interest in our work and for your keen attention to BrainCog. If you exceed the maximum memory capacity of GPU, you can choose to reduce the time 'T' of the linear layer or decrease the batch size to an appropriate range. Additionally, regarding LTC, we will look into it when we have some spare time.