|
--- |
|
license: mit |
|
base_model: xlm-roberta-base |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- squad_v2 |
|
model-index: |
|
- name: xlm-roberta-base-finetuned-squad2 |
|
results: [] |
|
language: |
|
- en |
|
- ar |
|
- de |
|
- el |
|
- es |
|
- hi |
|
- ro |
|
- ru |
|
- th |
|
- tr |
|
- vi |
|
- zh |
|
metrics: |
|
- exact_match |
|
- f1 |
|
pipeline_tag: question-answering |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
## Model description |
|
|
|
XLM-RoBERTa is a multilingual version of RoBERTa developed by Facebook AI. It is pre-trained on 2.5TB of filtered CommonCrawl data containing 100 languages. |
|
It is an extension of RoBERTa, which is itself a variant of the BERT model. XLM-RoBERTa is designed to handle multiple languages and demonstrate strong performance across a wide range of tasks, making it highly useful for multilingual natural language processing (NLP) applications. |
|
|
|
**Language model:** xlm-roberta-base |
|
**Language:** English |
|
**Downstream-task:** Question-Answering |
|
**Training data:** Train-set SQuAD 2.0 |
|
**Evaluation data:** Evaluation-set SQuAD 2.0 |
|
**Hardware Accelerator used**: GPU Tesla T4 |
|
|
|
## Intended uses & limitations |
|
|
|
Multilingual Question-Answering |
|
|
|
For Question-Answering in English- |
|
|
|
```python |
|
!pip install transformers |
|
from transformers import pipeline |
|
model_checkpoint = "IProject-10/bert-base-uncased-finetuned-squad2" |
|
question_answerer = pipeline("question-answering", model=model_checkpoint) |
|
|
|
context = """ |
|
The Statue of Unity is the world's tallest statue, with a height of 182 metres (597 feet), located near Kevadia in the state of Gujarat, India. |
|
""" |
|
|
|
question = "What is the height of statue of Unity?" |
|
question_answerer(question=question, context=context) |
|
``` |
|
For Question-Answering in Hindi- |
|
|
|
```python |
|
!pip install transformers |
|
from transformers import pipeline |
|
model_checkpoint = "IProject-10/bert-base-uncased-finetuned-squad2" |
|
question_answerer = pipeline("question-answering", model=model_checkpoint) |
|
|
|
context = """ |
|
स्टैच्यू ऑफ यूनिटी दुनिया की सबसे ऊंची प्रतिमा है, जिसकी ऊंचाई 182 मीटर (597 फीट) है, जो भारत के गुजरात राज्य में केवडिया के पास स्थित है। |
|
""" |
|
|
|
question = "स्टैच्यू ऑफ यूनिटी की ऊंचाई कितनी है?" |
|
question_answerer(question=question, context=context) |
|
``` |
|
|
|
For Question-Answering in Spanish- |
|
|
|
```python |
|
!pip install transformers |
|
from transformers import pipeline |
|
model_checkpoint = "IProject-10/bert-base-uncased-finetuned-squad2" |
|
question_answerer = pipeline("question-answering", model=model_checkpoint) |
|
|
|
context = """ |
|
La Estatua de la Unidad es la estatua más alta del mundo, con una altura de 182 metros (597 pies), ubicada cerca de Kevadia en el estado de Gujarat, India. |
|
""" |
|
|
|
question = "¿Cuál es la altura de la estatua de la Unidad?" |
|
question_answerer(question=question, context=context) |
|
``` |
|
|
|
## Results |
|
|
|
Evaluation on SQuAD 2.0 validation dataset: |
|
|
|
``` |
|
exact: 75.51587635812348, |
|
f1: 78.7328391907263, |
|
total: 11873, |
|
HasAns_exact: 73.00944669365722, |
|
HasAns_f1: 79.45259779208723, |
|
HasAns_total: 5928, |
|
NoAns_exact: 78.01513877207738, |
|
NoAns_f1: 78.01513877207738, |
|
NoAns_total: 5945, |
|
best_exact: 75.51587635812348, |
|
best_exact_thresh: 0.999241054058075, |
|
best_f1: 78.73283919072665, |
|
best_f1_thresh: 0.999241054058075, |
|
total_time_in_seconds: 218.97641910400125, |
|
samples_per_second: 54.220450076686134, |
|
latency_in_seconds: 0.018443225730986376 |
|
``` |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 3e-05 |
|
- train_batch_size: 16 |
|
- eval_batch_size: 16 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 3 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | |
|
|:-------------:|:-----:|:-----:|:---------------:| |
|
| 1.0539 | 1.0 | 8333 | 0.9962 | |
|
| 0.8013 | 2.0 | 16666 | 0.8910 | |
|
| 0.5918 | 3.0 | 24999 | 0.9802 | |
|
|
|
This model is a fine-tuned version of [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) on the squad_v2 dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.9802 |
|
|
|
### Framework versions |
|
|
|
- Transformers 4.31.0 |
|
- Pytorch 2.0.1+cu118 |
|
- Datasets 2.14.3 |
|
- Tokenizers 0.13.3 |
