Upload FIMODE
Browse files- config.json +46 -12
config.json
CHANGED
|
@@ -6,22 +6,30 @@
|
|
| 6 |
"dim_model": 512,
|
| 7 |
"dim_time": 512,
|
| 8 |
"dropout": 0.1,
|
|
|
|
| 9 |
"num_encoder_blocks": 4,
|
| 10 |
"num_heads": 8,
|
| 11 |
"residual_mlp": {
|
| 12 |
"dropout": 0.1,
|
| 13 |
-
"hidden_act": {
|
|
|
|
|
|
|
| 14 |
"hidden_layers": [
|
| 15 |
1024
|
| 16 |
],
|
| 17 |
"in_features": 512,
|
|
|
|
| 18 |
"out_features": 512,
|
| 19 |
-
"output_act": {
|
|
|
|
|
|
|
| 20 |
}
|
| 21 |
},
|
| 22 |
"combiner_net": {
|
| 23 |
"dropout": 0.1,
|
| 24 |
-
"hidden_act": {
|
|
|
|
|
|
|
| 25 |
"hidden_layers": [
|
| 26 |
1024,
|
| 27 |
1024,
|
|
@@ -29,12 +37,17 @@
|
|
| 29 |
1024
|
| 30 |
],
|
| 31 |
"in_features": 1024,
|
|
|
|
| 32 |
"out_features": 512,
|
| 33 |
-
"output_act": {
|
|
|
|
|
|
|
| 34 |
},
|
| 35 |
"init_cond_net": {
|
| 36 |
"dropout": 0.1,
|
| 37 |
-
"hidden_act": {
|
|
|
|
|
|
|
| 38 |
"hidden_layers": [
|
| 39 |
1024,
|
| 40 |
1024,
|
|
@@ -42,24 +55,37 @@
|
|
| 42 |
1024
|
| 43 |
],
|
| 44 |
"in_features": 513,
|
|
|
|
| 45 |
"out_features": 2,
|
| 46 |
-
"output_act": {
|
|
|
|
|
|
|
| 47 |
},
|
| 48 |
"load_in_8bit": false,
|
| 49 |
"loss_configs": {
|
|
|
|
|
|
|
|
|
|
| 50 |
"ode_solver": "rk4"
|
| 51 |
},
|
| 52 |
"model_type": "fimode",
|
| 53 |
-
"normalization_time": {
|
| 54 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
"time_encoding": {
|
|
|
|
| 56 |
"out_features": 512
|
| 57 |
},
|
| 58 |
"torch_dtype": "float32",
|
| 59 |
"transformers_version": "4.51.3",
|
| 60 |
"trunk_net": {
|
| 61 |
"dropout": 0.1,
|
| 62 |
-
"hidden_act": {
|
|
|
|
|
|
|
| 63 |
"hidden_layers": [
|
| 64 |
1024,
|
| 65 |
1024,
|
|
@@ -67,16 +93,24 @@
|
|
| 67 |
1024
|
| 68 |
],
|
| 69 |
"in_features": 512,
|
|
|
|
| 70 |
"out_features": 512,
|
| 71 |
-
"output_act": {
|
|
|
|
|
|
|
| 72 |
},
|
| 73 |
"use_bf16": false,
|
| 74 |
"vector_field_net": {
|
| 75 |
"dropout": 0.1,
|
| 76 |
-
"hidden_act": {
|
|
|
|
|
|
|
| 77 |
"hidden_layers": [],
|
| 78 |
"in_features": 512,
|
|
|
|
| 79 |
"out_features": 2,
|
| 80 |
-
"output_act": {
|
|
|
|
|
|
|
| 81 |
}
|
| 82 |
}
|
|
|
|
| 6 |
"dim_model": 512,
|
| 7 |
"dim_time": 512,
|
| 8 |
"dropout": 0.1,
|
| 9 |
+
"name": "fim.models.blocks.Transformer",
|
| 10 |
"num_encoder_blocks": 4,
|
| 11 |
"num_heads": 8,
|
| 12 |
"residual_mlp": {
|
| 13 |
"dropout": 0.1,
|
| 14 |
+
"hidden_act": {
|
| 15 |
+
"name": "torch.nn.SELU"
|
| 16 |
+
},
|
| 17 |
"hidden_layers": [
|
| 18 |
1024
|
| 19 |
],
|
| 20 |
"in_features": 512,
|
| 21 |
+
"name": "fim.models.blocks.MLP",
|
| 22 |
"out_features": 512,
|
| 23 |
+
"output_act": {
|
| 24 |
+
"name": "torch.nn.Identity"
|
| 25 |
+
}
|
| 26 |
}
|
| 27 |
},
|
| 28 |
"combiner_net": {
|
| 29 |
"dropout": 0.1,
|
| 30 |
+
"hidden_act": {
|
| 31 |
+
"name": "torch.nn.SELU"
|
| 32 |
+
},
|
| 33 |
"hidden_layers": [
|
| 34 |
1024,
|
| 35 |
1024,
|
|
|
|
| 37 |
1024
|
| 38 |
],
|
| 39 |
"in_features": 1024,
|
| 40 |
+
"name": "fim.models.blocks.MLP",
|
| 41 |
"out_features": 512,
|
| 42 |
+
"output_act": {
|
| 43 |
+
"name": "torch.nn.Identity"
|
| 44 |
+
}
|
| 45 |
},
|
| 46 |
"init_cond_net": {
|
| 47 |
"dropout": 0.1,
|
| 48 |
+
"hidden_act": {
|
| 49 |
+
"name": "torch.nn.SELU"
|
| 50 |
+
},
|
| 51 |
"hidden_layers": [
|
| 52 |
1024,
|
| 53 |
1024,
|
|
|
|
| 55 |
1024
|
| 56 |
],
|
| 57 |
"in_features": 513,
|
| 58 |
+
"name": "fim.models.blocks.MLP",
|
| 59 |
"out_features": 2,
|
| 60 |
+
"output_act": {
|
| 61 |
+
"name": "torch.nn.Identity"
|
| 62 |
+
}
|
| 63 |
},
|
| 64 |
"load_in_8bit": false,
|
| 65 |
"loss_configs": {
|
| 66 |
+
"loss_scale_drift": 1.0,
|
| 67 |
+
"loss_scale_init_cond": 1.0,
|
| 68 |
+
"loss_scale_unsuperv_loss": 10.0,
|
| 69 |
"ode_solver": "rk4"
|
| 70 |
},
|
| 71 |
"model_type": "fimode",
|
| 72 |
+
"normalization_time": {
|
| 73 |
+
"name": "fim.models.blocks.MinMaxNormalization"
|
| 74 |
+
},
|
| 75 |
+
"normalization_values": {
|
| 76 |
+
"name": "fim.models.blocks.MinMaxNormalization"
|
| 77 |
+
},
|
| 78 |
"time_encoding": {
|
| 79 |
+
"name": "fim.models.blocks.SineTimeEncoding",
|
| 80 |
"out_features": 512
|
| 81 |
},
|
| 82 |
"torch_dtype": "float32",
|
| 83 |
"transformers_version": "4.51.3",
|
| 84 |
"trunk_net": {
|
| 85 |
"dropout": 0.1,
|
| 86 |
+
"hidden_act": {
|
| 87 |
+
"name": "torch.nn.SELU"
|
| 88 |
+
},
|
| 89 |
"hidden_layers": [
|
| 90 |
1024,
|
| 91 |
1024,
|
|
|
|
| 93 |
1024
|
| 94 |
],
|
| 95 |
"in_features": 512,
|
| 96 |
+
"name": "fim.models.blocks.MLP",
|
| 97 |
"out_features": 512,
|
| 98 |
+
"output_act": {
|
| 99 |
+
"name": "torch.nn.Identity"
|
| 100 |
+
}
|
| 101 |
},
|
| 102 |
"use_bf16": false,
|
| 103 |
"vector_field_net": {
|
| 104 |
"dropout": 0.1,
|
| 105 |
+
"hidden_act": {
|
| 106 |
+
"name": "torch.nn.SELU"
|
| 107 |
+
},
|
| 108 |
"hidden_layers": [],
|
| 109 |
"in_features": 512,
|
| 110 |
+
"name": "fim.models.blocks.MLP",
|
| 111 |
"out_features": 2,
|
| 112 |
+
"output_act": {
|
| 113 |
+
"name": "torch.nn.Identity"
|
| 114 |
+
}
|
| 115 |
}
|
| 116 |
}
|