TURNA / apps /home.py
gokceuludogan's picture
Update apps/home.py
7510de1 verified
import requests
import streamlit as st
import time
from transformers import pipeline
import os
from .utils import query
def write():
st.markdown(
"""
<h1 style="text-align:left;">TURNA</h1>
""",
unsafe_allow_html=True,
)
st.write("#")
col = st.columns(2)
col[0].image("images/turna-logo.png", width=100)
st.markdown(
"""
<h3 style="text-align:left;">a Turkish encoder-decoder language model</h3>
<p style="text-align:right;"><p>
""", unsafe_allow_html=True,
)
st.markdown(
"""
Welcome to our Huggingface space, where you can explore the capabilities of TURNA.
**Key Features of TURNA:**
- **Powerful Architecture:** TURNA contains 1.1B parameters, and was pre-trained with an encoder-decoder architecture following the UL2 framework on 43B tokens from various domains.
- **Diverse Training Data:** Our model is trained on a varied dataset of 43 billion tokens, covering a wide array of domains.
- **Broad Applications:** TURNA is fine-tuned for a variety of generation and understanding tasks, including:
- Summarization
- Paraphrasing
- News title generation
- Sentiment classification
- Text categorization
- Named entity recognition
- Part-of-speech tagging
- Semantic textual similarity
- Natural language inference
Explore various applications powered by **TURNA** using the **Navigation** bar.
Refer to our [paper](https://arxiv.org/abs/2401.14373) for more details.
### Citation
```bibtex
@misc{uludogan2024turna,
title={TURNA: A Turkish Encoder-Decoder Language Model for Enhanced Understanding and Generation},
author={Gökçe Uludoğan and Zeynep Yirmibeşoğlu Balal and Furkan Akkurt and Melikşah Türker and Onur Güngör and Susan Üsküdarlı},
year={2024},
eprint={2401.14373},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
""")
st.markdown(
"""
<p style="text-align:right;"><em>TURNA can generate toxic content or provide erroneous information. Double-check before usage. </em><p>
""",
unsafe_allow_html=True,
)