ENAS-pytorch
ENAS-pytorch copied to clipboard
carpedm20
PyTorch implementation of "Efficient Neural Architecture Search via Parameters Sharing"
If want to run on multi gpu, when self.shared is forwarding , should use Modulelist's data (like self._w_h(which is a type of ModuleList)). Otherwise will raise an error :( RuntimeError:...
I don't think it can finish search architectures of CNN in 11.5 hours, because the parameters sharing is not as efficient as in RNN.
@dukebw Hi, thanks for your code. I have a detailed problem: In your code, you build embedding : ``` num_total_tokens = sum(self.num_tokens) self.encoder = torch.nn.Embedding(num_total_tokens, args.controller_hid) ``` This code shows...
Hello @carpedm20 , Thanks a lot for this nice implementation of the ENAS paper. Did you manage to reproduce their results by retraining the model from scratch? Thanks, Best
Is there any idea about how to deal with 1x1 convolution for channel alignment for parameter sharing CNN? Since the size changes as the graph change.
Otherwise, I will do it myself and check it in! Thanks for your time!
It is clear that controller falls into a local optimal while it can't find better actions from REINFORCE. I think unknown `c` of `c/valid ppl`, moving average baseline and temperature...
I encountered this strange error. Here is the output, thank you. Before, it was showing that the error cannot run on CPU and GPU at the same time, I added...
Bumps [scipy](https://github.com/scipy/scipy) from 1.2.1 to 1.10.0. Release notes Sourced from scipy's releases. SciPy 1.10.0 Release Notes SciPy 1.10.0 is the culmination of 6 months of hard work. It contains many...
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot] avatar"){kind=avatar}