datasets:
- SajjadAyoubi/persian_qa
language:
- fa
pipeline_tag: question-answering
license: apache-2.0
library_name: transformers
tags:
- roberta
- question-answering
- Persian
Tara-Roberta-Base-FA-QA
Tara-Roberta-Base-FA-QA is a fine-tuned version of the facebookAI/roberta-base model for question-answering tasks, trained on the SajjadAyoubi/persian_qa dataset. This model is designed to understand and generate answers to questions posed in Persian.
Model Description
This model was fine-tuned on a dataset containing Persian question-answering pairs. It leverages the roberta-base architecture to provide answers based on the context provided. The training process was performed with a focus on improving the model's ability to handle Persian text and answer questions effectively.
Training Details
The model was trained for 3 epochs with the following training and validation losses:
Epoch 1:
- Training Loss: 2.0713
- Validation Loss: 2.1061
Epoch 2:
- Training Loss: 2.1558
- Validation Loss: 2.0121
Epoch 3:
- Training Loss: 2.0951
- Validation Loss: 2.0168
Evaluation Results
The model achieved the following results:
- Training Loss: Decreased over the epochs, indicating effective learning.
- Validation Loss: Slight variations were observed, reflecting the model's performance on unseen data.
Usage
To use this model for question-answering tasks, load it with the transformers library:
from transformers import AutoTokenizer, AutoModelForQuestionAnswering, pipeline
model = "hosseinhimself/tara-roberta-base-fa-qa"
# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(model)
model = AutoModelForQuestionAnswering.from_pretrained(model)
# Create a QA pipeline
qa_pipeline = pipeline("question-answering", model=model, tokenizer=tokenizer)
# Example usage
context = "CONTEXT"
question = "QUESTION"
# Modify the pipeline to return more answers
results = qa_pipeline(question=question, context=context, top_k=5) # top_k specifies the number of answers
# Display the answers
for idx, result in enumerate(results):
print(f"Answer {idx+1}: {result['answer']}")
Datasets
The model was fine-tuned using the SajjadAyoubi/persian_qa dataset.
Languages
The model supports the Persian language.
Additional Information
For more details on how to fine-tune similar models or to report issues, please visit the Hugging Face documentation.