---
library_name: transformers
tags:
- argumentation
license: apache-2.0
datasets:
- Kleo/ArgKP_2021_GR
language:
- el
metrics:
- precision
base_model:
- ilsp/Meltemi-7B-v1
pipeline_tag: text-classification
---

# Model Card for Model ID

<!-- Provide a quick summary of what the model is/does. -->
This is a Meltemi-7b-v1 adapter model for a sequence classification task. 
It classifies keypoint-argument pairs as Matching/Non-matching. It was developed in the process of the KeyPoint Matching subtask of the [Key Point Analysis|Quantitative Argument Summarization Shared Task](https://github.com/IBM/KPA_2021_shared_task) as a solution for a low-resource language, Greek. 
The classifier was trained on the official shared task's dataset (ArgKP-2021) in a machine translated version for Greek with madlad-400-3b. For details refer to ArgKP-2021-GR dataset. 


## Model Details

### Model Description

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

This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.

- **Developed by:** https://huggingface.co/Kleo
- **Shared by [optional]:** https://huggingface.co/Kleo
- **Model type:** adapter
- **Language(s) (NLP):** el/GR
- **License:** Apache license 2.0
- **Finetuned from model [optional]:**  [ilsp/Meltemi-7B-v1](https://huggingface.co/ilsp/Meltemi-7B-v1)

### Model Sources [optional]

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

- **Repository:**  https://github.com/Kleo-Karap/KPA_thesis
- **Paper [optional]:** [More Information Needed]

## Uses

<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->

### Direct Use

<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->

[More Information Needed]



## Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

The model has been finetuned on the train set of the ArgKP-2021-GR dataset, which is the result of machine translation.

## How to Get Started with the Model

Use the code below to get started with the model.

```
import torch
from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel, PeftConfig

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16)


peft_model_id = "Kleo/meltemi_base_finetuning_kpm_kp_arg"
config = PeftConfig.from_pretrained(peft_model_id)
model = AutoModelForSequenceClassification.from_pretrained(config.base_model_name_or_path, return_dict=True, quantization_config=bnb_config, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)

# Load the Lora model
model = PeftModel.from_pretrained(model, peft_model_id)

tokenizer.pad_token_id = tokenizer.eos_token_id
tokenizer.pad_token = tokenizer.eos_token

model.eval()
# Initialize the classifier pipeline
classifier = pipeline(
    "text-classification",
    model=model,
    tokenizer=tokenizer,
    device_map="auto"
)

# List of inputs
texts = [
    "Keypoint: Ο εμβολιασμός μπορεί να έχει απρόβλεπτες παρενέργειες; Argument: Οι παιδικοί εμβολιασμοί ρουτίνας δεν θα πρέπει να είναι υποχρεωτικοί, καθώς τα παιδιά μπορεί να μην είναι σε θέση να αντέξουν τις επιπτώσεις τους",
    "Keypoint: Ο εμβολιασμός μπορεί να έχει απρόβλεπτες παρενέργειες; Argument: Ένα εμβόλιο, το οποίο δεν έχει ελεγχθεί επαρκώς και του οποίου οι επιπτώσεις δεν είναι γνωστές, δεν ενδείκνυται για τα παιδιά",
    "Keypoint: Ο εμβολιασμός μπορεί να έχει απρόβλεπτες παρενέργειες; Argument: Ο παιδικός εμβολιασμός δεν είναι απαραίτητος και μπορεί να επιφέρει άλλες ασθένειες στα παιδιά",
    "Keypoint: Στις Ηνωμένες Πολιτείες δεν υπάρχει ασφάλεια; Argument: σε ορισμένες πολιτείες εξακολουθούν να υπάρχουν πολλές ληστείες και εγκλήματα στα οποία εμπλέκονται αθώοι άνθρωποι",
    "Keypoint: Στις Ηνωμένες Πολιτείες δεν υπάρχει ασφάλεια; Argument: η κουλτούρα των ΗΠΑ προωθεί τον υλισμό",
    "Keypoint: Στις Ηνωμένες Πολιτείες δεν υπάρχει ασφάλεια; Argument: όχι, επειδή οι φόροι είναι υψηλοί και δαπανηροί",
    "Keypoint: Στις Ηνωμένες Πολιτείες δεν υπάρχει ασφάλεια; Argument: η υγειονομική περίθαλψη και η εκπαίδευση είναι υπερβολικά ακριβές για τη μεσαία τάξη",
    "Keypoint: Στις Ηνωμένες Πολιτείες δεν υπάρχει ασφάλεια; Argument: δεν είναι όλα όπως τις ταινίες, η Αμερική έχει πολλή ανισότητα"
]

# Perform inference for multiple inputs
results = classifier(texts)

# Print results with sentence numbers
for idx, result in enumerate(results, start=1):
    print(f"Sentence {idx}: Predicted Label: {result['label']}, Score: {result['score']:.4f}")
```


## Training Details

### Training Data

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

[ArgKP_2021_GR]](https://huggingface.co/datasets/Kleo/ArgKP_2021_GR)

### Training Procedure

<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->

#### Preprocessing 
Social media text removal
-Remove URLs, HTML tags, emojis and certain symbols


#### Training Hyperparameters

- **Training regime:**  bf16 Mixed Precision <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->

#### Speeds, Sizes, Times [optional]

<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->


## Evaluation

<!-- This section describes the evaluation protocols and provides the results. -->

### Testing Data, Factors & Metrics

#### Testing Data

<!-- This should link to a Dataset Card if possible. -->

[More Information Needed]


#### Metrics

<!-- These are the evaluation metrics being used, ideally with a description of why. -->

mean Average Precision (mAP)

### Results

[More Information Needed]

#### Summary

## Environmental Impact

<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->

Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).

- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]

## Technical Specifications

### Model Architecture and Objective

[More Information Needed]

### Compute Infrastructure
The model has been trained with the available [Kaggle](https://www.kaggle.com/) resources:
- Number of nodes: 1
- Number of GPUs per node: 1
- GPU type: NVIDIA P100
- GPU memory: 16GB


## Citation

<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->

**BibTeX:**

```
@masterthesis{3456844,
    title = "Key Point Analysis in Greek: A New Dataset and Baselines",
    authorField = "Καραπαναγιώτου, Κλεοπάτρα",
    year = "2025",
    school = "ΠΜΣ Γλωσσική Τεχνολογία, Εθνικό και Καποδιστριακό Πανεπιστήμιο Αθηνών"
}
```

**APA:**

Karapanagiotou, K. (2025). Key Point Analysis in Greek: A New Dataset and Baselines [MSc Thesis, National and Kapodistrian University of Athens]. Pergamos.https://pergamos.lib.uoa.gr/uoa/dl/frontend/el/browse/3456844


## Model Card Contact

https://huggingface.co/Kleo