Covid19_Text_Model / README.md
MattStammers's picture
Update README.md
eb8217c
metadata
language: en
license: mit
model_id: Covid19_Text_Model
tags:
  - text-generation
developers: Matt Stammers
model_type: BERT
model_summary: This model looks to compare texts for relevance to Covid-19
shared_by: Matt Stammers
finetuned_from: https://thigm85.github.io/data/cord19/cord19-query-title-label.csv
repo: >-
  https://huggingface.co/MattStammers/Covid19_Text_Model?text=Comprehensive+overview+of+COVID-19.+Comprehensive+overview+of+Flu
paper: N/A
widget:
  - text: Comprehensive overview of COVID-19. Comprehensive overview of Flu
    example_title: Covid 19 Article Status. Label_0 = Covid-19 probability
    output:
      - label: Covid-19-article
        score: 0.6
      - label: Non-Covid-19-article
        score: 0.4
demo: >-
  https://huggingface.co/MattStammers/Covid19_Text_Model?text=Comprehensive+overview+of+COVID-19.+Comprehensive+overview+of+Flu
direct_use: Test it out here"
downstream_use: This is a standalone app
out_of_scope_use: >-
  The model will not work with any very complex sentences or to compare more
  than 3 statements
bias_risks_limitations: >-
  Biases inherent in the google BERT base also apply here. Should not be used
  for clinical tasks. This is a toy demonstration app only.
bias_recommendations: Do not be surprised if unusual results are obtained
get_started_code: |2-

      ``` python 
      # Use a pipeline as a high-level helper
          from transformers import pipeline

          pipe = pipeline("text-classification", model="MattStammers/Covid19_Text_Model")
      # Load model directly
          from transformers import AutoTokenizer, AutoModelForSequenceClassification

          tokenizer = AutoTokenizer.from_pretrained("MattStammers/MattStammers/Covid19_Text_Model")
          model = AutoModelForSequenceClassification.from_pretrained("MattStammers/Covid19_Text_Model")
      ```
                          
training_data: https://thigm85.github.io/data/cord19/cord19-query-title-label.csv
preprocessing: Sentence Pairs to analyse similarity
training_regime: User Defined
speeds_sizes_times: Not Relevant
metrics: Not Given
pipeline_tag: text-classification

This is a basic inference BERT model which has been fine-tuned to discriminate between covid19 and non-covid-19 relevant texts.

Unlike past models I have created this one raw and uploaded it as a standalone git repo to experiment with upload options. Not as streamlined as using the Huggingface card generation system but definitely simpler to do.

This is also my first experiment with ONNX.

Training data can be obtained as follows:

import pandas as pd

training_data = pd.read_csv("https://thigm85.github.io/data/cord19/cord19-query-title-label.csv")
training_data.head()

Please do not use this for any clinical/applied purpose. It is a toy app only.