Update README.md
Browse files
README.md
CHANGED
|
@@ -17,8 +17,7 @@ tags:
|
|
| 17 |
- 4bit-quantization
|
| 18 |
datasets:
|
| 19 |
- iamshnoo/alpaca-cleaned-bengali
|
| 20 |
-
|
| 21 |
-
base_model: google/gemma-2-27b-it
|
| 22 |
model_type: gemma2
|
| 23 |
---
|
| 24 |
|
|
@@ -28,7 +27,7 @@ model_type: gemma2
|
|
| 28 |
|
| 29 |
## Model Description
|
| 30 |
|
| 31 |
-
This model is a fine-tuned version of [google/gemma-2-
|
| 32 |
|
| 33 |
### Key Features
|
| 34 |
|
|
@@ -48,7 +47,7 @@ This model is a fine-tuned version of [google/gemma-2-27b-it](https://huggingfac
|
|
| 48 |
- **Total Examples**: ~8,000 high-quality Bengali examples
|
| 49 |
|
| 50 |
### Training Configuration
|
| 51 |
-
- **Base Model**: google/gemma-2-
|
| 52 |
- **Fine-tuning Method**: LoRA (Low-Rank Adaptation)
|
| 53 |
- **Quantization**: 4-bit using BitsAndBytesConfig
|
| 54 |
- **LoRA Rank**: 16
|
|
@@ -76,7 +75,7 @@ import torch
|
|
| 76 |
|
| 77 |
# Load the model and tokenizer
|
| 78 |
base_model = AutoModelForCausalLM.from_pretrained(
|
| 79 |
-
"google/gemma-2-
|
| 80 |
torch_dtype=torch.float16,
|
| 81 |
device_map="auto"
|
| 82 |
)
|
|
@@ -177,23 +176,23 @@ See the [complete notebook](https://github.com/your-repo/gemma3-bengali-multimod
|
|
| 177 |
|
| 178 |
## Model Card Authors
|
| 179 |
|
| 180 |
-
Created by the
|
| 181 |
|
| 182 |
## Citation
|
| 183 |
|
| 184 |
```bibtex
|
| 185 |
@misc{gemma2-bengali-multimodal,
|
| 186 |
-
title={Gemma 2
|
| 187 |
-
author={
|
| 188 |
-
year={
|
| 189 |
url={https://huggingface.co/retro56/gemma3-4b-bengali-multimodal-persona}
|
| 190 |
}
|
| 191 |
```
|
| 192 |
|
| 193 |
## License
|
| 194 |
|
| 195 |
-
This model is licensed under the Gemma License. See the [original model](https://huggingface.co/google/gemma-2-
|
| 196 |
|
| 197 |
---
|
| 198 |
|
| 199 |
-
**Built with ❤️ for the Bengali AI community**
|
|
|
|
| 17 |
- 4bit-quantization
|
| 18 |
datasets:
|
| 19 |
- iamshnoo/alpaca-cleaned-bengali
|
| 20 |
+
base_model: google/gemma-2-4b-it
|
|
|
|
| 21 |
model_type: gemma2
|
| 22 |
---
|
| 23 |
|
|
|
|
| 27 |
|
| 28 |
## Model Description
|
| 29 |
|
| 30 |
+
This model is a fine-tuned version of [google/gemma-2-4b-it](https://huggingface.co/google/gemma-2-4b-it) specifically optimized for Bengali language conversations and multimodal AI persona applications. The model has been trained to provide natural, helpful responses in Bengali and can be integrated with voice synthesis for complete multimodal AI experiences.
|
| 31 |
|
| 32 |
### Key Features
|
| 33 |
|
|
|
|
| 47 |
- **Total Examples**: ~8,000 high-quality Bengali examples
|
| 48 |
|
| 49 |
### Training Configuration
|
| 50 |
+
- **Base Model**: google/gemma-2-4b-it
|
| 51 |
- **Fine-tuning Method**: LoRA (Low-Rank Adaptation)
|
| 52 |
- **Quantization**: 4-bit using BitsAndBytesConfig
|
| 53 |
- **LoRA Rank**: 16
|
|
|
|
| 75 |
|
| 76 |
# Load the model and tokenizer
|
| 77 |
base_model = AutoModelForCausalLM.from_pretrained(
|
| 78 |
+
"google/gemma-2-4b-it",
|
| 79 |
torch_dtype=torch.float16,
|
| 80 |
device_map="auto"
|
| 81 |
)
|
|
|
|
| 176 |
|
| 177 |
## Model Card Authors
|
| 178 |
|
| 179 |
+
Created by the Personify research team for advancing Bengali language AI capabilities.
|
| 180 |
|
| 181 |
## Citation
|
| 182 |
|
| 183 |
```bibtex
|
| 184 |
@misc{gemma2-bengali-multimodal,
|
| 185 |
+
title={Gemma 2 4B Bengali Multimodal Persona},
|
| 186 |
+
author={Personify Research Team},
|
| 187 |
+
year={2025},
|
| 188 |
url={https://huggingface.co/retro56/gemma3-4b-bengali-multimodal-persona}
|
| 189 |
}
|
| 190 |
```
|
| 191 |
|
| 192 |
## License
|
| 193 |
|
| 194 |
+
This model is licensed under the Gemma License. See the [original model](https://huggingface.co/google/gemma-2-4b-it) for complete license terms.
|
| 195 |
|
| 196 |
---
|
| 197 |
|
| 198 |
+
**Built with ❤️ for the Bengali AI community**
|