xyzedd / cssr Goto Github PK

The value of hidden layer size is not specified in the config files. With 0 hidden layers, the stack of encoder and decoder in an autoencoder is not created. I ran the code but found that input is bypassing the encoder section of autoencoder and directly passing to latent_conv.

Thanks for your great work.

Could you provide detailed configured for experiments on tinyimagenet？

I ran the command "python main.py --gpu 0 --ds exps/tinyimagenet/spl_a.json --config configs/rcssr/tinyimagenet.json --method cssr_ft". However, the classification accuracy on the closed set is only about 0.20.

Could you help with it?

Is there a timeline for this project? I'm really interested in the specific design and implementation! thx