File size: 4,381 Bytes
7ac06e8 bfb338f 7ac06e8 bfb338f cc384cb 09526dd bfb338f 7ac06e8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 |
---
library_name: peft
base_model: meta-llama/Meta-Llama-3-8B-Instruct
license: apache-2.0
language:
- en
---
## Overview
The model is a LoRa Adaptor based on LLaMA-3-8B-Instruct. The model has been trained on a [re-annotated version](https://github.com/Teddy-Li/MulVOIEL/tree/master/CaRB/data) of the [CaRB dataset](https://github.com/dair-iitd/CaRB).
The model produces multi-valent Open IE tuples, i.e. relations with various numbers of arguments (1, 2, or more). We provide an example below:
Consider the following sentence (taken from the CaRB dev set):
`Earlier this year , President Bush made a final `` take - it - or - leave it '' offer on the minimum wage`
Our model would extract the following relation from the sentence:
<<span style="color:#2471A3">President Bush</span>, <span style="color:#A93226">made</span>, <span style="color:#138D75">a final "take-it-or-leave-it" offer</span>, <span style="color:#B7950B ">on the minimum wage</span>, <span style="color:#B9770E">earlier this year</span>>
where we include <span style="color:#2471A3">President Bush</span> as the <span style="color:#2471A3">subject</span>, <span style="color:#A93226">made</span> as the <span style="color:#A93226">object</span>, <span style="color:#138D75">a final "take-it-or-leave-it" offer</span> as the<span style="color:#138D75">direct object</span>, and <span style="color:#B7950B ">on the minimum wage</span> and <span style="color:#B9770E">earlier this year</span>> as salient <span style="color:#B7950B">_compl_</span><span style="color:#B9770E">_ements_</span>.
We briefly describe how to use our model in the below, and provide further details in our [MulVOIEL repository on Github](https://github.com/Teddy-Li/MulVOIEL/)
## Getting Started
### Model Output Format
Given a sentence, the model produces textual predictions in the following format:
`<subj> ,, (<auxi> ###) <predicate> ,, (<prep1> ###) <obj1>, (<prep2> ###) <obj2>, ...`
### How to Use
1. Install the relevant libraries (using the LLaMA3-8b-instruct model as an example):
```bash
pip install transformers datasets peft torch
```
2. Load the model and perform inference (example):
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch
from llamaOIE import parse_outstr_to_triples
from llamaOIE_dataset import prepare_input
base_model_name = "meta-llama/Meta-Llama-3-8B-Instruct"
peft_adapter_name = "Teddy487/LLaMA3-8b-for-OpenIE"
model = AutoModelForCausalLM.from_pretrained(base_model_name)
model = PeftModel.from_pretrained(model, peft_adapter_name)
tokenizer = AutoTokenizer.from_pretrained(base_model_name)
input_text = "Earlier this year , President Bush made a final `` take - it - or - leave it '' offer on the minimum wage"
input_text, _ = prepare_input({'s': input_text}, tokenizer, has_labels=False)
input_ids = tokenizer(input_text, return_tensors="pt").input_ids
outputs = model.generate(input_ids)
outstr = tokenizer.decode(outputs[0][len(input_ids):], skip_special_tokens=True)
triples = parse_outstr_to_triples(outstr)
for tpl in triples:
print(tpl)
```
🍺
## Model Performance
The primary benefit of our model is the ability to extract finer-grained information for predicates. On the other hand, we also report performance on a roughly comparable basis with prior SOTA open IE models, where our method is comparable and even superior to prior models, while producing finer-grained and more complex outputs. We report evaluation results in (macro) F-1 metric, as well as in the average [Levenshtein Distance](https://pypi.org/project/python-Levenshtein/) between gold and predicted relations:
| Model | Levenshtein Distance | Macro F-1 |
| --- | --- | --- |
| [LoRA LLaMA2-7b](https://huggingface.co/Teddy487/LLaMA2-7b-for-OpenIE) | 5.85 | 50.2 |
| [LoRA LLaMA3-8b](https://huggingface.co/Teddy487/LLaMA3-8b-for-OpenIE) | **5.04** | **55.3** |
| RNN OIE * | - | 49.0 |
| IMOJIE * | - | 53.5 |
| Open IE 6 * | - | 54.0/52.7 |
Note that the precision and recall values are not directly comparable, because we evaluate the model prediction at a finer granularity, and we use different train/dev/test arrangements as the original CaRB dataset, hence the asterisk.
### Framework versions
- PEFT 0.10.0
- PEFT 0.5.0 |