search

tejasph / care_nlp_psych

Cancer Ambulatory Resource Enhancement through early prediction of ER visits
MIT License
0 stars 0 forks source link

Creating a Compatible Pytorch Environment #14

Open tejasph opened 8 months ago

tejasph commented 8 months ago

Our environment will very much depend on pytorch, so basing dependencies around this is important.

https://pytorch.org/get-started/locally/

https://github.com/pytorch/pytorch#from-source

tejasph commented 8 months ago

First step, install pytorch, then install any additional dependencies needed using --freeze-installed

tejasph commented 8 months ago

conda install pytorch torchvision -c pytorch

tejasph commented 8 months ago

Final pytorch environment python = 3.10 pytorch =1.13.0 torchvision==0.14.0 torchaudio==0.13.0 torchtext== 0.14.0

tejasph commented 7 months ago

Found updated documenation for torchtext, may want to upgrade torchtext to a new version

https://github.com/pytorch/text

tejasph commented 7 months ago

New compatability list as part of my cnn_content branch:

name: care_nlp channels:

tejasph commented 7 months ago

Actually, the new data processing requires torchdata 0.6.0 which is compatible with pytorch 2.0. Going to have to adjust packages once again.

tejasph commented 7 months ago

new enviroment.yml core

name: care_nlp channels: