Question about speed and memory?

[SunGaofeng]

Thanks to your great week on this repo, I have some questions about speed and memory of this model, 1, Do you use a single V100 to train this model? How long does it take to complete on the two dataset, moving mnist and KTH actions? 2, In your default settings, batch_size = 4, why use such a small batch_size? Is it because that the model is too large? What's the memory of your V100, 32G or 16G?

[GuptaSonam]

Hi SunGaofeng,

Did you try running the code? For me, training for MNIST dataset on Ti gtx 1080 (12 gb) takes 1 week using a single gpu. Not sure if I am doing something wrong or the model is supposed to be that slow.