Update README.md
Browse files
README.md
CHANGED
|
@@ -234,4 +234,6 @@ When pre-training Lippershey-Base, use the following guidelines to evaluate the
|
|
| 234 |
|
| 235 |
* Validation Loss (MSE): Measures how accurately the model reconstructs the continuous coefficients. A healthy convergence on structured MILPs should see `val/loss` drop below 0.15 (L1 average absolute error $\approx 0.38$ on $[-1, 1]$ scale).
|
| 236 |
* Grad Norm (L2): Should stabilize between 0.3 and 1.0. Constant clipping to 1.0 indicates learning rate is too high; an decaying gradient norm towards 0 without loss convergence indicates gradient vanishing.
|
| 237 |
-
* Prediction Mean Convergence: Check `train/pred_mean` against `train/true_mean`. They should overlap near the $0$ axis, indicating the model's global prediction scale matches the mathematical distribution of your planning domain.
|
|
|
|
|
|
|
|
|
| 234 |
|
| 235 |
* Validation Loss (MSE): Measures how accurately the model reconstructs the continuous coefficients. A healthy convergence on structured MILPs should see `val/loss` drop below 0.15 (L1 average absolute error $\approx 0.38$ on $[-1, 1]$ scale).
|
| 236 |
* Grad Norm (L2): Should stabilize between 0.3 and 1.0. Constant clipping to 1.0 indicates learning rate is too high; an decaying gradient norm towards 0 without loss convergence indicates gradient vanishing.
|
| 237 |
+
* Prediction Mean Convergence: Check `train/pred_mean` against `train/true_mean`. They should overlap near the $0$ axis, indicating the model's global prediction scale matches the mathematical distribution of your planning domain.
|
| 238 |
+
|
| 239 |
+
|