Upload folder using huggingface_hub
Browse files- hf_upload_new/README.md +17 -3
hf_upload_new/README.md
CHANGED
|
@@ -1,10 +1,24 @@
|
|
| 1 |
---
|
|
|
|
| 2 |
tags:
|
| 3 |
- hrm
|
|
|
|
| 4 |
- sudoku
|
| 5 |
- pytorch
|
| 6 |
- safetensors
|
| 7 |
---
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
license: mit
|
| 3 |
tags:
|
| 4 |
- hrm
|
| 5 |
+
- hierarchical-reasoning
|
| 6 |
- sudoku
|
| 7 |
- pytorch
|
| 8 |
- safetensors
|
| 9 |
---
|
| 10 |
+
|
| 11 |
+
# HRM Sudoku Solver
|
| 12 |
+
|
| 13 |
+
This model is a Hierarchical Reasoning Model (HRM) trained on the Sudoku-Extreme-1K dataset.
|
| 14 |
+
|
| 15 |
+
## Performance
|
| 16 |
+
- **Exact Accuracy**: 60.6%
|
| 17 |
+
- **Token Accuracy**: 88.2%
|
| 18 |
+
- **Training**: 20,000 epochs with 1000x augmentation
|
| 19 |
+
|
| 20 |
+
## Files
|
| 21 |
+
- `model.safetensors`: Optimized weights.
|
| 22 |
+
- `config.yaml`: Architecture configuration.
|
| 23 |
+
- `hrm_act_v1.py`: Model source code.
|
| 24 |
+
- `losses.py`: Loss functions.
|