README.md
Browse files
README.md
CHANGED
|
@@ -1,22 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
|
| 2 |
-
# OMDA: Arabic-English Chat LLM
|
| 3 |
|
| 4 |
**Model Name:** OMDA
|
| 5 |
-
**Architecture:** OMDA-Decoder
|
| 6 |
-
**Tokenizer:** OMDATokenizer
|
| 7 |
-
**Languages:** Arabic, English
|
| 8 |
-
**Type:** Chat/Instruction-following
|
| 9 |
**Author:** Binomda
|
| 10 |
-
**Date:** 2025-06-28
|
| 11 |
|
| 12 |
-
## Model
|
| 13 |
-
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
## Intended Use
|
| 22 |
- Chatbots, assistants, translation, and educational tools for Arabic/English.
|
|
@@ -33,5 +68,4 @@
|
|
| 33 |
- For research and educational purposes only.
|
| 34 |
|
| 35 |
## Export & Deployment
|
| 36 |
-
- See below for HuggingFace, llama.cpp, and ollama export
|
| 37 |
-
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- ar
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- chat
|
| 7 |
+
- translation
|
| 8 |
+
license: apache-2.0
|
| 9 |
+
base_model: OMDA-Decoder
|
| 10 |
+
pipeline_tag: text-generation
|
| 11 |
+
---
|
| 12 |
|
| 13 |
+
# OMDA: Bilingual Arabic-English Chat LLM
|
| 14 |
|
| 15 |
**Model Name:** OMDA
|
| 16 |
+
**Architecture:** OMDA-Decoder (Custom Architecture)
|
| 17 |
+
**Tokenizer:** OMDATokenizer (Custom Tokenizer)
|
| 18 |
+
**Languages:** Arabic (Primary), English
|
| 19 |
+
**Model Type:** Chat/Instruction-following
|
| 20 |
**Author:** Binomda
|
| 21 |
+
**Release Date:** 2025-06-28
|
| 22 |
|
| 23 |
+
## Model Overview
|
| 24 |
+
OMDA is a compact bilingual language model specifically designed for Arabic-English conversational AI applications. Built with a custom decoder architecture, it excels at understanding and generating natural responses in both languages.
|
| 25 |
+
|
| 26 |
+
## Model Specifications
|
| 27 |
+
| Parameter | Value |
|
| 28 |
+
|--------------------|-----------|
|
| 29 |
+
| Layers | 6 |
|
| 30 |
+
| Hidden Size | 512 |
|
| 31 |
+
| Attention Heads | 8 |
|
| 32 |
+
| FFN Dimension | 2048 |
|
| 33 |
+
| Max Sequence Length| 512 |
|
| 34 |
+
| Vocabulary Size | 128,004 |
|
| 35 |
+
| Training Data | 1,000 curated Arabic-English conversation pairs |
|
| 36 |
+
|
| 37 |
+
## Intended Uses
|
| 38 |
+
✅ Chatbot development
|
| 39 |
+
✅ Bilingual assistant applications
|
| 40 |
+
✅ Educational tools
|
| 41 |
+
✅ Basic translation tasks
|
| 42 |
+
|
| 43 |
+
## Quick Start
|
| 44 |
+
```python
|
| 45 |
+
from transformers import pipeline
|
| 46 |
+
|
| 47 |
+
# Initialize chat pipeline
|
| 48 |
+
chatbot = pipeline("text-generation", model="BINOMDA/OMDA")
|
| 49 |
+
|
| 50 |
+
# Arabic input example
|
| 51 |
+
ar_responser = chatbot("ما هو رأيك في التكنولوجيا الحديثة؟")
|
| 52 |
+
|
| 53 |
+
# English input example
|
| 54 |
+
en_response = chatbot("Explain artificial intelligence simply")
|
| 55 |
|
| 56 |
## Intended Use
|
| 57 |
- Chatbots, assistants, translation, and educational tools for Arabic/English.
|
|
|
|
| 68 |
- For research and educational purposes only.
|
| 69 |
|
| 70 |
## Export & Deployment
|
| 71 |
+
- See below for HuggingFace, llama.cpp, and ollama export
|
|
|