Update README.md
Browse files
README.md
CHANGED
|
@@ -72,7 +72,15 @@ NeoLLM incorporates several cutting-edge components:
|
|
| 72 |
- **Polynomial Composition Activations**: PolyNorm activation functions in MLP layers for enhanced dynamics
|
| 73 |
- **Advanced Normalization**: LayerNorm Scaling (LNS) and Gradient-Preserving Activation Scaling (GPAS)
|
| 74 |
- **Efficient Linear Attention**: Gated Delta Networks for improved computational efficiency
|
|
|
|
| 75 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 76 |
|
| 77 |
|
| 78 |
### Architecture Details
|
|
|
|
| 72 |
- **Polynomial Composition Activations**: PolyNorm activation functions in MLP layers for enhanced dynamics
|
| 73 |
- **Advanced Normalization**: LayerNorm Scaling (LNS) and Gradient-Preserving Activation Scaling (GPAS)
|
| 74 |
- **Efficient Linear Attention**: Gated Delta Networks for improved computational efficiency
|
| 75 |
+
## Installation
|
| 76 |
|
| 77 |
+
Before using this model, install the required dependencies:
|
| 78 |
+
|
| 79 |
+
```bash
|
| 80 |
+
pip install git+https://github.com/huggingface/transformers.git@main
|
| 81 |
+
pip install "cut-cross-entropy @ git+https://github.com/apple/ml-cross-entropy.git"
|
| 82 |
+
pip install flash-linear-attention
|
| 83 |
+
```
|
| 84 |
|
| 85 |
|
| 86 |
### Architecture Details
|