TuningAI/Llama2_7B_Cover_letter_generator

Model Name: Llama2_7B_Cover_letter_generator

Description:

Llama2_7B_Cover_letter_generator is a powerful, custom language model that has been meticulously fine-tuned to excel at generating cover letters for various job positions. It serves as an invaluable tool for automating the creation of personalized cover letters, tailored to specific job descriptions.

Base Model:

This model is based on the Meta's meta-llama/Llama-2-7b-hf architecture, making it a highly capable foundation for generating human-like text responses.

Dataset :

This model was fine-tuned on a custom dataset meticulously curated with more than 200 unique examples. The dataset incorporates both manual entries and contributions from GPT3.5, GPT4, and Falcon 180B models.

Fine-tuning Techniques:

Fine-tuning was performed using QLoRA (Quantized LoRA), an extension of LoRA that introduces quantization for enhanced parameter efficiency. The model benefits from 4-bit NormalFloat (NF4) quantization and Double Quantization techniques, ensuring optimized performance.

Use Cases:

• Automating Cover Letter Creation: Llama2_7B_Cover_letter_generator can be used to rapidly generate cover letters for a wide range of job openings, saving time and effort for job seekers.

Performance:

• Llama2_7B_Cover_letter_generator exhibits impressive performance in generating context-aware cover letters with high coherence and relevance to job descriptions.
• It maintains a low perplexity score, indicating its ability to generate text that aligns well with user input and desired contexts.
• The model's quantization techniques enhance its efficiency without significantly compromising performance.

Limitations:

While the model excels in generating cover letters, it may occasionally produce text that requires minor post-processing for perfection.

• It may not fully capture highly specific or niche job requirements, and some manual customization might be necessary for certain applications.
• Llama2_7B_Cover_letter_generator's performance may vary depending on the complexity and uniqueness of the input prompts.
• Users should be mindful of potential biases in the generated content and perform appropriate reviews to ensure inclusivity and fairness.

Training procedure

The following bitsandbytes quantization config was used during training:

• load_in_8bit: False
• load_in_4bit: True
• llm_int8_threshold: 6.0
• llm_int8_skip_modules: None
• llm_int8_enable_fp32_cpu_offload: False
• llm_int8_has_fp16_weight: False
• bnb_4bit_quant_type: nf4
• bnb_4bit_use_double_quant: False
• bnb_4bit_compute_dtype: float16

Framework versions

• PEFT 0.4.0

How to Get Started with the Model

! huggingface-cli login

from transformers import pipeline
from transformers import AutoTokenizer
from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM , BitsAndBytesConfig
import torch

#config = PeftConfig.from_pretrained("ayoubkirouane/Llama2_13B_startup_hf")
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=getattr(torch, "float16"),
    bnb_4bit_use_double_quant=False)
model = AutoModelForCausalLM.from_pretrained(
        "meta-llama/Llama-2-7b-hf",
        quantization_config=bnb_config,
        device_map={"": 0})
model.config.use_cache = False
model.config.pretraining_tp = 1
model = PeftModel.from_pretrained(model, "TuningAI/Llama2_7B_Cover_letter_generator")
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf" , trust_remote_code=True)
tokenizer.pad_token = tokenizer.eos_token
tokenizer.padding_side = "right"
Instruction = "Given a user's information about the target job, you will generate a Cover letter for this job based on this information."
while 1:
  input_text = input(">>>")
  logging.set_verbosity(logging.CRITICAL)
  prompt = f"### Instruction\n{Instruction}.\n ###Input \n\n{input_text}. ### Output:"
  pipe = pipeline(task="text-generation", model=model, tokenizer=tokenizer,max_length=400)
  result = pipe(prompt)
  print(result[0]['generated_text'].replace(prompt, ''))