search

kohpangwei / influence-release

MIT License
776 stars 175 forks source link

Local issue running Hospital Readmission notebook #23

Closed bricha2 closed 3 years ago

bricha2 commented 3 years ago

I have been working on replicating the hospital readmission notebook locally and I am getting output that differs substantially from the expected output. Below, I show side-by-side my output vs the expected output. I think the major issue is that the number of parameters that my model outputs is 122 when it's supposed to be 127. I can't localize where this change is occurring because the dimensions of my input data are as expected: 127 columns. Any guidance you can provide on why this might be occurring would be greatly appreciated!

Screen Shot 2021-06-10 at 7 07 29 PM Screen Shot 2021-06-10 at 7 07 53 PM
kohpangwei commented 3 years ago

Hi Brianna,

Hmm, I'm not sure why the output dimensions are changing. If you run the hospital notebook in this repo without any changes (https://github.com/kohpangwei/influence-release/blob/master/scripts/hospital_readmission.ipynb), do you get the expected number of parameters? If you don't, then it could perhaps be due to version changes (since this code is quite old)?

Is your input_dim 127?

bricha2 commented 3 years ago

Hello! Yeah, I did have issues running this repo as is. The docker installation was too outdated for my system, it seems. I got the correct input_dim, but the model still ended up with a different number of parameters (122). Even when I change the dataset (to the adult dataset), I am still getting the same issue where the original dimensions (97) are somehow converted to 122 in the model. I wasn't able to figure out the cause of this, but I assume it was a version issue. Thanks for your time and assistance, anyhow!

kohpangwei commented 3 years ago

Ok. Thank you for raising this issue!