README.md

---
license: apache-2.0
language: ti
library_name: sentence-transformers
pipeline_tag: sentence-similarity
tags:
- sentence-transformers
- feature-extraction
- sentence-similarity
- transformers
widget:
- text: "ግራፋይት ኣብ መላእ ዓለም ዳርጋ ብምዕሩይ ዝርጋሐ’ዩ ዝርከብ"
---

# TiELECTRA BiEncoder Model

This model is a bi-encoder model for the Tigrinya language based on [TiELECTRA-small](https://huggingface.co/fgaim/tielectra-small).
The model maps sentences & paragraphs to a 256 dimensional dense vector space and can be used for tasks like text embedding, clustering, or semantic search.

This is part of a work that introduces monolingual bi-encoder language models for Tigrinya. For a larger and more powerful model look at [TiRoBERTa-bi-encoder](https://huggingface.co/fgaim/tiroberta-bi-encoder). The models are based on the [sentence-transformers](https://www.sbert.net) architecture and are trained on Tigrinya question-answering and information retrieval datasets. The models are designed to support semantic search tasks, such as information retrieval, text representation, and question answering.

## Using Model with Sentence-Transformers

Using this model becomes easy when you have sentence-transformers installed:

```shell
pip install -U sentence-transformers
```

Then use the model as follows:

```python
from sentence_transformers import SentenceTransformer
sentences = ["ሓደ ሰብኣይ ፈረስ ይጋልብ ኣሎ።", "ሓንቲ ጓል ክራር ትጻወት ኣላ።"]

model = SentenceTransformer('fgaim/tielectra-bi-encoder')
embeddings = model.encode(sentences)
print(embeddings)
```

## Using Model with 🤗 Transformers

Use the transformers library as follows:
Pass the input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.

```python
import torch
from transformers import AutoModel, AutoTokenizer


# Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0]  # First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# Sentences we want sentence embeddings for
sentences = ["ሓደ ሰብኣይ ፈረስ ይጋልብ ኣሎ።", "ሓንቲ ጓል ክራር ትጻወት ኣላ።"]

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained("fgaim/tielectra-bi-encoder")
model = AutoModel.from_pretrained("fgaim/tielectra-bi-encoder")

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors="pt")

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling. In this case, mean pooling.
sentence_embeddings = mean_pooling(model_output, encoded_input["attention_mask"])

print("Sentence embeddings:", sentence_embeddings)
```

## Architecture

### Base Model

The model properties:

| Model Size | Layers  | Attn. Heads | Hidden Size  | FFN  | Parameters  | Max. Seq  |
|------------|----|----|-----|------|------|------|
| SMALL      | 12 | 4  | 256 | 1024 | 14M  | 512  |

### BiEncoder Model

- Max Seq Length: `512`
- Word embedding dimension: `256`

```text
SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: ElectraModel

  (1): Pooling({'word_embedding_dimension': 256, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)
```

## Citation

If you use this model in your product or research, you can cite it as follows:

```bibtex
@misc{gaim-2024-semantic-search,
  title     = {{Semantic Search Models for Tigrinya}},
  author    = {Fitsum Gaim},
  month     = {January},
  year      = {2024},
  publisher = {Hugging Face Hub},
  doi       = {10.57967/hf/6068},
  url       = {https://huggingface.co/fgaim/tiroberta-bi-encoder}
}
```