Model Card for CoverGenie

<!-- Provide a quick summary of what the model is/does. -->

The goal of this project is to build a fine-grained mini-ChatGPT (named “CoverGenie”) , which is designed to generate resumes and cover letters based on job descriptions from the tech field. By nature,it is a language generation task, and it takes the job description as input to a sequence of text and turns it into a structured, certain style of resumes and cover letters. This might involve parameter efficient finetuning, reinforcement learning and prompting engineering to some extent.

Model Details

Model Description

<!-- Provide a longer summary of what this model is. -->

• Model type: T5 (Text-to-Text-Transfer-Transformer)
• Language(s) (NLP): [More Information Needed]
• License: Apache-2.0
• Finetuned from model: FlanT5 Large

Model Sources [optional]

<!-- Provide the basic links for the model. -->

• Repository: [More Information Needed]
• Paper [optional]: https://arxiv.org/pdf/2210.11416.pdf

Uses

It Can Generate Cover letter if we are able to input the Job description and Resume of a candidate.

How to Get Started with the Model

Use the code below to get started with the model.

<details> <summary> Click to expand </summary>

from transformers import GenerationConfig
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
from transformers import GenerationConfig
import nltk
nltk.download('punkt')
max_source_length=512 
tokenizer = AutoTokenizer.from_pretrained("Hariharavarshan/Cover_genie")
model = AutoModelForSeq2SeqLM.from_pretrained("Hariharavarshan/Cover_genie")
JD='''<Job description Text>'''
resume_text= '''<Resume Text>'''
final_text="give me a cover letter based on the a job description and a resume. Job description:"+JD +" Resume:"+ resume_text
generation_config = GenerationConfig.from_pretrained("google/flan-t5-large",temperature=2.0)
inputs = tokenizer(final_text, max_length=max_source_length, truncation=True, return_tensors="pt")
output = model.generate(**inputs, num_beams=3, do_sample=True, min_length=1000,
                               max_length=10000,generation_config=generation_config,num_return_sequences=3)
decoded_output = tokenizer.batch_decode(output, skip_special_tokens=True)[0]
generated_Coverletter = nltk.sent_tokenize(decoded_output.strip())

Developed by: Hariharavarshan,Jayathilaga,Sara,Meiyu