Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Enhanced Affine Model
|
| 2 |
+
|
| 3 |
+
## Improvements Made
|
| 4 |
+
|
| 5 |
+
### Chat Template Enhancements
|
| 6 |
+
- **Specialized Problem Solving**: Added environment-specific reasoning strategies
|
| 7 |
+
- **Mathematical Focus**: Enhanced guidance for Project Euler and constraint satisfaction problems
|
| 8 |
+
- **Stack VM Expertise**: Detailed instructions for virtual machine execution tracing
|
| 9 |
+
- **Systematic Approach**: Step-by-step problem solving methodology
|
| 10 |
+
|
| 11 |
+
### Generation Configuration
|
| 12 |
+
- **Optimized Sampling**: Temperature 0.7, top_p 0.9, top_k 40 for balanced creativity and precision
|
| 13 |
+
- **Repetition Control**: Penalty 1.1 to reduce redundant outputs
|
| 14 |
+
- **Extended Context**: Max 2048 tokens for complex reasoning
|
| 15 |
+
|
| 16 |
+
### Target Improvements
|
| 17 |
+
- **HVM Environment**: Enhanced constraint satisfaction and VM execution tracing
|
| 18 |
+
- **ELR Environment**: Better mathematical reasoning for Project Euler problems
|
| 19 |
+
- **Maintained Performance**: Preserved strengths in SAT, ABD, and DED environments
|
| 20 |
+
|
| 21 |
+
## Model Architecture
|
| 22 |
+
- **Base**: GptOssForCausalLM with Mixture of Experts
|
| 23 |
+
- **Size**: 36 layers, 128 experts, 4 active per token
|
| 24 |
+
- **Quantization**: MXFP4 for efficiency
|
| 25 |
+
- **Context**: 131K tokens with YARN RoPE scaling
|
| 26 |
+
|
| 27 |
+
## Usage
|
| 28 |
+
This model is optimized for the Affine protocol with enhanced reasoning capabilities across all challenge environments.
|