Need for your kindly help with code problem

[GuokaiLiu]

Hello Udibr, Thanks for sharing your great work with the noisy_labels. I'm now doing a relevant research on the noisy data problem. Noisy-Labels-Problem-Collection and find your amazing code.

I have a question that how should I run your 161103, 161202 and 161230 Jupyter notebook code?

1

When I run it, the following comes out: [Errno 2] No such file or directory: 'data/160919-run-plot.results.pkl'

2

When I execute bash run_all.bash, the following comes out: name 'basestring' is not defined

3

How can I get the data folder and get the above three jupyter notebooks to run rightly?

---

I wish you could kindly help me since I'm doing my master's thesis and being in a puzzle about it for a long time.

Thank you very much! Looking forward to your reply : ）

[udibr]

for the basestring error I'm guessing you are using python3 and I was using python 2.7.

[udibr]

The file 160919-run-plot is created by running run.bash commands with --FN=160919-run-plot as one of the parameters. See the commented out lines in the 3rd cell of https://github.com/udibr/noisy_labels/blob/master/161103-run-plot.ipynb

[udibr]

Please let me know if you need any more help and if you get any results on your data set.

[GuokaiLiu]

Hi, Udibr, Sincerely thanks for your reply. I have changed to use python 2.7 today according to your insightful suggestion and get some data, here is the result. jupyter notebook .

Some errors might be caused by the packages. So could you please describe your configures? Such as:

1. Linux/Windows/MacOS?

2. tenforflow and keras version?

   And in order to get all the plot data in
   https://github.com/udibr/noisy_labels/blob/master/161103-run-plot.ipynb Should I run the code one by one?

# ./run.bash --FN=data/$1
# ./run.bash --FN=data/$1 --model=simple --beta=0
# ./run.bash --FN=data/$1 --model=complex --beta=0 --pretrain=2
# ./run.bash --FN=data/$1 --model=reed_hard --beta=0.8
# ./run.bash --FN=data/$1 --model=reed_soft --beta=0.95

Thank you very much! ： ） Best Regards!

[udibr]

Youre run looks great. Don’t see any error happening in it. If you run all the steps one by one you should get all graphs at the end

[GuokaiLiu]

Hello, Udibr, Thank you for your tireless answers. Final results come out of trial and error for me in these days. 161103-run-plot4

Here remains one problem to be solved. Due to the limited Nvidia GPU 1080Ti display memory, it seems that I can only run one seed of five to avoid out of memory. I think there may exist some tricks or settings to solve the problem.

Running Environment Configuration:

• Linux 16.04
• Anaconda 3 (create an environment with python 2.7)
• tensorflow=1.2.0
• keras=1.2.2
• theano=0.9
• keras.json: tensorflow→theano

Looking forward to your wonderful post-sharing. With all my best wishes! Guokai