icd-coding-benchmark
icd-coding-benchmark copied to clipboard
dalgu90
Automatic ICD coding benchmark based on the MIMIC dataset
As discussed, we are porting to a design specified in this article: https://huggingface.co/blog/transformers-design-philosophy. The aim is not only to follow the principle listed above, but we also intend to make...
I think we should have a folder inside `src/modules` called `layers`. Please see the example below to get a better idea: Consider the DCAN code (https://agit.ai/jsx/DCAN/src/branch/master/models.py). It has multiple "building...
Often, users don't want to train the model; they just want to load the pretrained weights (which we can provide - we can perhaps upload them on Google Drive) and...
Currently, we load splits from a file (which we take from CAML). Let's create classes for other kinds of splits - random splitting, for starters.
Again, all libraries have Unit Tests, but it's a pretty time-consuming process. So, maybe, we can ask potential future contributors to do this. For now, let's focus on adding approaches.