search

D-X-Y / AutoDL-Projects

Automated deep learning algorithms implemented in PyTorch.
MIT License
1.57k stars 282 forks source link

DARTS1V-search-NASNet-space.sh how to interpret results #66

Closed abduallahmohamed closed 4 years ago

abduallahmohamed commented 4 years ago

Hi, sorry for asking too much

I ran bash -i ./scripts-search/DARTS1V-search-NASNet-space.sh cifar10 -1 and got the log and .pth file, is there a script I need to run so I can tell exactly what's the architecture I found?

Thanks

abduallahmohamed commented 4 years ago

I think I resolved it ! thanks

abduallahmohamed commented 4 years ago

I'm sorry about this but I do have hard time trying to draw the cell generated from the search, for instance: how would this looks like? ['|nor_conv_3x3~0|+|nor_conv_3x3~0|nor_conv_3x3~1|+|skip_connect~0|skip_connect~1|skip_connect~2|' ] as a cell? image does this looks right?

D-X-Y commented 4 years ago

For |nor_conv_3x3~0|+|nor_conv_3x3~0|nor_conv_3x3~1|+|skip_connect~0|skip_connect~1|skip_connect~2|, it will be parsed as follows:

node-1 = nor_conv_3x3(node-0, i.e., inputs)
node-2 = nor_conv_3x3(node-0) + nor_conv_3x3(node-1)
node-3 = node-0 + node-1 + node 2

Yeap, your figure looks correct.