clinical-adapters/

n2c2-pfeiffer-adapter-bert-ast

from transformers import AutoTokenizer
from adapters import AutoAdapterModel

model_adapter = "clinical-adapters/n2c2-pfieffer-adapter-bert-ast"
pretrained_model_name_or_path = "bert-base-uncased"
tokenizer = AutoTokenizer.from_pretrained(pretrained_model_name_or_path, model_max_length=150)
special_tokens_dict = {"additional_special_tokens": ["[entity]"]}
num_added_toks = tokenizer.add_special_tokens(special_tokens_dict,False)
model = AutoAdapterModel.from_pretrained(pretrained_model_name_or_path = pretrained_model_name_or_path,
num_labels=3 , id2label = {0: 'PRESENT', 1: 'ABSENT', 2:'POSSIBLE'})
model.resize_token_embeddings(len(tokenizer))
ast = model.load_adapter(model_adapter,with_head=True)
model.active_adapters = ast

import torch

id2label = {0: 'PRESENT', 1: 'ABSENT', 2:'POSSIBLE'}
sentence = [ "Patient denies [entity] SOB [entity]",
"Patient do not have [entity] fever [entity]",
"had [entity] abnormal ett [entity] and referred for cath",
"The patient recovered during the night and now denies any [entity] shortness of breath [entity].",
"Patient with [entity] severe fever [entity].",
"Patient should abstain from [entity] painkillers [entity]"]
model.to('cpu')
for s in sentence :
tokenized_input = tokenizer(s, return_tensors="pt", padding=True)
outputs = model(**tokenized_input)
predicted_labels = torch.argmax(outputs.logits, dim=1)
print(id2label[predicted_labels.item()])