Update model configuration
Browse files- config.json +4 -234
config.json
CHANGED
|
@@ -1,42 +1,4 @@
|
|
| 1 |
{
|
| 2 |
-
"architectures": [
|
| 3 |
-
"DeepONet"
|
| 4 |
-
],
|
| 5 |
-
"framework": "pytorch",
|
| 6 |
-
"model_name": "DeepONet",
|
| 7 |
-
"model_type": "deeponet",
|
| 8 |
-
"source_config": "config/config.yaml",
|
| 9 |
-
"project": {
|
| 10 |
-
"name": "DeepONet",
|
| 11 |
-
"paper": "DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators",
|
| 12 |
-
"arxiv": "1910.03193v3",
|
| 13 |
-
"paper_url": "https://arxiv.org/abs/1910.03193",
|
| 14 |
-
"implementation_policy": "paper_first_independent_reimplementation",
|
| 15 |
-
"paper_scale": true
|
| 16 |
-
},
|
| 17 |
-
"paths": {
|
| 18 |
-
"checkpoint": "weight/best_model.pth",
|
| 19 |
-
"results": "results",
|
| 20 |
-
"cache": "results/cache"
|
| 21 |
-
},
|
| 22 |
-
"runtime": {
|
| 23 |
-
"seed": 0,
|
| 24 |
-
"device": "auto",
|
| 25 |
-
"dtype": "float32",
|
| 26 |
-
"num_workers": 0
|
| 27 |
-
},
|
| 28 |
-
"function_space": {
|
| 29 |
-
"type": "grf",
|
| 30 |
-
"length_scale": 0.2,
|
| 31 |
-
"grf_grid_size": 1000,
|
| 32 |
-
"jitter": 1e-13,
|
| 33 |
-
"interpolation": "cubic",
|
| 34 |
-
"generation_chunk_size": 128,
|
| 35 |
-
"chebyshev": {
|
| 36 |
-
"degree": 10,
|
| 37 |
-
"coefficient_bound": 1.0
|
| 38 |
-
}
|
| 39 |
-
},
|
| 40 |
"model_defaults": {
|
| 41 |
"architecture": "deeponet",
|
| 42 |
"activation": "relu",
|
|
@@ -79,222 +41,30 @@
|
|
| 79 |
"output_bias": true
|
| 80 |
}
|
| 81 |
},
|
| 82 |
-
"training_defaults": {
|
| 83 |
-
"optimizer": "adam",
|
| 84 |
-
"learning_rate": 0.001,
|
| 85 |
-
"batch_size": null,
|
| 86 |
-
"evaluation_batch_size": 8192,
|
| 87 |
-
"weight_decay": 0.0,
|
| 88 |
-
"print_every": 1000,
|
| 89 |
-
"evaluate_every": 1000,
|
| 90 |
-
"checkpoint_metric": "test_mse",
|
| 91 |
-
"resume": false
|
| 92 |
-
},
|
| 93 |
-
"solver_defaults": {
|
| 94 |
-
"method": "RK45",
|
| 95 |
-
"rtol": 1e-07,
|
| 96 |
-
"atol": 1e-09,
|
| 97 |
-
"pde_newton_tolerance": 1e-10,
|
| 98 |
-
"pde_newton_max_iterations": 20
|
| 99 |
-
},
|
| 100 |
"experiments": {
|
| 101 |
"antiderivative": {
|
| 102 |
-
"equation": "ds/dx = u(x), s(0) = 0",
|
| 103 |
"sensor_points": 100,
|
| 104 |
"trunk_dim": 1,
|
| 105 |
-
"
|
| 106 |
-
"train_size": 10000,
|
| 107 |
-
"test_size": 100000,
|
| 108 |
-
"iterations": 50000,
|
| 109 |
-
"default_variant": "unstacked_bias",
|
| 110 |
-
"sweeps": {
|
| 111 |
-
"architectures": [
|
| 112 |
-
"fnn",
|
| 113 |
-
"unstacked_no_bias",
|
| 114 |
-
"unstacked_bias",
|
| 115 |
-
"stacked_no_bias",
|
| 116 |
-
"stacked_bias"
|
| 117 |
-
],
|
| 118 |
-
"fnn_depths": [
|
| 119 |
-
2,
|
| 120 |
-
3,
|
| 121 |
-
4
|
| 122 |
-
],
|
| 123 |
-
"fnn_widths": [
|
| 124 |
-
20,
|
| 125 |
-
40,
|
| 126 |
-
100
|
| 127 |
-
],
|
| 128 |
-
"learning_rates": [
|
| 129 |
-
0.0001,
|
| 130 |
-
0.001,
|
| 131 |
-
0.01
|
| 132 |
-
]
|
| 133 |
-
}
|
| 134 |
},
|
| 135 |
"nonlinear_ode": {
|
| 136 |
-
"equation": "ds/dx = -s^2 + u(x), s(0) = 0",
|
| 137 |
"sensor_points": 100,
|
| 138 |
"trunk_dim": 1,
|
| 139 |
-
"
|
| 140 |
-
"train_size": 10000,
|
| 141 |
-
"test_size": 100000,
|
| 142 |
-
"iterations": 100000,
|
| 143 |
-
"default_variant": "unstacked_bias",
|
| 144 |
-
"trim_fraction": 0.001,
|
| 145 |
-
"ood_functions": [
|
| 146 |
-
"linear",
|
| 147 |
-
"sin_pi",
|
| 148 |
-
"sin_2pi"
|
| 149 |
-
],
|
| 150 |
-
"sweeps": {
|
| 151 |
-
"architectures": [
|
| 152 |
-
"stacked_no_bias",
|
| 153 |
-
"stacked_bias",
|
| 154 |
-
"unstacked_no_bias",
|
| 155 |
-
"unstacked_bias"
|
| 156 |
-
]
|
| 157 |
-
}
|
| 158 |
},
|
| 159 |
"pendulum": {
|
| 160 |
-
"equation": "ds1/dt = s2; ds2/dt = -k*sin(s1) + u(t); s(0) = (0,0)",
|
| 161 |
"sensor_points": 100,
|
| 162 |
"trunk_dim": 1,
|
| 163 |
-
"
|
| 164 |
-
"k": 1.0,
|
| 165 |
-
"train_size": 10000,
|
| 166 |
-
"test_size": 100000,
|
| 167 |
-
"iterations": 100000,
|
| 168 |
-
"default_variant": "unstacked_bias",
|
| 169 |
-
"convergence_iterations": 500000,
|
| 170 |
-
"convergence_test_size": 1000000,
|
| 171 |
-
"sweeps": {
|
| 172 |
-
"sensor_points": [
|
| 173 |
-
10,
|
| 174 |
-
20,
|
| 175 |
-
50,
|
| 176 |
-
100,
|
| 177 |
-
200
|
| 178 |
-
],
|
| 179 |
-
"domain_end": [
|
| 180 |
-
1.0,
|
| 181 |
-
2.0,
|
| 182 |
-
3.0,
|
| 183 |
-
4.0
|
| 184 |
-
],
|
| 185 |
-
"length_scale": [
|
| 186 |
-
0.1,
|
| 187 |
-
0.2,
|
| 188 |
-
0.5,
|
| 189 |
-
1.0
|
| 190 |
-
],
|
| 191 |
-
"k": [
|
| 192 |
-
0.5,
|
| 193 |
-
1.0,
|
| 194 |
-
2.0
|
| 195 |
-
],
|
| 196 |
-
"width": [
|
| 197 |
-
20,
|
| 198 |
-
40,
|
| 199 |
-
100
|
| 200 |
-
],
|
| 201 |
-
"train_size": [
|
| 202 |
-
1000,
|
| 203 |
-
5000,
|
| 204 |
-
10000
|
| 205 |
-
],
|
| 206 |
-
"chebyshev_degree": [
|
| 207 |
-
5,
|
| 208 |
-
10,
|
| 209 |
-
20
|
| 210 |
-
]
|
| 211 |
-
}
|
| 212 |
},
|
| 213 |
"diffusion_reaction": {
|
| 214 |
-
"equation": "s_t = D*s_xx + k*s^2 + u(x)",
|
| 215 |
"sensor_points": 100,
|
| 216 |
"trunk_dim": 2,
|
| 217 |
-
"domain_end": 1.0,
|
| 218 |
-
"train_functions": 100,
|
| 219 |
-
"points_per_function": 1000,
|
| 220 |
-
"test_functions": 1000,
|
| 221 |
-
"test_points_per_function": 1000,
|
| 222 |
-
"test_size": 1000000,
|
| 223 |
-
"iterations": 500000,
|
| 224 |
"default_variant": "unstacked_bias",
|
| 225 |
-
"diffusion": 0.01,
|
| 226 |
-
"reaction": 0.01,
|
| 227 |
-
"space_points": 100,
|
| 228 |
-
"time_points": 100,
|
| 229 |
"branch_depth": 2,
|
| 230 |
"trunk_depth": 3,
|
| 231 |
"width": 100,
|
| 232 |
-
"latent_dim": 100
|
| 233 |
-
"sweeps": {
|
| 234 |
-
"train_functions": [
|
| 235 |
-
50,
|
| 236 |
-
100,
|
| 237 |
-
200,
|
| 238 |
-
400
|
| 239 |
-
],
|
| 240 |
-
"points_per_function": [
|
| 241 |
-
10,
|
| 242 |
-
100,
|
| 243 |
-
1000,
|
| 244 |
-
10000
|
| 245 |
-
]
|
| 246 |
-
}
|
| 247 |
-
}
|
| 248 |
-
},
|
| 249 |
-
"inference": {
|
| 250 |
-
"batch_size": 8192,
|
| 251 |
-
"ood_query_points": 100,
|
| 252 |
-
"save_inputs": true
|
| 253 |
-
},
|
| 254 |
-
"smoke_test": {
|
| 255 |
-
"project": {
|
| 256 |
-
"paper_scale": false
|
| 257 |
-
},
|
| 258 |
-
"function_space": {
|
| 259 |
-
"grf_grid_size": 32,
|
| 260 |
-
"generation_chunk_size": 16
|
| 261 |
-
},
|
| 262 |
-
"training_defaults": {
|
| 263 |
-
"batch_size": 16,
|
| 264 |
-
"evaluation_batch_size": 32,
|
| 265 |
-
"print_every": 1,
|
| 266 |
-
"evaluate_every": 1
|
| 267 |
-
},
|
| 268 |
-
"experiments": {
|
| 269 |
-
"antiderivative": {
|
| 270 |
-
"sensor_points": 16,
|
| 271 |
-
"train_size": 32,
|
| 272 |
-
"test_size": 24,
|
| 273 |
-
"iterations": 2
|
| 274 |
-
},
|
| 275 |
-
"nonlinear_ode": {
|
| 276 |
-
"sensor_points": 16,
|
| 277 |
-
"train_size": 24,
|
| 278 |
-
"test_size": 16,
|
| 279 |
-
"iterations": 2
|
| 280 |
-
},
|
| 281 |
-
"pendulum": {
|
| 282 |
-
"sensor_points": 16,
|
| 283 |
-
"train_size": 24,
|
| 284 |
-
"test_size": 16,
|
| 285 |
-
"iterations": 2
|
| 286 |
-
},
|
| 287 |
-
"diffusion_reaction": {
|
| 288 |
-
"sensor_points": 16,
|
| 289 |
-
"train_functions": 4,
|
| 290 |
-
"points_per_function": 8,
|
| 291 |
-
"test_functions": 2,
|
| 292 |
-
"test_points_per_function": 8,
|
| 293 |
-
"test_size": 16,
|
| 294 |
-
"space_points": 12,
|
| 295 |
-
"time_points": 12,
|
| 296 |
-
"iterations": 2
|
| 297 |
-
}
|
| 298 |
}
|
| 299 |
}
|
| 300 |
}
|
|
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"model_defaults": {
|
| 3 |
"architecture": "deeponet",
|
| 4 |
"activation": "relu",
|
|
|
|
| 41 |
"output_bias": true
|
| 42 |
}
|
| 43 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
"experiments": {
|
| 45 |
"antiderivative": {
|
|
|
|
| 46 |
"sensor_points": 100,
|
| 47 |
"trunk_dim": 1,
|
| 48 |
+
"default_variant": "unstacked_bias"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
},
|
| 50 |
"nonlinear_ode": {
|
|
|
|
| 51 |
"sensor_points": 100,
|
| 52 |
"trunk_dim": 1,
|
| 53 |
+
"default_variant": "unstacked_bias"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
},
|
| 55 |
"pendulum": {
|
|
|
|
| 56 |
"sensor_points": 100,
|
| 57 |
"trunk_dim": 1,
|
| 58 |
+
"default_variant": "unstacked_bias"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
},
|
| 60 |
"diffusion_reaction": {
|
|
|
|
| 61 |
"sensor_points": 100,
|
| 62 |
"trunk_dim": 2,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
"default_variant": "unstacked_bias",
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
"branch_depth": 2,
|
| 65 |
"trunk_depth": 3,
|
| 66 |
"width": 100,
|
| 67 |
+
"latent_dim": 100
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
}
|
| 69 |
}
|
| 70 |
}
|