nascell-automl

nascell-automl copied to clipboard

NijatZeynalov

Hi @wallarm I have tried tensorflow 1.4.0 and 1.5.0, however while running the train.py file I am getting this issue, ` Module not found error: No module named '_pywrap_tensorflow_internal' `

MkMushtaq

Why are you taking the output of the last time step to generate the neural net?

1

https://github.com/wallarm/nascell-automl/blob/18e5e59b94d56590f044be28c97cfa84886ec295/train.py#L46 Won't this return the output of the last time step alone? To generate the neural network, we need output from every time step right? What am I missing here?

TharunMohandoss

I think the loss in the code id wrong, how can you explain about it?

2

self.cross_entropy_loss = tf.nn.softmax_cross_entropy_with_logits(logits=self.logprobs[:, -1, :], labels=self.states) Why you use softmax_cross_entropy_with_logits here， the first state is "[10.0, 128.0, 1.0, 1.0]*args.max_layers"，so does the labels. The final output of RNN contributes to the...

moving-on

i didn't found the CNN on your link？ can u provided more detailed link？ thank u~

johsnows

grad*reward?is this rigorous? i used this code to apply to svhn dataset, and got very low accuracy, and sometimes the accuracy will accuracy will increase and then decrease, so i...

johsnows

Hi, I was going through your wonderful blog post of NAS with RL and realized there is a tiny bug within the codebase. The dropout rate extracted from `actions` is...

helloeve