Add v6 results
Browse files- results_v6.json +382 -0
results_v6.json
ADDED
|
@@ -0,0 +1,382 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"main_tasks": {
|
| 3 |
+
"Complex Fn (4D)": {
|
| 4 |
+
"Vanilla": {
|
| 5 |
+
"mean": 0.0574857605000337,
|
| 6 |
+
"std": 0.008335319220438129,
|
| 7 |
+
"scores": [
|
| 8 |
+
0.0631193146109581,
|
| 9 |
+
0.06363634765148163,
|
| 10 |
+
0.04570161923766136
|
| 11 |
+
],
|
| 12 |
+
"params": 4993,
|
| 13 |
+
"hidden": 48
|
| 14 |
+
},
|
| 15 |
+
"SinGLU": {
|
| 16 |
+
"mean": 0.014336802531033754,
|
| 17 |
+
"std": 0.011309525106379352,
|
| 18 |
+
"scores": [
|
| 19 |
+
0.03023100085556507,
|
| 20 |
+
0.007935304194688797,
|
| 21 |
+
0.004844102542847395
|
| 22 |
+
],
|
| 23 |
+
"params": 4811,
|
| 24 |
+
"hidden": 31
|
| 25 |
+
},
|
| 26 |
+
"Hybrid": {
|
| 27 |
+
"mean": 0.07689925034840901,
|
| 28 |
+
"std": 0.02610708296796532,
|
| 29 |
+
"scores": [
|
| 30 |
+
0.06204311549663544,
|
| 31 |
+
0.11359916627407074,
|
| 32 |
+
0.055055469274520874
|
| 33 |
+
],
|
| 34 |
+
"params": 4959,
|
| 35 |
+
"hidden": 34
|
| 36 |
+
},
|
| 37 |
+
"Adaptive": {
|
| 38 |
+
"mean": 2.157564640045166,
|
| 39 |
+
"std": 0.06025844793099204,
|
| 40 |
+
"scores": [
|
| 41 |
+
2.085890531539917,
|
| 42 |
+
2.233323335647583,
|
| 43 |
+
2.153480052947998
|
| 44 |
+
],
|
| 45 |
+
"params": 5099,
|
| 46 |
+
"hidden": 33
|
| 47 |
+
}
|
| 48 |
+
},
|
| 49 |
+
"Nested Fn (2D)": {
|
| 50 |
+
"Vanilla": {
|
| 51 |
+
"mean": 0.0486922413110733,
|
| 52 |
+
"std": 0.0030859308547071134,
|
| 53 |
+
"scores": [
|
| 54 |
+
0.0464077852666378,
|
| 55 |
+
0.05305477976799011,
|
| 56 |
+
0.046614158898591995
|
| 57 |
+
],
|
| 58 |
+
"params": 2961,
|
| 59 |
+
"hidden": 37
|
| 60 |
+
},
|
| 61 |
+
"SinGLU": {
|
| 62 |
+
"mean": 0.00021879897879747054,
|
| 63 |
+
"std": 9.755789869089583e-05,
|
| 64 |
+
"scores": [
|
| 65 |
+
8.296288433484733e-05,
|
| 66 |
+
0.00030763907125219703,
|
| 67 |
+
0.00026579498080536723
|
| 68 |
+
],
|
| 69 |
+
"params": 2993,
|
| 70 |
+
"hidden": 24
|
| 71 |
+
},
|
| 72 |
+
"Hybrid": {
|
| 73 |
+
"mean": 0.001271639756547908,
|
| 74 |
+
"std": 0.000170726739254775,
|
| 75 |
+
"scores": [
|
| 76 |
+
0.0013282388681545854,
|
| 77 |
+
0.0010400698520243168,
|
| 78 |
+
0.0014466105494648218
|
| 79 |
+
],
|
| 80 |
+
"params": 2959,
|
| 81 |
+
"hidden": 26
|
| 82 |
+
},
|
| 83 |
+
"Adaptive": {
|
| 84 |
+
"mean": 0.19985504945119223,
|
| 85 |
+
"std": 0.015309994965033512,
|
| 86 |
+
"scores": [
|
| 87 |
+
0.2051493227481842,
|
| 88 |
+
0.1790262758731842,
|
| 89 |
+
0.21538954973220825
|
| 90 |
+
],
|
| 91 |
+
"params": 3009,
|
| 92 |
+
"hidden": 25
|
| 93 |
+
}
|
| 94 |
+
},
|
| 95 |
+
"Spiral": {
|
| 96 |
+
"Vanilla": {
|
| 97 |
+
"mean": 0.851111114025116,
|
| 98 |
+
"std": 0.12812993960807575,
|
| 99 |
+
"scores": [
|
| 100 |
+
0.9366666674613953,
|
| 101 |
+
0.6700000166893005,
|
| 102 |
+
0.9466666579246521
|
| 103 |
+
],
|
| 104 |
+
"params": 2999,
|
| 105 |
+
"hidden": 37
|
| 106 |
+
},
|
| 107 |
+
"SinGLU": {
|
| 108 |
+
"mean": 0.4422222177187602,
|
| 109 |
+
"std": 0.020427534068557608,
|
| 110 |
+
"scores": [
|
| 111 |
+
0.46666666865348816,
|
| 112 |
+
0.4166666567325592,
|
| 113 |
+
0.44333332777023315
|
| 114 |
+
],
|
| 115 |
+
"params": 3018,
|
| 116 |
+
"hidden": 24
|
| 117 |
+
},
|
| 118 |
+
"Hybrid": {
|
| 119 |
+
"mean": 1.0,
|
| 120 |
+
"std": 0.0,
|
| 121 |
+
"scores": [
|
| 122 |
+
1.0,
|
| 123 |
+
1.0,
|
| 124 |
+
1.0
|
| 125 |
+
],
|
| 126 |
+
"params": 2986,
|
| 127 |
+
"hidden": 26
|
| 128 |
+
},
|
| 129 |
+
"Adaptive": {
|
| 130 |
+
"mean": 0.7211111187934875,
|
| 131 |
+
"std": 0.046374760288990124,
|
| 132 |
+
"scores": [
|
| 133 |
+
0.6866666674613953,
|
| 134 |
+
0.7866666913032532,
|
| 135 |
+
0.6899999976158142
|
| 136 |
+
],
|
| 137 |
+
"params": 3035,
|
| 138 |
+
"hidden": 25
|
| 139 |
+
}
|
| 140 |
+
},
|
| 141 |
+
"Checkerboard": {
|
| 142 |
+
"Vanilla": {
|
| 143 |
+
"mean": 0.5788888931274414,
|
| 144 |
+
"std": 0.022825147689377753,
|
| 145 |
+
"scores": [
|
| 146 |
+
0.596666693687439,
|
| 147 |
+
0.54666668176651,
|
| 148 |
+
0.5933333039283752
|
| 149 |
+
],
|
| 150 |
+
"params": 2999,
|
| 151 |
+
"hidden": 37
|
| 152 |
+
},
|
| 153 |
+
"SinGLU": {
|
| 154 |
+
"mean": 0.9377777775128683,
|
| 155 |
+
"std": 0.014989694108291244,
|
| 156 |
+
"scores": [
|
| 157 |
+
0.9166666865348816,
|
| 158 |
+
0.9466666579246521,
|
| 159 |
+
0.949999988079071
|
| 160 |
+
],
|
| 161 |
+
"params": 3018,
|
| 162 |
+
"hidden": 24
|
| 163 |
+
},
|
| 164 |
+
"Hybrid": {
|
| 165 |
+
"mean": 0.9022222359975179,
|
| 166 |
+
"std": 0.02060804929952367,
|
| 167 |
+
"scores": [
|
| 168 |
+
0.9133333563804626,
|
| 169 |
+
0.8733333349227905,
|
| 170 |
+
0.9200000166893005
|
| 171 |
+
],
|
| 172 |
+
"params": 2986,
|
| 173 |
+
"hidden": 26
|
| 174 |
+
},
|
| 175 |
+
"Adaptive": {
|
| 176 |
+
"mean": 0.5888889034589132,
|
| 177 |
+
"std": 0.029979416198453995,
|
| 178 |
+
"scores": [
|
| 179 |
+
0.54666668176651,
|
| 180 |
+
0.6133333444595337,
|
| 181 |
+
0.6066666841506958
|
| 182 |
+
],
|
| 183 |
+
"params": 3035,
|
| 184 |
+
"hidden": 25
|
| 185 |
+
}
|
| 186 |
+
},
|
| 187 |
+
"High-Freq": {
|
| 188 |
+
"Vanilla": {
|
| 189 |
+
"mean": 1.099808136622111,
|
| 190 |
+
"std": 0.01846947085143113,
|
| 191 |
+
"scores": [
|
| 192 |
+
1.0741740465164185,
|
| 193 |
+
1.108283281326294,
|
| 194 |
+
1.1169670820236206
|
| 195 |
+
],
|
| 196 |
+
"params": 7999,
|
| 197 |
+
"hidden": 62
|
| 198 |
+
},
|
| 199 |
+
"SinGLU": {
|
| 200 |
+
"mean": 1.0171122153600056,
|
| 201 |
+
"std": 0.06186627648533991,
|
| 202 |
+
"scores": [
|
| 203 |
+
1.0643826723098755,
|
| 204 |
+
0.9297174215316772,
|
| 205 |
+
1.0572365522384644
|
| 206 |
+
],
|
| 207 |
+
"params": 8214,
|
| 208 |
+
"hidden": 41
|
| 209 |
+
},
|
| 210 |
+
"Hybrid": {
|
| 211 |
+
"mean": 1.1086766123771667,
|
| 212 |
+
"std": 0.2812712331673168,
|
| 213 |
+
"scores": [
|
| 214 |
+
0.8738470673561096,
|
| 215 |
+
1.5041409730911255,
|
| 216 |
+
0.9480417966842651
|
| 217 |
+
],
|
| 218 |
+
"params": 7997,
|
| 219 |
+
"hidden": 44
|
| 220 |
+
},
|
| 221 |
+
"Adaptive": {
|
| 222 |
+
"mean": 0.9636624654134115,
|
| 223 |
+
"std": 0.08904664624357356,
|
| 224 |
+
"scores": [
|
| 225 |
+
0.8458895087242126,
|
| 226 |
+
1.0611637830734253,
|
| 227 |
+
0.9839341044425964
|
| 228 |
+
],
|
| 229 |
+
"params": 8046,
|
| 230 |
+
"hidden": 43
|
| 231 |
+
}
|
| 232 |
+
},
|
| 233 |
+
"Memorization": {
|
| 234 |
+
"Vanilla": {
|
| 235 |
+
"mean": 0.15677066644032797,
|
| 236 |
+
"std": 0.010408440580230162,
|
| 237 |
+
"scores": [
|
| 238 |
+
0.1653069108724594,
|
| 239 |
+
0.14211732149124146,
|
| 240 |
+
0.16288776695728302
|
| 241 |
+
],
|
| 242 |
+
"params": 4926,
|
| 243 |
+
"hidden": 46
|
| 244 |
+
},
|
| 245 |
+
"SinGLU": {
|
| 246 |
+
"mean": 9.313488410119438e-07,
|
| 247 |
+
"std": 1.0663730164468857e-06,
|
| 248 |
+
"scores": [
|
| 249 |
+
1.7445064770527097e-07,
|
| 250 |
+
1.801714688554057e-07,
|
| 251 |
+
2.4394244064751547e-06
|
| 252 |
+
],
|
| 253 |
+
"params": 5067,
|
| 254 |
+
"hidden": 31
|
| 255 |
+
},
|
| 256 |
+
"Hybrid": {
|
| 257 |
+
"mean": 4.9525905675788335e-06,
|
| 258 |
+
"std": 4.7364751238755085e-06,
|
| 259 |
+
"scores": [
|
| 260 |
+
2.1999953787599225e-06,
|
| 261 |
+
1.1617433301580604e-05,
|
| 262 |
+
1.040343022395973e-06
|
| 263 |
+
],
|
| 264 |
+
"params": 4807,
|
| 265 |
+
"hidden": 32
|
| 266 |
+
},
|
| 267 |
+
"Adaptive": {
|
| 268 |
+
"mean": 1.2049987119233663e-09,
|
| 269 |
+
"std": 7.631317175211726e-10,
|
| 270 |
+
"scores": [
|
| 271 |
+
2.2784609754467056e-09,
|
| 272 |
+
7.647771860241903e-10,
|
| 273 |
+
5.717579742992029e-10
|
| 274 |
+
],
|
| 275 |
+
"params": 5124,
|
| 276 |
+
"hidden": 32
|
| 277 |
+
}
|
| 278 |
+
},
|
| 279 |
+
"OOD": {
|
| 280 |
+
"Vanilla": {
|
| 281 |
+
"mean": 1.5319531361262004,
|
| 282 |
+
"std": 0.2293290940636717,
|
| 283 |
+
"scores": [
|
| 284 |
+
1.5319531361262004
|
| 285 |
+
],
|
| 286 |
+
"params": 0,
|
| 287 |
+
"hidden": 0
|
| 288 |
+
},
|
| 289 |
+
"SinGLU": {
|
| 290 |
+
"mean": 5.897943178812663,
|
| 291 |
+
"std": 0.8500770525046415,
|
| 292 |
+
"scores": [
|
| 293 |
+
5.897943178812663
|
| 294 |
+
],
|
| 295 |
+
"params": 0,
|
| 296 |
+
"hidden": 0
|
| 297 |
+
},
|
| 298 |
+
"Hybrid": {
|
| 299 |
+
"mean": 4.738887468973796,
|
| 300 |
+
"std": 1.5966265517441278,
|
| 301 |
+
"scores": [
|
| 302 |
+
4.738887468973796
|
| 303 |
+
],
|
| 304 |
+
"params": 0,
|
| 305 |
+
"hidden": 0
|
| 306 |
+
},
|
| 307 |
+
"Adaptive": {
|
| 308 |
+
"mean": 5.617943445841472,
|
| 309 |
+
"std": 0.2315950615741721,
|
| 310 |
+
"scores": [
|
| 311 |
+
5.617943445841472
|
| 312 |
+
],
|
| 313 |
+
"params": 0,
|
| 314 |
+
"hidden": 0
|
| 315 |
+
}
|
| 316 |
+
}
|
| 317 |
+
},
|
| 318 |
+
"ood": {
|
| 319 |
+
"Vanilla": {
|
| 320 |
+
"id_mean": 0.21685118973255157,
|
| 321 |
+
"id_std": 0.011753120779597635,
|
| 322 |
+
"ood_mean": 1.5319531361262004,
|
| 323 |
+
"ood_std": 0.2293290940636717,
|
| 324 |
+
"params": 5097,
|
| 325 |
+
"degradation": 7.064536459383043
|
| 326 |
+
},
|
| 327 |
+
"SinGLU": {
|
| 328 |
+
"id_mean": 0.24567786852518717,
|
| 329 |
+
"id_std": 0.013796241102558409,
|
| 330 |
+
"ood_mean": 5.897943178812663,
|
| 331 |
+
"ood_std": 0.8500770525046415,
|
| 332 |
+
"params": 5109,
|
| 333 |
+
"degradation": 24.006815160918734
|
| 334 |
+
},
|
| 335 |
+
"Hybrid": {
|
| 336 |
+
"id_mean": 0.0039149638420591755,
|
| 337 |
+
"id_std": 0.0004443769118260595,
|
| 338 |
+
"ood_mean": 4.738887468973796,
|
| 339 |
+
"ood_std": 1.5966265517441278,
|
| 340 |
+
"params": 4819,
|
| 341 |
+
"degradation": 1210.4549774031263
|
| 342 |
+
},
|
| 343 |
+
"Adaptive": {
|
| 344 |
+
"id_mean": 0.3005977620681127,
|
| 345 |
+
"id_std": 0.04048296135061511,
|
| 346 |
+
"ood_mean": 5.617943445841472,
|
| 347 |
+
"ood_std": 0.2315950615741721,
|
| 348 |
+
"params": 5025,
|
| 349 |
+
"degradation": 18.689239092101083
|
| 350 |
+
}
|
| 351 |
+
},
|
| 352 |
+
"alpha_analysis": {
|
| 353 |
+
"Complex Fn (4D)": [
|
| 354 |
+
0.45252057909965515,
|
| 355 |
+
0.048741355538368225
|
| 356 |
+
],
|
| 357 |
+
"Nested Fn (2D)": [
|
| 358 |
+
0.4708990454673767,
|
| 359 |
+
0.06593184918165207
|
| 360 |
+
],
|
| 361 |
+
"Spiral": [
|
| 362 |
+
0.4788985252380371,
|
| 363 |
+
0.05437469482421875
|
| 364 |
+
],
|
| 365 |
+
"Checkerboard": [
|
| 366 |
+
0.4901539087295532,
|
| 367 |
+
0.04592375084757805
|
| 368 |
+
],
|
| 369 |
+
"High-Freq": [
|
| 370 |
+
0.4693261981010437,
|
| 371 |
+
0.06162083521485329
|
| 372 |
+
],
|
| 373 |
+
"Memorization": [
|
| 374 |
+
0.5054463148117065,
|
| 375 |
+
0.010797806084156036
|
| 376 |
+
]
|
| 377 |
+
},
|
| 378 |
+
"ood_alphas": {
|
| 379 |
+
"ID": 0.4717688262462616,
|
| 380 |
+
"OOD": 0.4758133590221405
|
| 381 |
+
}
|
| 382 |
+
}
|