Sheikh / README.md
megharudushi's picture
Fix YAML metadata with proper language codes
144f1ca verified
---
language: bn
language_bcp47:
- bn
- bn-IN
- bn-BD
license: apache-2.0
base_model: microsoft/DialoGPT-medium
tags:
- bengali
- bangla
- transformer
- causal-lm
- instruction-following
- nlp
- text-generation
- conversational-ai
- educational
- general-knowledge
model-index:
- name: Sheikh Bengali AI
results: []
---
# Sheikh Bengali AI Model
## Model Description
**Sheikh** is a Bengali (Bangla) language AI model trained for instruction following and conversational tasks. Built on top of Microsoft's DialoGPT-medium, this model has been fine-tuned with Bengali instruction-following data to understand and generate responses in Bengali language.
## Model Details
- **Model Type:** Language Model, Text Generation
- **Architecture:** GPT-2 based (DialoGPT-medium)
- **Base Model:** microsoft/DialoGPT-medium
- **Parameters:** 355M
- **Language:** Bengali (Bangla)
- **Training Data:** Alpaca Bangla instruction dataset
- **Model Size:** 1.4GB
- **License:** Apache 2.0
## Intended Use
This model is designed for:
- Bengali language text generation
- Instruction following and question answering
- Educational content creation
- Cultural and historical knowledge responses
- General conversational AI in Bengali
## Usage
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
# Load the model
tokenizer = AutoTokenizer.from_pretrained("megharudushi/Sheikh")
model = AutoModelForCausalLM.from_pretrained("megharudushi/Sheikh")
# Generate Bengali response
input_text = "বাংলাদেশের রাজধানী কী?"
inputs = tokenizer.encode(input_text, return_tensors="pt")
outputs = model.generate(inputs, max_length=150, temperature=0.8)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)
```
### Example Prompts
- Educational: "গণিতের মৌলিক নীতি বলুন"
- Cultural: "বাংলা সাহিত্যের বিখ্যাত কবি কারা?"
- General: "স্বাস্থ্যকর থাকার উপায় বলুন"
- Historical: "বাংলাদেশের স্বাধীনতার ইতিহাস বর্ণনা করুন"
## Model Performance
- Supports Bengali language understanding and generation
- Trained on Bengali instruction-following dataset
- Optimized for educational and conversational contexts
- Cultural knowledge preservation for Bengali language
## Limitations
- Trained primarily on Bengali instruction data
- May have limitations in very specialized domains
- Performance depends on input quality and clarity
- Model size limited by computational resources
## Training Details
- **Base Model:** microsoft/DialoGPT-medium
- **Fine-tuning Data:** Alpaca Bangla dataset
- **Training Approach:** Instruction following
- **Language Focus:** Bengali (Bangla) language
## Citation
If you use this model, please cite:
```bibtex
@misc{SheikhBengaliAI,
title={Sheikh Bengali AI Model},
author={megharudushi},
year={2025},
url={https://huggingface.co/megharudushi/Sheikh},
note={Bengali language instruction-following model based on DialoGPT-medium}
}
```
## License
This model is released under the Apache 2.0 License.
## Contributing
This model is part of the Bengali AI initiative to make Bengali language AI more accessible to the community.
---
**Created:** December 21, 2025
**Repository:** https://huggingface.co/megharudushi/Sheikh
**Base Model:** microsoft/DialoGPT-medium
**Language:** Bengali (Bangla)