Upload training_history.json with huggingface_hub
Browse files- training_history.json +1042 -0
training_history.json
ADDED
|
@@ -0,0 +1,1042 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"epoch": 1,
|
| 4 |
+
"train_loss": 0.5333103537559509,
|
| 5 |
+
"val_loss": 0.5222890377044678,
|
| 6 |
+
"val_mse": 0.32089319825172424,
|
| 7 |
+
"val_cos": 0.9922126531600952,
|
| 8 |
+
"lr": 6e-06,
|
| 9 |
+
"layer_weights": [
|
| 10 |
+
0.18799281120300293,
|
| 11 |
+
0.48695486783981323,
|
| 12 |
+
0.32505232095718384
|
| 13 |
+
]
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"epoch": 2,
|
| 17 |
+
"train_loss": 0.5328588336706161,
|
| 18 |
+
"val_loss": 0.5206942558288574,
|
| 19 |
+
"val_mse": 0.31876885890960693,
|
| 20 |
+
"val_cos": 0.9918535351753235,
|
| 21 |
+
"lr": 1.2e-05,
|
| 22 |
+
"layer_weights": [
|
| 23 |
+
0.1880057156085968,
|
| 24 |
+
0.48691338300704956,
|
| 25 |
+
0.325080931186676
|
| 26 |
+
]
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 3,
|
| 30 |
+
"train_loss": 0.5306472877661387,
|
| 31 |
+
"val_loss": 0.5182968974113464,
|
| 32 |
+
"val_mse": 0.31562013924121857,
|
| 33 |
+
"val_cos": 0.9912093877792358,
|
| 34 |
+
"lr": 1.8e-05,
|
| 35 |
+
"layer_weights": [
|
| 36 |
+
0.1880248337984085,
|
| 37 |
+
0.4868509769439697,
|
| 38 |
+
0.32512423396110535
|
| 39 |
+
]
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 4,
|
| 43 |
+
"train_loss": 0.5288386543591818,
|
| 44 |
+
"val_loss": 0.5155321359634399,
|
| 45 |
+
"val_mse": 0.3120589703321457,
|
| 46 |
+
"val_cos": 0.9903027713298798,
|
| 47 |
+
"lr": 2.4e-05,
|
| 48 |
+
"layer_weights": [
|
| 49 |
+
0.18804395198822021,
|
| 50 |
+
0.48678505420684814,
|
| 51 |
+
0.32517102360725403
|
| 52 |
+
]
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"epoch": 5,
|
| 56 |
+
"train_loss": 0.5236251354217529,
|
| 57 |
+
"val_loss": 0.5126239061355591,
|
| 58 |
+
"val_mse": 0.30844202637672424,
|
| 59 |
+
"val_cos": 0.9890482425689697,
|
| 60 |
+
"lr": 3e-05,
|
| 61 |
+
"layer_weights": [
|
| 62 |
+
0.1880473494529724,
|
| 63 |
+
0.4867120683193207,
|
| 64 |
+
0.3252406120300293
|
| 65 |
+
]
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 6,
|
| 69 |
+
"train_loss": 0.5210557878017426,
|
| 70 |
+
"val_loss": 0.5096661448478699,
|
| 71 |
+
"val_mse": 0.3051048368215561,
|
| 72 |
+
"val_cos": 0.9869758486747742,
|
| 73 |
+
"lr": 3.6e-05,
|
| 74 |
+
"layer_weights": [
|
| 75 |
+
0.18804167211055756,
|
| 76 |
+
0.48665091395378113,
|
| 77 |
+
0.32530736923217773
|
| 78 |
+
]
|
| 79 |
+
},
|
| 80 |
+
{
|
| 81 |
+
"epoch": 7,
|
| 82 |
+
"train_loss": 0.5185237427552541,
|
| 83 |
+
"val_loss": 0.5066331923007965,
|
| 84 |
+
"val_mse": 0.3021552562713623,
|
| 85 |
+
"val_cos": 0.9837483167648315,
|
| 86 |
+
"lr": 4.2e-05,
|
| 87 |
+
"layer_weights": [
|
| 88 |
+
0.18802917003631592,
|
| 89 |
+
0.48659923672676086,
|
| 90 |
+
0.3253715932369232
|
| 91 |
+
]
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"epoch": 8,
|
| 95 |
+
"train_loss": 0.5167387425899506,
|
| 96 |
+
"val_loss": 0.5039149075746536,
|
| 97 |
+
"val_mse": 0.3002340942621231,
|
| 98 |
+
"val_cos": 0.9791701436042786,
|
| 99 |
+
"lr": 4.8e-05,
|
| 100 |
+
"layer_weights": [
|
| 101 |
+
0.18801729381084442,
|
| 102 |
+
0.4865494668483734,
|
| 103 |
+
0.32543328404426575
|
| 104 |
+
]
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 9,
|
| 108 |
+
"train_loss": 0.512136181195577,
|
| 109 |
+
"val_loss": 0.5010921210050583,
|
| 110 |
+
"val_mse": 0.2990950047969818,
|
| 111 |
+
"val_cos": 0.9724186956882477,
|
| 112 |
+
"lr": 4.998932514657232e-05,
|
| 113 |
+
"layer_weights": [
|
| 114 |
+
0.1879856437444687,
|
| 115 |
+
0.4865161180496216,
|
| 116 |
+
0.3254982531070709
|
| 117 |
+
]
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"epoch": 10,
|
| 121 |
+
"train_loss": 0.5113767782847086,
|
| 122 |
+
"val_loss": 0.49832451343536377,
|
| 123 |
+
"val_mse": 0.2983831316232681,
|
| 124 |
+
"val_cos": 0.96485435962677,
|
| 125 |
+
"lr": 4.9933307091588796e-05,
|
| 126 |
+
"layer_weights": [
|
| 127 |
+
0.1879592388868332,
|
| 128 |
+
0.48649585247039795,
|
| 129 |
+
0.32554492354393005
|
| 130 |
+
]
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 11,
|
| 134 |
+
"train_loss": 0.506663978099823,
|
| 135 |
+
"val_loss": 0.4950540214776993,
|
| 136 |
+
"val_mse": 0.2976083904504776,
|
| 137 |
+
"val_cos": 0.9557604789733887,
|
| 138 |
+
"lr": 4.982938460687583e-05,
|
| 139 |
+
"layer_weights": [
|
| 140 |
+
0.18791629374027252,
|
| 141 |
+
0.48649364709854126,
|
| 142 |
+
0.3255901038646698
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"epoch": 12,
|
| 147 |
+
"train_loss": 0.5058708017071089,
|
| 148 |
+
"val_loss": 0.4917283356189728,
|
| 149 |
+
"val_mse": 0.296984001994133,
|
| 150 |
+
"val_cos": 0.9461317956447601,
|
| 151 |
+
"lr": 4.967775735898179e-05,
|
| 152 |
+
"layer_weights": [
|
| 153 |
+
0.18786482512950897,
|
| 154 |
+
0.4865027368068695,
|
| 155 |
+
0.3256324827671051
|
| 156 |
+
]
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"epoch": 13,
|
| 160 |
+
"train_loss": 0.5017159804701805,
|
| 161 |
+
"val_loss": 0.488487184047699,
|
| 162 |
+
"val_mse": 0.29643353819847107,
|
| 163 |
+
"val_cos": 0.9366123378276825,
|
| 164 |
+
"lr": 4.947871666974437e-05,
|
| 165 |
+
"layer_weights": [
|
| 166 |
+
0.1877976357936859,
|
| 167 |
+
0.4865366518497467,
|
| 168 |
+
0.32566574215888977
|
| 169 |
+
]
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 14,
|
| 173 |
+
"train_loss": 0.49695071826378506,
|
| 174 |
+
"val_loss": 0.48568475246429443,
|
| 175 |
+
"val_mse": 0.29597683250904083,
|
| 176 |
+
"val_cos": 0.928336501121521,
|
| 177 |
+
"lr": 4.923264495657319e-05,
|
| 178 |
+
"layer_weights": [
|
| 179 |
+
0.18773548305034637,
|
| 180 |
+
0.4865611493587494,
|
| 181 |
+
0.32570332288742065
|
| 182 |
+
]
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 15,
|
| 186 |
+
"train_loss": 0.49470319350560504,
|
| 187 |
+
"val_loss": 0.48278090357780457,
|
| 188 |
+
"val_mse": 0.2955007553100586,
|
| 189 |
+
"val_cos": 0.919767826795578,
|
| 190 |
+
"lr": 4.894001499771015e-05,
|
| 191 |
+
"layer_weights": [
|
| 192 |
+
0.1876676231622696,
|
| 193 |
+
0.48658791184425354,
|
| 194 |
+
0.3257444500923157
|
| 195 |
+
]
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 16,
|
| 199 |
+
"train_loss": 0.4919345850745837,
|
| 200 |
+
"val_loss": 0.4799541234970093,
|
| 201 |
+
"val_mse": 0.295034795999527,
|
| 202 |
+
"val_cos": 0.9114325046539307,
|
| 203 |
+
"lr": 4.86013890238794e-05,
|
| 204 |
+
"layer_weights": [
|
| 205 |
+
0.18757948279380798,
|
| 206 |
+
0.48664289712905884,
|
| 207 |
+
0.3257776200771332
|
| 208 |
+
]
|
| 209 |
+
},
|
| 210 |
+
{
|
| 211 |
+
"epoch": 17,
|
| 212 |
+
"train_loss": 0.4882906923691432,
|
| 213 |
+
"val_loss": 0.47724851965904236,
|
| 214 |
+
"val_mse": 0.29457418620586395,
|
| 215 |
+
"val_cos": 0.9034886062145233,
|
| 216 |
+
"lr": 4.821741763807186e-05,
|
| 217 |
+
"layer_weights": [
|
| 218 |
+
0.18750329315662384,
|
| 219 |
+
0.4866565465927124,
|
| 220 |
+
0.32584020495414734
|
| 221 |
+
]
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 18,
|
| 225 |
+
"train_loss": 0.4863167131940524,
|
| 226 |
+
"val_loss": 0.47495734691619873,
|
| 227 |
+
"val_mse": 0.2941738963127136,
|
| 228 |
+
"val_cos": 0.8967853486537933,
|
| 229 |
+
"lr": 4.778883856554004e-05,
|
| 230 |
+
"layer_weights": [
|
| 231 |
+
0.18742914497852325,
|
| 232 |
+
0.48668304085731506,
|
| 233 |
+
0.3258878290653229
|
| 234 |
+
]
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 19,
|
| 238 |
+
"train_loss": 0.48106920967499417,
|
| 239 |
+
"val_loss": 0.47280479967594147,
|
| 240 |
+
"val_mse": 0.2937912791967392,
|
| 241 |
+
"val_cos": 0.8905029296875,
|
| 242 |
+
"lr": 4.7316475236404454e-05,
|
| 243 |
+
"layer_weights": [
|
| 244 |
+
0.18736256659030914,
|
| 245 |
+
0.48669952154159546,
|
| 246 |
+
0.3259379267692566
|
| 247 |
+
]
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"epoch": 20,
|
| 251 |
+
"train_loss": 0.4799882446726163,
|
| 252 |
+
"val_loss": 0.4706132858991623,
|
| 253 |
+
"val_mse": 0.2933909595012665,
|
| 254 |
+
"val_cos": 0.8841319978237152,
|
| 255 |
+
"lr": 4.6801235203595195e-05,
|
| 256 |
+
"layer_weights": [
|
| 257 |
+
0.18728071451187134,
|
| 258 |
+
0.48676493763923645,
|
| 259 |
+
0.3259543478488922
|
| 260 |
+
]
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 21,
|
| 264 |
+
"train_loss": 0.4774326408902804,
|
| 265 |
+
"val_loss": 0.46855370700359344,
|
| 266 |
+
"val_mse": 0.29300159215927124,
|
| 267 |
+
"val_cos": 0.8781753182411194,
|
| 268 |
+
"lr": 4.624410839916798e-05,
|
| 269 |
+
"layer_weights": [
|
| 270 |
+
0.1872127205133438,
|
| 271 |
+
0.4868311583995819,
|
| 272 |
+
0.3259561061859131
|
| 273 |
+
]
|
| 274 |
+
},
|
| 275 |
+
{
|
| 276 |
+
"epoch": 22,
|
| 277 |
+
"train_loss": 0.4739230324824651,
|
| 278 |
+
"val_loss": 0.4666079431772232,
|
| 279 |
+
"val_mse": 0.29262329638004303,
|
| 280 |
+
"val_cos": 0.8725720942020416,
|
| 281 |
+
"lr": 4.564616523234511e-05,
|
| 282 |
+
"layer_weights": [
|
| 283 |
+
0.18713925778865814,
|
| 284 |
+
0.48689666390419006,
|
| 285 |
+
0.3259640634059906
|
| 286 |
+
]
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"epoch": 23,
|
| 290 |
+
"train_loss": 0.471984604994456,
|
| 291 |
+
"val_loss": 0.4646792411804199,
|
| 292 |
+
"val_mse": 0.29221203923225403,
|
| 293 |
+
"val_cos": 0.867102712392807,
|
| 294 |
+
"lr": 4.500855453293532e-05,
|
| 295 |
+
"layer_weights": [
|
| 296 |
+
0.1870780885219574,
|
| 297 |
+
0.48695069551467896,
|
| 298 |
+
0.3259712755680084
|
| 299 |
+
]
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"epoch": 24,
|
| 303 |
+
"train_loss": 0.47153498729070026,
|
| 304 |
+
"val_loss": 0.4629760682582855,
|
| 305 |
+
"val_mse": 0.29188139736652374,
|
| 306 |
+
"val_cos": 0.8621969223022461,
|
| 307 |
+
"lr": 4.433250134408401e-05,
|
| 308 |
+
"layer_weights": [
|
| 309 |
+
0.1870197206735611,
|
| 310 |
+
0.48700079321861267,
|
| 311 |
+
0.32597944140434265
|
| 312 |
+
]
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 25,
|
| 316 |
+
"train_loss": 0.4680219367146492,
|
| 317 |
+
"val_loss": 0.460899218916893,
|
| 318 |
+
"val_mse": 0.2914563864469528,
|
| 319 |
+
"val_cos": 0.8562657833099365,
|
| 320 |
+
"lr": 4.361930456859455e-05,
|
| 321 |
+
"layer_weights": [
|
| 322 |
+
0.1869507133960724,
|
| 323 |
+
0.4870733320713043,
|
| 324 |
+
0.3259759545326233
|
| 325 |
+
]
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 26,
|
| 329 |
+
"train_loss": 0.4661514386534691,
|
| 330 |
+
"val_loss": 0.4593184292316437,
|
| 331 |
+
"val_mse": 0.29106342792510986,
|
| 332 |
+
"val_cos": 0.8519134819507599,
|
| 333 |
+
"lr": 4.287033447334286e-05,
|
| 334 |
+
"layer_weights": [
|
| 335 |
+
0.18687810003757477,
|
| 336 |
+
0.4871317148208618,
|
| 337 |
+
0.3259901702404022
|
| 338 |
+
]
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 27,
|
| 342 |
+
"train_loss": 0.46488772084315616,
|
| 343 |
+
"val_loss": 0.457811176776886,
|
| 344 |
+
"val_mse": 0.2907838821411133,
|
| 345 |
+
"val_cos": 0.8475415408611298,
|
| 346 |
+
"lr": 4.208703005657999e-05,
|
| 347 |
+
"layer_weights": [
|
| 348 |
+
0.18682761490345,
|
| 349 |
+
0.48717233538627625,
|
| 350 |
+
0.32600003480911255
|
| 351 |
+
]
|
| 352 |
+
},
|
| 353 |
+
{
|
| 354 |
+
"epoch": 28,
|
| 355 |
+
"train_loss": 0.4627537876367569,
|
| 356 |
+
"val_loss": 0.4561166316270828,
|
| 357 |
+
"val_mse": 0.2903905361890793,
|
| 358 |
+
"val_cos": 0.8428108096122742,
|
| 359 |
+
"lr": 4.1270896283180896e-05,
|
| 360 |
+
"layer_weights": [
|
| 361 |
+
0.1867944747209549,
|
| 362 |
+
0.48722437024116516,
|
| 363 |
+
0.32598114013671875
|
| 364 |
+
]
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"epoch": 29,
|
| 368 |
+
"train_loss": 0.46068041771650314,
|
| 369 |
+
"val_loss": 0.4543476104736328,
|
| 370 |
+
"val_mse": 0.2899843454360962,
|
| 371 |
+
"val_cos": 0.8378618955612183,
|
| 372 |
+
"lr": 4.0423501193151416e-05,
|
| 373 |
+
"layer_weights": [
|
| 374 |
+
0.18675556778907776,
|
| 375 |
+
0.4872865080833435,
|
| 376 |
+
0.3259579837322235
|
| 377 |
+
]
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 30,
|
| 381 |
+
"train_loss": 0.4596499999364217,
|
| 382 |
+
"val_loss": 0.4527883976697922,
|
| 383 |
+
"val_mse": 0.2896469980478287,
|
| 384 |
+
"val_cos": 0.83345165848732,
|
| 385 |
+
"lr": 3.954647288894883e-05,
|
| 386 |
+
"layer_weights": [
|
| 387 |
+
0.18669842183589935,
|
| 388 |
+
0.48734673857688904,
|
| 389 |
+
0.3259548842906952
|
| 390 |
+
]
|
| 391 |
+
},
|
| 392 |
+
{
|
| 393 |
+
"epoch": 31,
|
| 394 |
+
"train_loss": 0.45678772032260895,
|
| 395 |
+
"val_loss": 0.450978547334671,
|
| 396 |
+
"val_mse": 0.28921158611774445,
|
| 397 |
+
"val_cos": 0.8284347951412201,
|
| 398 |
+
"lr": 3.864149640740417e-05,
|
| 399 |
+
"layer_weights": [
|
| 400 |
+
0.18662679195404053,
|
| 401 |
+
0.48743027448654175,
|
| 402 |
+
0.3259429335594177
|
| 403 |
+
]
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 32,
|
| 407 |
+
"train_loss": 0.4548913861314456,
|
| 408 |
+
"val_loss": 0.4496942460536957,
|
| 409 |
+
"val_mse": 0.2888818830251694,
|
| 410 |
+
"val_cos": 0.8249230980873108,
|
| 411 |
+
"lr": 3.7710310482256526e-05,
|
| 412 |
+
"layer_weights": [
|
| 413 |
+
0.1865749955177307,
|
| 414 |
+
0.48748308420181274,
|
| 415 |
+
0.32594192028045654
|
| 416 |
+
]
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 33,
|
| 420 |
+
"train_loss": 0.4544881780942281,
|
| 421 |
+
"val_loss": 0.44842529296875,
|
| 422 |
+
"val_mse": 0.28857842087745667,
|
| 423 |
+
"val_cos": 0.8214012980461121,
|
| 424 |
+
"lr": 3.675470420351921e-05,
|
| 425 |
+
"layer_weights": [
|
| 426 |
+
0.18652117252349854,
|
| 427 |
+
0.48755523562431335,
|
| 428 |
+
0.3259236216545105
|
| 429 |
+
]
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 34,
|
| 433 |
+
"train_loss": 0.4538674329717954,
|
| 434 |
+
"val_loss": 0.4470330774784088,
|
| 435 |
+
"val_mse": 0.288252592086792,
|
| 436 |
+
"val_cos": 0.8175208568572998,
|
| 437 |
+
"lr": 3.5776513580096315e-05,
|
| 438 |
+
"layer_weights": [
|
| 439 |
+
0.18647873401641846,
|
| 440 |
+
0.48761439323425293,
|
| 441 |
+
0.3259068429470062
|
| 442 |
+
]
|
| 443 |
+
},
|
| 444 |
+
{
|
| 445 |
+
"epoch": 35,
|
| 446 |
+
"train_loss": 0.4523365447918574,
|
| 447 |
+
"val_loss": 0.44579267501831055,
|
| 448 |
+
"val_mse": 0.28796349465847015,
|
| 449 |
+
"val_cos": 0.8140607476234436,
|
| 450 |
+
"lr": 3.47776180122539e-05,
|
| 451 |
+
"layer_weights": [
|
| 452 |
+
0.18644081056118011,
|
| 453 |
+
0.4876665472984314,
|
| 454 |
+
0.32589268684387207
|
| 455 |
+
]
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"epoch": 36,
|
| 459 |
+
"train_loss": 0.4503123164176941,
|
| 460 |
+
"val_loss": 0.44486451148986816,
|
| 461 |
+
"val_mse": 0.28770390152931213,
|
| 462 |
+
"val_cos": 0.8115726113319397,
|
| 463 |
+
"lr": 3.375993668072324e-05,
|
| 464 |
+
"layer_weights": [
|
| 465 |
+
0.18638741970062256,
|
| 466 |
+
0.4877305328845978,
|
| 467 |
+
0.32588207721710205
|
| 468 |
+
]
|
| 469 |
+
},
|
| 470 |
+
{
|
| 471 |
+
"epoch": 37,
|
| 472 |
+
"train_loss": 0.4485436826944351,
|
| 473 |
+
"val_loss": 0.44352900981903076,
|
| 474 |
+
"val_mse": 0.2873896658420563,
|
| 475 |
+
"val_cos": 0.8078540861606598,
|
| 476 |
+
"lr": 3.272542485937369e-05,
|
| 477 |
+
"layer_weights": [
|
| 478 |
+
0.18635866045951843,
|
| 479 |
+
0.48779621720314026,
|
| 480 |
+
0.3258451223373413
|
| 481 |
+
]
|
| 482 |
+
},
|
| 483 |
+
{
|
| 484 |
+
"epoch": 38,
|
| 485 |
+
"train_loss": 0.4482949525117874,
|
| 486 |
+
"val_loss": 0.4422347843647003,
|
| 487 |
+
"val_mse": 0.28710322082042694,
|
| 488 |
+
"val_cos": 0.8042083978652954,
|
| 489 |
+
"lr": 3.1676070158539825e-05,
|
| 490 |
+
"layer_weights": [
|
| 491 |
+
0.18631692230701447,
|
| 492 |
+
0.4878581166267395,
|
| 493 |
+
0.32582494616508484
|
| 494 |
+
]
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 39,
|
| 498 |
+
"train_loss": 0.4465568463007609,
|
| 499 |
+
"val_loss": 0.4412134736776352,
|
| 500 |
+
"val_mse": 0.286797359585762,
|
| 501 |
+
"val_cos": 0.8015177249908447,
|
| 502 |
+
"lr": 3.0613888706220336e-05,
|
| 503 |
+
"layer_weights": [
|
| 504 |
+
0.18628276884555817,
|
| 505 |
+
0.4878806471824646,
|
| 506 |
+
0.32583656907081604
|
| 507 |
+
]
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 40,
|
| 511 |
+
"train_loss": 0.4467907374103864,
|
| 512 |
+
"val_loss": 0.4401810020208359,
|
| 513 |
+
"val_mse": 0.2865670323371887,
|
| 514 |
+
"val_cos": 0.7986135482788086,
|
| 515 |
+
"lr": 2.954092127448591e-05,
|
| 516 |
+
"layer_weights": [
|
| 517 |
+
0.18625368177890778,
|
| 518 |
+
0.4879050552845001,
|
| 519 |
+
0.32584133744239807
|
| 520 |
+
]
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 41,
|
| 524 |
+
"train_loss": 0.4456794833143552,
|
| 525 |
+
"val_loss": 0.4392598569393158,
|
| 526 |
+
"val_mse": 0.28631071746349335,
|
| 527 |
+
"val_cos": 0.7961412072181702,
|
| 528 |
+
"lr": 2.8459229358538407e-05,
|
| 529 |
+
"layer_weights": [
|
| 530 |
+
0.18623541295528412,
|
| 531 |
+
0.4879443645477295,
|
| 532 |
+
0.3258202373981476
|
| 533 |
+
]
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 42,
|
| 537 |
+
"train_loss": 0.44552704443534213,
|
| 538 |
+
"val_loss": 0.4383067339658737,
|
| 539 |
+
"val_mse": 0.2860827147960663,
|
| 540 |
+
"val_cos": 0.7934961020946503,
|
| 541 |
+
"lr": 2.7370891215954568e-05,
|
| 542 |
+
"layer_weights": [
|
| 543 |
+
0.18620307743549347,
|
| 544 |
+
0.4880008101463318,
|
| 545 |
+
0.32579612731933594
|
| 546 |
+
]
|
| 547 |
+
},
|
| 548 |
+
{
|
| 549 |
+
"epoch": 43,
|
| 550 |
+
"train_loss": 0.4442807783683141,
|
| 551 |
+
"val_loss": 0.43739478290081024,
|
| 552 |
+
"val_mse": 0.28582078218460083,
|
| 553 |
+
"val_cos": 0.7910674512386322,
|
| 554 |
+
"lr": 2.6277997873724182e-05,
|
| 555 |
+
"layer_weights": [
|
| 556 |
+
0.18618352711200714,
|
| 557 |
+
0.48801717162132263,
|
| 558 |
+
0.3257993459701538
|
| 559 |
+
]
|
| 560 |
+
},
|
| 561 |
+
{
|
| 562 |
+
"epoch": 44,
|
| 563 |
+
"train_loss": 0.44318560759226483,
|
| 564 |
+
"val_loss": 0.4364076852798462,
|
| 565 |
+
"val_mse": 0.2855755388736725,
|
| 566 |
+
"val_cos": 0.7883493006229401,
|
| 567 |
+
"lr": 2.5182649110754324e-05,
|
| 568 |
+
"layer_weights": [
|
| 569 |
+
0.1861647516489029,
|
| 570 |
+
0.4880358576774597,
|
| 571 |
+
0.32579949498176575
|
| 572 |
+
]
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 45,
|
| 576 |
+
"train_loss": 0.4432884429891904,
|
| 577 |
+
"val_loss": 0.4354058504104614,
|
| 578 |
+
"val_mse": 0.2853531688451767,
|
| 579 |
+
"val_cos": 0.7855288088321686,
|
| 580 |
+
"lr": 2.4086949423558526e-05,
|
| 581 |
+
"layer_weights": [
|
| 582 |
+
0.18614305555820465,
|
| 583 |
+
0.48807674646377563,
|
| 584 |
+
0.3257801830768585
|
| 585 |
+
]
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 46,
|
| 589 |
+
"train_loss": 0.44228366017341614,
|
| 590 |
+
"val_loss": 0.4349256455898285,
|
| 591 |
+
"val_mse": 0.28515173494815826,
|
| 592 |
+
"val_cos": 0.784398078918457,
|
| 593 |
+
"lr": 2.2993003982881975e-05,
|
| 594 |
+
"layer_weights": [
|
| 595 |
+
0.18610072135925293,
|
| 596 |
+
0.4881318211555481,
|
| 597 |
+
0.3257673978805542
|
| 598 |
+
]
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"epoch": 47,
|
| 602 |
+
"train_loss": 0.4409554104010264,
|
| 603 |
+
"val_loss": 0.4342317283153534,
|
| 604 |
+
"val_mse": 0.2850031703710556,
|
| 605 |
+
"val_cos": 0.7824316620826721,
|
| 606 |
+
"lr": 2.19029145890313e-05,
|
| 607 |
+
"layer_weights": [
|
| 608 |
+
0.18606385588645935,
|
| 609 |
+
0.48818841576576233,
|
| 610 |
+
0.32574766874313354
|
| 611 |
+
]
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 48,
|
| 615 |
+
"train_loss": 0.4404771775007248,
|
| 616 |
+
"val_loss": 0.4333122968673706,
|
| 617 |
+
"val_mse": 0.28479498624801636,
|
| 618 |
+
"val_cos": 0.7798527181148529,
|
| 619 |
+
"lr": 2.081877563368006e-05,
|
| 620 |
+
"layer_weights": [
|
| 621 |
+
0.18604432046413422,
|
| 622 |
+
0.48823291063308716,
|
| 623 |
+
0.32572272419929504
|
| 624 |
+
]
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 49,
|
| 628 |
+
"train_loss": 0.44003237038850784,
|
| 629 |
+
"val_loss": 0.43286144733428955,
|
| 630 |
+
"val_mse": 0.2846349775791168,
|
| 631 |
+
"val_cos": 0.7787232100963593,
|
| 632 |
+
"lr": 1.974267007590835e-05,
|
| 633 |
+
"layer_weights": [
|
| 634 |
+
0.18603070080280304,
|
| 635 |
+
0.4882526397705078,
|
| 636 |
+
0.3257165849208832
|
| 637 |
+
]
|
| 638 |
+
},
|
| 639 |
+
{
|
| 640 |
+
"epoch": 50,
|
| 641 |
+
"train_loss": 0.4389326920111974,
|
| 642 |
+
"val_loss": 0.4319637417793274,
|
| 643 |
+
"val_mse": 0.2844291925430298,
|
| 644 |
+
"val_cos": 0.7762110233306885,
|
| 645 |
+
"lr": 1.867666544020798e-05,
|
| 646 |
+
"layer_weights": [
|
| 647 |
+
0.18601518869400024,
|
| 648 |
+
0.488275945186615,
|
| 649 |
+
0.32570886611938477
|
| 650 |
+
]
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 51,
|
| 654 |
+
"train_loss": 0.43824509531259537,
|
| 655 |
+
"val_loss": 0.4313496947288513,
|
| 656 |
+
"val_mse": 0.2842629551887512,
|
| 657 |
+
"val_cos": 0.7745520770549774,
|
| 658 |
+
"lr": 1.7622809844142137e-05,
|
| 659 |
+
"layer_weights": [
|
| 660 |
+
0.18598434329032898,
|
| 661 |
+
0.48831209540367126,
|
| 662 |
+
0.3257036507129669
|
| 663 |
+
]
|
| 664 |
+
},
|
| 665 |
+
{
|
| 666 |
+
"epoch": 52,
|
| 667 |
+
"train_loss": 0.43857457985480625,
|
| 668 |
+
"val_loss": 0.4309784919023514,
|
| 669 |
+
"val_mse": 0.2840999513864517,
|
| 670 |
+
"val_cos": 0.7736950814723969,
|
| 671 |
+
"lr": 1.6583128063291576e-05,
|
| 672 |
+
"layer_weights": [
|
| 673 |
+
0.1859699934720993,
|
| 674 |
+
0.48831743001937866,
|
| 675 |
+
0.32571253180503845
|
| 676 |
+
]
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 53,
|
| 680 |
+
"train_loss": 0.43759987751642865,
|
| 681 |
+
"val_loss": 0.4305753856897354,
|
| 682 |
+
"val_mse": 0.28399814665317535,
|
| 683 |
+
"val_cos": 0.7725889086723328,
|
| 684 |
+
"lr": 1.5559617641047886e-05,
|
| 685 |
+
"layer_weights": [
|
| 686 |
+
0.185959592461586,
|
| 687 |
+
0.488329142332077,
|
| 688 |
+
0.3257112503051758
|
| 689 |
+
]
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 54,
|
| 693 |
+
"train_loss": 0.4366108253598213,
|
| 694 |
+
"val_loss": 0.4301539361476898,
|
| 695 |
+
"val_mse": 0.2838665097951889,
|
| 696 |
+
"val_cos": 0.7714912593364716,
|
| 697 |
+
"lr": 1.4554245050728085e-05,
|
| 698 |
+
"layer_weights": [
|
| 699 |
+
0.1859455555677414,
|
| 700 |
+
0.4883480966091156,
|
| 701 |
+
0.3257063329219818
|
| 702 |
+
]
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 55,
|
| 706 |
+
"train_loss": 0.43590281655391055,
|
| 707 |
+
"val_loss": 0.4297642707824707,
|
| 708 |
+
"val_mse": 0.28371478617191315,
|
| 709 |
+
"val_cos": 0.7705464065074921,
|
| 710 |
+
"lr": 1.3568941917384036e-05,
|
| 711 |
+
"layer_weights": [
|
| 712 |
+
0.1859353631734848,
|
| 713 |
+
0.48836269974708557,
|
| 714 |
+
0.3257019817829132
|
| 715 |
+
]
|
| 716 |
+
},
|
| 717 |
+
{
|
| 718 |
+
"epoch": 56,
|
| 719 |
+
"train_loss": 0.4365125621358554,
|
| 720 |
+
"val_loss": 0.429342657327652,
|
| 721 |
+
"val_mse": 0.28359802067279816,
|
| 722 |
+
"val_cos": 0.769413411617279,
|
| 723 |
+
"lr": 1.2605601306566205e-05,
|
| 724 |
+
"layer_weights": [
|
| 725 |
+
0.1859167218208313,
|
| 726 |
+
0.4883984625339508,
|
| 727 |
+
0.3256847858428955
|
| 728 |
+
]
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 57,
|
| 732 |
+
"train_loss": 0.43557916829983395,
|
| 733 |
+
"val_loss": 0.4290514290332794,
|
| 734 |
+
"val_mse": 0.2834867388010025,
|
| 735 |
+
"val_cos": 0.768702358007431,
|
| 736 |
+
"lr": 1.1666074087171627e-05,
|
| 737 |
+
"layer_weights": [
|
| 738 |
+
0.18590416014194489,
|
| 739 |
+
0.4884169399738312,
|
| 740 |
+
0.32567888498306274
|
| 741 |
+
]
|
| 742 |
+
},
|
| 743 |
+
{
|
| 744 |
+
"epoch": 58,
|
| 745 |
+
"train_loss": 0.43708648284276325,
|
| 746 |
+
"val_loss": 0.4289564788341522,
|
| 747 |
+
"val_mse": 0.28341861069202423,
|
| 748 |
+
"val_cos": 0.7685448229312897,
|
| 749 |
+
"lr": 1.0752165375364593e-05,
|
| 750 |
+
"layer_weights": [
|
| 751 |
+
0.18589580059051514,
|
| 752 |
+
0.48841819167137146,
|
| 753 |
+
0.32568594813346863
|
| 754 |
+
]
|
| 755 |
+
},
|
| 756 |
+
{
|
| 757 |
+
"epoch": 59,
|
| 758 |
+
"train_loss": 0.43515104552110034,
|
| 759 |
+
"val_loss": 0.4286029487848282,
|
| 760 |
+
"val_mse": 0.2833453267812729,
|
| 761 |
+
"val_cos": 0.767537385225296,
|
| 762 |
+
"lr": 9.865631066402137e-06,
|
| 763 |
+
"layer_weights": [
|
| 764 |
+
0.18588438630104065,
|
| 765 |
+
0.4884311556816101,
|
| 766 |
+
0.32568442821502686
|
| 767 |
+
]
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 60,
|
| 771 |
+
"train_loss": 0.43382184704144794,
|
| 772 |
+
"val_loss": 0.4281291216611862,
|
| 773 |
+
"val_mse": 0.2832487225532532,
|
| 774 |
+
"val_cos": 0.7661833763122559,
|
| 775 |
+
"lr": 9.008174461027724e-06,
|
| 776 |
+
"layer_weights": [
|
| 777 |
+
0.18587639927864075,
|
| 778 |
+
0.4884514808654785,
|
| 779 |
+
0.3256721496582031
|
| 780 |
+
]
|
| 781 |
+
},
|
| 782 |
+
{
|
| 783 |
+
"epoch": 61,
|
| 784 |
+
"train_loss": 0.43559310336907703,
|
| 785 |
+
"val_loss": 0.4276932626962662,
|
| 786 |
+
"val_mse": 0.2831578105688095,
|
| 787 |
+
"val_cos": 0.7649426162242889,
|
| 788 |
+
"lr": 8.181442992915e-06,
|
| 789 |
+
"layer_weights": [
|
| 790 |
+
0.18586839735507965,
|
| 791 |
+
0.48846176266670227,
|
| 792 |
+
0.3256698548793793
|
| 793 |
+
]
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 62,
|
| 797 |
+
"train_loss": 0.4351862221956253,
|
| 798 |
+
"val_loss": 0.42760802805423737,
|
| 799 |
+
"val_mse": 0.2830965518951416,
|
| 800 |
+
"val_cos": 0.7648014426231384,
|
| 801 |
+
"lr": 7.387025063449082e-06,
|
| 802 |
+
"layer_weights": [
|
| 803 |
+
0.18585878610610962,
|
| 804 |
+
0.4884742796421051,
|
| 805 |
+
0.3256669044494629
|
| 806 |
+
]
|
| 807 |
+
},
|
| 808 |
+
{
|
| 809 |
+
"epoch": 63,
|
| 810 |
+
"train_loss": 0.4348592087626457,
|
| 811 |
+
"val_loss": 0.42764292657375336,
|
| 812 |
+
"val_mse": 0.28304730355739594,
|
| 813 |
+
"val_cos": 0.7650326788425446,
|
| 814 |
+
"lr": 6.626446989926652e-06,
|
| 815 |
+
"layer_weights": [
|
| 816 |
+
0.18584749102592468,
|
| 817 |
+
0.4884894788265228,
|
| 818 |
+
0.3256630599498749
|
| 819 |
+
]
|
| 820 |
+
},
|
| 821 |
+
{
|
| 822 |
+
"epoch": 64,
|
| 823 |
+
"train_loss": 0.43365951875845593,
|
| 824 |
+
"val_loss": 0.4275168776512146,
|
| 825 |
+
"val_mse": 0.2829972058534622,
|
| 826 |
+
"val_cos": 0.7647294402122498,
|
| 827 |
+
"lr": 5.901170073038523e-06,
|
| 828 |
+
"layer_weights": [
|
| 829 |
+
0.18583695590496063,
|
| 830 |
+
0.4884980618953705,
|
| 831 |
+
0.32566502690315247
|
| 832 |
+
]
|
| 833 |
+
},
|
| 834 |
+
{
|
| 835 |
+
"epoch": 65,
|
| 836 |
+
"train_loss": 0.43526748071114224,
|
| 837 |
+
"val_loss": 0.4272725284099579,
|
| 838 |
+
"val_mse": 0.2829395681619644,
|
| 839 |
+
"val_cos": 0.7640494108200073,
|
| 840 |
+
"lr": 5.2125877892686496e-06,
|
| 841 |
+
"layer_weights": [
|
| 842 |
+
0.18583163619041443,
|
| 843 |
+
0.4885026216506958,
|
| 844 |
+
0.32566580176353455
|
| 845 |
+
]
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 66,
|
| 849 |
+
"train_loss": 0.4331621627012889,
|
| 850 |
+
"val_loss": 0.42718012630939484,
|
| 851 |
+
"val_mse": 0.28290486335754395,
|
| 852 |
+
"val_cos": 0.7638223767280579,
|
| 853 |
+
"lr": 4.562023113604041e-06,
|
| 854 |
+
"layer_weights": [
|
| 855 |
+
0.1858270764350891,
|
| 856 |
+
0.488506942987442,
|
| 857 |
+
0.32566604018211365
|
| 858 |
+
]
|
| 859 |
+
},
|
| 860 |
+
{
|
| 861 |
+
"epoch": 67,
|
| 862 |
+
"train_loss": 0.43357910464207333,
|
| 863 |
+
"val_loss": 0.42704926431179047,
|
| 864 |
+
"val_mse": 0.2828690856695175,
|
| 865 |
+
"val_cos": 0.7634696364402771,
|
| 866 |
+
"lr": 3.950725977699396e-06,
|
| 867 |
+
"layer_weights": [
|
| 868 |
+
0.1858239322900772,
|
| 869 |
+
0.48851150274276733,
|
| 870 |
+
0.3256644904613495
|
| 871 |
+
]
|
| 872 |
+
},
|
| 873 |
+
{
|
| 874 |
+
"epoch": 68,
|
| 875 |
+
"train_loss": 0.4325304701924324,
|
| 876 |
+
"val_loss": 0.4268566370010376,
|
| 877 |
+
"val_mse": 0.28282904624938965,
|
| 878 |
+
"val_cos": 0.7629209756851196,
|
| 879 |
+
"lr": 3.3798708683800305e-06,
|
| 880 |
+
"layer_weights": [
|
| 881 |
+
0.18581977486610413,
|
| 882 |
+
0.48851752281188965,
|
| 883 |
+
0.32566267251968384
|
| 884 |
+
]
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 69,
|
| 888 |
+
"train_loss": 0.43360988547404605,
|
| 889 |
+
"val_loss": 0.42667947709560394,
|
| 890 |
+
"val_mse": 0.2827933728694916,
|
| 891 |
+
"val_cos": 0.7624136805534363,
|
| 892 |
+
"lr": 2.850554571097211e-06,
|
| 893 |
+
"layer_weights": [
|
| 894 |
+
0.18581505119800568,
|
| 895 |
+
0.4885236620903015,
|
| 896 |
+
0.3256613314151764
|
| 897 |
+
]
|
| 898 |
+
},
|
| 899 |
+
{
|
| 900 |
+
"epoch": 70,
|
| 901 |
+
"train_loss": 0.43410984923442203,
|
| 902 |
+
"val_loss": 0.42659904062747955,
|
| 903 |
+
"val_mse": 0.2827706038951874,
|
| 904 |
+
"val_cos": 0.7621987164020538,
|
| 905 |
+
"lr": 2.3637940626713346e-06,
|
| 906 |
+
"layer_weights": [
|
| 907 |
+
0.18581299483776093,
|
| 908 |
+
0.4885265827178955,
|
| 909 |
+
0.3256604075431824
|
| 910 |
+
]
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"epoch": 71,
|
| 914 |
+
"train_loss": 0.4316417450706164,
|
| 915 |
+
"val_loss": 0.4265323132276535,
|
| 916 |
+
"val_mse": 0.2827492654323578,
|
| 917 |
+
"val_cos": 0.7620260715484619,
|
| 918 |
+
"lr": 1.9205245573716197e-06,
|
| 919 |
+
"layer_weights": [
|
| 920 |
+
0.18581034243106842,
|
| 921 |
+
0.48852840065956116,
|
| 922 |
+
0.325661301612854
|
| 923 |
+
]
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"epoch": 72,
|
| 927 |
+
"train_loss": 0.4345224474867185,
|
| 928 |
+
"val_loss": 0.42647726833820343,
|
| 929 |
+
"val_mse": 0.2827341854572296,
|
| 930 |
+
"val_cos": 0.7618777751922607,
|
| 931 |
+
"lr": 1.5215977100864392e-06,
|
| 932 |
+
"layer_weights": [
|
| 933 |
+
0.18580801784992218,
|
| 934 |
+
0.48853209614753723,
|
| 935 |
+
0.3256599009037018
|
| 936 |
+
]
|
| 937 |
+
},
|
| 938 |
+
{
|
| 939 |
+
"epoch": 73,
|
| 940 |
+
"train_loss": 0.4332921927173932,
|
| 941 |
+
"val_loss": 0.4264543503522873,
|
| 942 |
+
"val_mse": 0.28272223472595215,
|
| 943 |
+
"val_cos": 0.761829286813736,
|
| 944 |
+
"lr": 1.1677799800364958e-06,
|
| 945 |
+
"layer_weights": [
|
| 946 |
+
0.1858067810535431,
|
| 947 |
+
0.4885338544845581,
|
| 948 |
+
0.3256593346595764
|
| 949 |
+
]
|
| 950 |
+
},
|
| 951 |
+
{
|
| 952 |
+
"epoch": 74,
|
| 953 |
+
"train_loss": 0.43199672053257626,
|
| 954 |
+
"val_loss": 0.42647068202495575,
|
| 955 |
+
"val_mse": 0.28271588683128357,
|
| 956 |
+
"val_cos": 0.7618985176086426,
|
| 957 |
+
"lr": 8.597511581746626e-07,
|
| 958 |
+
"layer_weights": [
|
| 959 |
+
0.18580596148967743,
|
| 960 |
+
0.48853540420532227,
|
| 961 |
+
0.3256585896015167
|
| 962 |
+
]
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 75,
|
| 966 |
+
"train_loss": 0.4340783307949702,
|
| 967 |
+
"val_loss": 0.426461398601532,
|
| 968 |
+
"val_mse": 0.2827081084251404,
|
| 969 |
+
"val_cos": 0.7618857026100159,
|
| 970 |
+
"lr": 5.981030611018234e-07,
|
| 971 |
+
"layer_weights": [
|
| 972 |
+
0.1858053356409073,
|
| 973 |
+
0.48853573203086853,
|
| 974 |
+
0.3256588876247406
|
| 975 |
+
]
|
| 976 |
+
},
|
| 977 |
+
{
|
| 978 |
+
"epoch": 76,
|
| 979 |
+
"train_loss": 0.43331120908260345,
|
| 980 |
+
"val_loss": 0.42645110189914703,
|
| 981 |
+
"val_mse": 0.28270475566387177,
|
| 982 |
+
"val_cos": 0.7618592083454132,
|
| 983 |
+
"lr": 3.833383940080232e-07,
|
| 984 |
+
"layer_weights": [
|
| 985 |
+
0.18580502271652222,
|
| 986 |
+
0.4885362386703491,
|
| 987 |
+
0.32565873861312866
|
| 988 |
+
]
|
| 989 |
+
},
|
| 990 |
+
{
|
| 991 |
+
"epoch": 77,
|
| 992 |
+
"train_loss": 0.43248408287763596,
|
| 993 |
+
"val_loss": 0.42645207047462463,
|
| 994 |
+
"val_mse": 0.28270311653614044,
|
| 995 |
+
"val_cos": 0.7618663012981415,
|
| 996 |
+
"lr": 2.158697848236607e-07,
|
| 997 |
+
"layer_weights": [
|
| 998 |
+
0.18580475449562073,
|
| 999 |
+
0.48853710293769836,
|
| 1000 |
+
0.3256581723690033
|
| 1001 |
+
]
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 78,
|
| 1005 |
+
"train_loss": 0.43264390528202057,
|
| 1006 |
+
"val_loss": 0.4264531582593918,
|
| 1007 |
+
"val_mse": 0.2827022522687912,
|
| 1008 |
+
"val_cos": 0.7618719339370728,
|
| 1009 |
+
"lr": 9.60189914363363e-08,
|
| 1010 |
+
"layer_weights": [
|
| 1011 |
+
0.18580472469329834,
|
| 1012 |
+
0.48853710293769836,
|
| 1013 |
+
0.32565808296203613
|
| 1014 |
+
]
|
| 1015 |
+
},
|
| 1016 |
+
{
|
| 1017 |
+
"epoch": 79,
|
| 1018 |
+
"train_loss": 0.4327772284547488,
|
| 1019 |
+
"val_loss": 0.4264516681432724,
|
| 1020 |
+
"val_mse": 0.28270184993743896,
|
| 1021 |
+
"val_cos": 0.7618679404258728,
|
| 1022 |
+
"lr": 2.4016283496544613e-08,
|
| 1023 |
+
"layer_weights": [
|
| 1024 |
+
0.18580475449562073,
|
| 1025 |
+
0.4885372221469879,
|
| 1026 |
+
0.32565805315971375
|
| 1027 |
+
]
|
| 1028 |
+
},
|
| 1029 |
+
{
|
| 1030 |
+
"epoch": 80,
|
| 1031 |
+
"train_loss": 0.43244554350773495,
|
| 1032 |
+
"val_loss": 0.42645132541656494,
|
| 1033 |
+
"val_mse": 0.282701775431633,
|
| 1034 |
+
"val_cos": 0.7618669271469116,
|
| 1035 |
+
"lr": 0.0,
|
| 1036 |
+
"layer_weights": [
|
| 1037 |
+
0.18580475449562073,
|
| 1038 |
+
0.4885372221469879,
|
| 1039 |
+
0.32565805315971375
|
| 1040 |
+
]
|
| 1041 |
+
}
|
| 1042 |
+
]
|