LLIMONIIE: Large Language Instructed Model for Open Named Italian Information Extraction

LLIMONIE generalizes IE across diverse domains without requiring input ontologies.

  • Perform three tasks:
    • Open Named Entity Recognition

    • Open Relation Extraction

    • Joint Entity and Relation Extraction

πŸ’» Quick Start

Setup conda environment

Install the unsloth package following the repo guide

Clone the repository

git clone https://github.com/leonardoPiano/LLIMONIE.git

Run the generation

from PromptTemplates.instruct_prompt_templates import  NER,RE,JOINT
from LLM.Unsloth import UnslothLLM
model_path="leopiano98/LLIMONIIE_llama3-8b"

llimonie=UnslothLLM(model_path,inference=True)
task=NER
text="Alessandro Manzoni Γ¨ considerato uno dei maggiori romanzieri italiani di tutti i tempi per il suo celebre romanzo I promessi sposi"
messages = [{"role": "system", "content": task},
                        {"role": "user", "content": text}]
output= llimonie.generate(messages, max_new_tokens=512)
#output: Alessandro Manzoni[Writer|Person]; I promessi sposi[Novel|Book]; italiani[Nationality|Ethnicity] 
  • Developed by: leopiano98
  • License: apache-2.0
  • Finetuned from model : unsloth/llama-3-8b-instruct-bnb-4bit

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

Downloads last month
29
Safetensors
Model size
8.03B params
Tensor type
BF16
Β·
Inference API
Unable to determine this model's library. Check the docs .

Model tree for leopiano98/LLIMONIIE_llama3-8b

Quantizations
2 models