README.md

metadata

{}

Relation Extraction model for KBQA

This is the fine-tuned version of TinyLlama for the Relation Extraction task.

Model Details

The model is trained on the Dataset, which consists of questions and their related information, such as entities and relations. The relationships in the data are annotated from the Freebase dataset. The relationships in the data are annotated from the Freebase dataset.

How to use

You will need the transformers>=4.34 Do check the TinyLlama github page for more information.

Direct Use

import os
import torch
#from datasets import load_dataset
from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    BitsAndBytesConfig,
    HfArgumentParser,
    TrainingArguments,
    pipeline,
    logging,
)

from transformers import AutoTokenizer, pipeline, logging
model_name_or_path = "dice-research/Ft_TinnyLlama_QA_RE"
model_basename = "model"

use_triton = False

tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
tokenizer.pad_token = tokenizer.eos_token
model = AutoModelForCausalLM.from_pretrained(
        model_name_or_path, quantization_config=None, device_map="auto"
    )

pipe = pipeline("text-generation", model=model,tokenizer=tokenizer, torch_dtype=torch.bfloat16, device_map="auto")

def prompt_REQA(question): 
    messages = [
    {"role": "user", "content": question},
    ]
    
    prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    #response=pipe(sparql_prompt)
    response = pipe(prompt,max_new_tokens=20, do_sample=True, temperature=0.6, top_k=5, top_p=0.95)[0]['generated_text']
    return response.split('<|assistant|>\n')[1]

prompt_REQA("how many electronic arts games are available for sale in the united states of america?")