ADRv1 / README.md
paragon-analytics's picture
Update README.md
f8ddae9
|
raw
history blame
1.08 kB
metadata
license: mit
widget:
  - text: >-
      Took the pill, 12 hours later my muscles started to really hurt, then my
      ribs started to burn so bad I couldn't breath.

This model takes text (narrative of reasctions to medications) as input and returns a predicted severity score for the reaction (LABEL_1 is severe reaction). Please do NOT use for medical diagnosis. Example usage:

import torch
import tensorflow as tf
from transformers import RobertaTokenizer, RobertaModel
from transformers import AutoModelForSequenceClassification
from transformers import TFAutoModelForSequenceClassification
from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("paragon-analytics/ADRv1")  
model = AutoModelForSequenceClassification.from_pretrained("paragon-analytics/ADRv1")

def adr_predict(x):
    encoded_input = tokenizer(x, return_tensors='pt')
    output = model(**encoded_input)
    scores = output[0][0].detach().numpy()
    scores = tf.nn.softmax(scores)
    return scores.numpy()[1]

sentence = "I have severe pain."

adr_predict(sentence)