Add model config
Browse files- config.json +10 -10
config.json
CHANGED
|
@@ -34,14 +34,14 @@
|
|
| 34 |
"initial_mi_beta": 1.0,
|
| 35 |
"final_mi_beta": 1.0,
|
| 36 |
"mi_beta_shape": "constant",
|
| 37 |
-
"initial_tc_beta":
|
| 38 |
"final_tc_beta": 10.0,
|
| 39 |
-
"tc_beta_shape": "
|
| 40 |
"initial_dw_beta": 0.2,
|
| 41 |
"final_dw_beta": 1.0,
|
| 42 |
"dw_beta_shape": "custom",
|
| 43 |
"warmup_epoch_ratio": 0.2,
|
| 44 |
-
"free_bits": 0.
|
| 45 |
"focal_loss_alpha": 0.5,
|
| 46 |
"focal_loss_gamma": 2.0
|
| 47 |
},
|
|
@@ -49,9 +49,9 @@
|
|
| 49 |
"initial_mi_beta": 1.0,
|
| 50 |
"final_mi_beta": 1.0,
|
| 51 |
"mi_beta_shape": "constant",
|
| 52 |
-
"initial_tc_beta":
|
| 53 |
"final_tc_beta": 10.0,
|
| 54 |
-
"tc_beta_shape": "
|
| 55 |
"initial_dw_beta": 0.2,
|
| 56 |
"final_dw_beta": 1.0,
|
| 57 |
"dw_beta_shape": "custom",
|
|
@@ -60,7 +60,7 @@
|
|
| 60 |
"regularization": {
|
| 61 |
"encoder_dropout": 0.1,
|
| 62 |
"decoder_dropout": 0.1,
|
| 63 |
-
"free_bits": 0.
|
| 64 |
"focal_loss_alpha": 0.5,
|
| 65 |
"focal_loss_gamma": 2.0
|
| 66 |
},
|
|
@@ -72,9 +72,9 @@
|
|
| 72 |
"best_epoch": null
|
| 73 |
}
|
| 74 |
},
|
| 75 |
-
"final_train_loss":
|
| 76 |
-
"final_test_loss":
|
| 77 |
-
"best_train_loss":
|
| 78 |
-
"best_test_loss":
|
| 79 |
"total_epochs": 15
|
| 80 |
}
|
|
|
|
| 34 |
"initial_mi_beta": 1.0,
|
| 35 |
"final_mi_beta": 1.0,
|
| 36 |
"mi_beta_shape": "constant",
|
| 37 |
+
"initial_tc_beta": 0.0,
|
| 38 |
"final_tc_beta": 10.0,
|
| 39 |
+
"tc_beta_shape": "custom",
|
| 40 |
"initial_dw_beta": 0.2,
|
| 41 |
"final_dw_beta": 1.0,
|
| 42 |
"dw_beta_shape": "custom",
|
| 43 |
"warmup_epoch_ratio": 0.2,
|
| 44 |
+
"free_bits": 0.75,
|
| 45 |
"focal_loss_alpha": 0.5,
|
| 46 |
"focal_loss_gamma": 2.0
|
| 47 |
},
|
|
|
|
| 49 |
"initial_mi_beta": 1.0,
|
| 50 |
"final_mi_beta": 1.0,
|
| 51 |
"mi_beta_shape": "constant",
|
| 52 |
+
"initial_tc_beta": 0.0,
|
| 53 |
"final_tc_beta": 10.0,
|
| 54 |
+
"tc_beta_shape": "custom",
|
| 55 |
"initial_dw_beta": 0.2,
|
| 56 |
"final_dw_beta": 1.0,
|
| 57 |
"dw_beta_shape": "custom",
|
|
|
|
| 60 |
"regularization": {
|
| 61 |
"encoder_dropout": 0.1,
|
| 62 |
"decoder_dropout": 0.1,
|
| 63 |
+
"free_bits": 0.75,
|
| 64 |
"focal_loss_alpha": 0.5,
|
| 65 |
"focal_loss_gamma": 2.0
|
| 66 |
},
|
|
|
|
| 72 |
"best_epoch": null
|
| 73 |
}
|
| 74 |
},
|
| 75 |
+
"final_train_loss": 980.7457598876953,
|
| 76 |
+
"final_test_loss": 990.1863067626953,
|
| 77 |
+
"best_train_loss": 980.7457598876953,
|
| 78 |
+
"best_test_loss": 985.553515625,
|
| 79 |
"total_epochs": 15
|
| 80 |
}
|