README.md

---
license: cc-by-nc-4.0
pipeline_tag: question-answering
tags:
    - question-answering
    - transformers
    - generated_from_trainer
datasets:
    - squad_v2
    - LLukas22/nq-simplified
---

# deberta-v3-base-qa-en
This model is an extractive qa model.
It's a fine-tuned version of [deberta-v3-base](https://huggingface.co/microsoft/deberta-v3-base) on the following datasets: [squad_v2](https://huggingface.co/datasets/squad_v2), [LLukas22/nq-simplified](https://huggingface.co/datasets/LLukas22/nq-simplified).



## Usage

You can use the model like this:

```python
from transformers import pipeline

#Make predictions
model_name = "LLukas22/deberta-v3-base-qa-en"
nlp = pipeline('question-answering', model=model_name, tokenizer=model_name)

QA_input = {
    "question": "What's my name?",
    "context": "My name is Clara and I live in Berkeley."
}

result = nlp(QA_input)
print(result)
```
Alternatively you can load the model and tokenizer on their own:
```python
from transformers import AutoModelForQuestionAnswering, AutoTokenizer

#Make predictions
model_name = "LLukas22/deberta-v3-base-qa-en"
model = AutoModelForQuestionAnswering.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
```

## Training hyperparameters
The following hyperparameters were used during training:

- learning_rate: 2E-05
- per device batch size: 15
- effective batch size: 45
- seed: 42
- optimizer: AdamW with betas (0.9,0.999) and eps 1E-08
- weight decay: 1E-02
- D-Adaptation: False
- Warmup: False
- number of epochs: 10
- mixed_precision_training: bf16

## Training results
| Epoch | Train Loss | Validation Loss |
| ----- | ---------- | --------------- |
| 0 | 1.57 | 1.47 |
| 1 | 1.22 | 1.46 |
| 2 | 1.09 | 1.48 |
| 3 | 1.02 | 1.5 |

## Evaluation results
| Epoch | f1 | exact_match |
| ----- | ----- | ----- |
| 0 | 0.658 | 0.514 |
| 1 | 0.661 | 0.522 |
| 2 | 0.664 | 0.525 |
| 3 | 0.666 | 0.524 |

## Framework versions
- Transformers: 4.25.1
- PyTorch: 2.0.0+cu118
- PyTorch Lightning: 1.8.6
- Datasets: 2.7.1
- Tokenizers: 0.13.1
- Sentence Transformers: 2.2.2

## Additional Information
This model was trained as part of my Master's Thesis **'Evaluation of transformer based language models for use in service information systems'**. The source code is available on [Github](https://github.com/LLukas22/Retrieval-Augmented-QA).