---
license: mit
datasets:
- tsac
language:
- ar
---

This is a converted version of [Instadeep's](https://huggingface.co/InstaDeepAI) [TunBERT](https://github.com/instadeepai/tunbert/) from nemo to safetensors.

Make sure to read the original model [licence](https://github.com/instadeepai/tunbert/blob/main/LICENSE)
<details>
<summary>architectural changes </summary>
  
## original model head

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6527e89a8808d80ccff88b7a/b-uXLwsi4n1Tc7-OtHe9b.png)


## this model head

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6527e89a8808d80ccff88b7a/xG-tOQscrvxb4wQm_2n-r.png)

</details>

## Note
this is a WIP and any contributions are welcome 


# how to load the model 
```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("not-lain/TunBERT")
model = AutoModelForSequenceClassification.from_pretrained("not-lain/TunBERT",trust_remote_code=True)
```


# how to use the model 
```python
text = "[insert text here]"
inputs = tokenizer(text,return_tensors='pt')
output = model(**inputs)
```
or you can use the pipeline : 
```python
from transformers import pipeline

pipe = pipeline(model="not-lain/TunBERT",tokenizer = "not-lain/TunBERT",trust_remote_code=True)
pipe("text")
```
**IMPORTANT** : 
* Make sure to enable `trust_remote_code=True`