Add TabNet model README
Browse files
README.md
ADDED
|
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
tags:
|
| 4 |
+
- finance
|
| 5 |
+
- tabnet
|
| 6 |
+
- tabular
|
| 7 |
+
- pytorch
|
| 8 |
+
- mag7
|
| 9 |
+
- private-dataset-trained
|
| 10 |
+
library_name: pytorch
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# MAG7 TabNet Model
|
| 14 |
+
|
| 15 |
+
Public model artifact repo for an optimized TabNet classifier trained on a private MAG7 daily ML dataset.
|
| 16 |
+
|
| 17 |
+
The dataset used for training is private and is not distributed with this model:
|
| 18 |
+
|
| 19 |
+
- Private dataset: `kyLELEng/mag7-ml-daily-dataset-5y`
|
| 20 |
+
- Default target: `target_next_10d_outperform_qqq`
|
| 21 |
+
- Split method: chronological train / validation / test
|
| 22 |
+
- Training hardware: Hugging Face Jobs, `a10g-large`
|
| 23 |
+
|
| 24 |
+
After training completes, this repo contains:
|
| 25 |
+
|
| 26 |
+
- `tabnet_model.zip`
|
| 27 |
+
- `scaler.joblib`
|
| 28 |
+
- `feature_columns.json`
|
| 29 |
+
- `metrics.json`
|
| 30 |
+
- `trial_results.csv`
|
| 31 |
+
- `test_predictions.csv`
|
| 32 |
+
|
| 33 |
+
This model is for research only. It is not investment advice.
|