Why model storage memory takes a huge amount of space?

quark0 / darts

Differentiable architecture search for convolutional and recurrent networks

https://arxiv.org/abs/1806.09055

Apache License 2.0

3.92k stars 843 forks source link

Why model storage memory takes a huge amount of space? #100

Open haithanhp opened 5 years ago

haithanhp commented 5 years ago

Hi,

I see your saved models for imageNet are very huge (48.2MB) while in the code they are 4.9MB (#parameters). I also double checked MobileNet v2 only 14.02MB on Pytorch.

Could you explain why your model is very huge? It should be small in memory storage.

alphadl commented 5 years ago

Personally, 4.9M means the number of 4.9 million learnable parameters rather than storage.

haithanhp commented 5 years ago

Thanks for your information.

Do you know why the models in pytorch are very huge (48.2MB)? I see same scale models (Mobilenet v1, v2) with 4.4M and 3.9M (#parameters) only take little space (14.02MB or 16MB).