Alexander Chebykin comments

Results 14 comments of


                                            Alexander Chebykin

code for Section 3.2 and Figure 4

Hey @zachluo, I'm also looking for that code, were you able to find it in some other repository?

Is it actually reasonable to use single-class accuracy?

To explain my concern more clearly in an example: Accuracy for class 1 includes as a term True negative rate for classess 2..365. Suppose we remove a neuron that doesn't...

Change Shortcut for Adding Bookmark

@ardcore On Windows this works perfectly. Also I would suggest for you to add information about changing binding to the extension page in the Chrome Store, or at least to...

How to reproduce the results on CIFAR10

The code seems to have not been updated to fix this. The main problem is not loading the pretrained weights (thanks @zhangshuaitao for mentioning that), which can be fixed by...

Cityscapes experiment

Sorry for an off-topic question, but I have trouble even running the training on CityScapes: for 256x512 input I get 32x64 output, while the target is 256x512. And smaller output...

Torch source change: repo incompatible with versions 0.5.0 and 1.0.0

@cyizhuo that worked perfectly, thank you so much!

Pre-trained supernet weights release

Thanks for the update, looking forward to it

Cannot import name 'gfile' from 'tensorflow'

The Colab notebook linked from the Feature Visualization Distill article also fails now with this error. Easy fix is to add ``` !pip install tensorflow-gpu==1.15 ```

Ways to free memory?

Thanks for the pointer, I've set only lexical preservation to false, since all other options seem to impact comments, which I need.

Is there any plan on when to release the train and search code?

I'd also be interested in the train & search code - maybe the authors could mention the approximate timeline of the code release?