Upload sigma_v5_ablation/V4_DynMu_result.json with huggingface_hub
Browse files
sigma_v5_ablation/V4_DynMu_result.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"variant": "V4_DynMu",
|
| 3 |
+
"best_fr_acc": 0.18943321792586498,
|
| 4 |
+
"final_tf_acc": 0.4841638355933248,
|
| 5 |
+
"final_ppl": 74.94092686946688,
|
| 6 |
+
"mean_step_var": 0.13881311813990274,
|
| 7 |
+
"mean_alpha_grad": 1.332800347123945e-07,
|
| 8 |
+
"n_epochs": 3,
|
| 9 |
+
"history": {
|
| 10 |
+
"epoch": [
|
| 11 |
+
1,
|
| 12 |
+
2,
|
| 13 |
+
3
|
| 14 |
+
],
|
| 15 |
+
"train_loss": [
|
| 16 |
+
15.708697366907268,
|
| 17 |
+
13.720215440439086,
|
| 18 |
+
19.28240414610449
|
| 19 |
+
],
|
| 20 |
+
"ppl": [
|
| 21 |
+
903.8067424539117,
|
| 22 |
+
78.64624914333437,
|
| 23 |
+
74.94092686946688
|
| 24 |
+
],
|
| 25 |
+
"tf_acc": [
|
| 26 |
+
0.3843802626327522,
|
| 27 |
+
0.4694811916141823,
|
| 28 |
+
0.4841638355933248
|
| 29 |
+
],
|
| 30 |
+
"fr_acc": [
|
| 31 |
+
0.18943321792586498,
|
| 32 |
+
0.05183016534119475,
|
| 33 |
+
0.14025399641208464
|
| 34 |
+
],
|
| 35 |
+
"step_var": [
|
| 36 |
+
0.14010637998580933,
|
| 37 |
+
0.13857358694076538,
|
| 38 |
+
0.13775938749313354
|
| 39 |
+
],
|
| 40 |
+
"alpha_grad": [
|
| 41 |
+
3.998401041371835e-07,
|
| 42 |
+
0.0,
|
| 43 |
+
0.0
|
| 44 |
+
]
|
| 45 |
+
}
|
| 46 |
+
}
|