v12 results
Browse files- results_v12.json +304 -0
results_v12.json
ADDED
|
@@ -0,0 +1,304 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"tasks": {
|
| 3 |
+
"Complex Fn": {
|
| 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 |
+
"v12": {
|
| 27 |
+
"mean": 0.007541361652935545,
|
| 28 |
+
"std": 0.0027402010484466033,
|
| 29 |
+
"scores": [
|
| 30 |
+
0.003877621376886964,
|
| 31 |
+
0.008279701694846153,
|
| 32 |
+
0.010466761887073517
|
| 33 |
+
],
|
| 34 |
+
"params": 4915,
|
| 35 |
+
"hidden": 27
|
| 36 |
+
}
|
| 37 |
+
},
|
| 38 |
+
"Nested Fn": {
|
| 39 |
+
"Vanilla": {
|
| 40 |
+
"mean": 0.0486922413110733,
|
| 41 |
+
"std": 0.0030859308547071134,
|
| 42 |
+
"scores": [
|
| 43 |
+
0.0464077852666378,
|
| 44 |
+
0.05305477976799011,
|
| 45 |
+
0.046614158898591995
|
| 46 |
+
],
|
| 47 |
+
"params": 2961,
|
| 48 |
+
"hidden": 37
|
| 49 |
+
},
|
| 50 |
+
"SinGLU": {
|
| 51 |
+
"mean": 0.00021879897879747054,
|
| 52 |
+
"std": 9.755789869089583e-05,
|
| 53 |
+
"scores": [
|
| 54 |
+
8.296288433484733e-05,
|
| 55 |
+
0.00030763907125219703,
|
| 56 |
+
0.00026579498080536723
|
| 57 |
+
],
|
| 58 |
+
"params": 2993,
|
| 59 |
+
"hidden": 24
|
| 60 |
+
},
|
| 61 |
+
"v12": {
|
| 62 |
+
"mean": 0.000645058105874341,
|
| 63 |
+
"std": 0.0007162512363257481,
|
| 64 |
+
"scores": [
|
| 65 |
+
0.0001647444732952863,
|
| 66 |
+
0.00011288218229310587,
|
| 67 |
+
0.0016575476620346308
|
| 68 |
+
],
|
| 69 |
+
"params": 2983,
|
| 70 |
+
"hidden": 21
|
| 71 |
+
}
|
| 72 |
+
},
|
| 73 |
+
"Spiral": {
|
| 74 |
+
"Vanilla": {
|
| 75 |
+
"mean": 0.851111114025116,
|
| 76 |
+
"std": 0.12812993960807575,
|
| 77 |
+
"scores": [
|
| 78 |
+
0.9366666674613953,
|
| 79 |
+
0.6700000166893005,
|
| 80 |
+
0.9466666579246521
|
| 81 |
+
],
|
| 82 |
+
"params": 2999,
|
| 83 |
+
"hidden": 37
|
| 84 |
+
},
|
| 85 |
+
"SinGLU": {
|
| 86 |
+
"mean": 0.4422222177187602,
|
| 87 |
+
"std": 0.020427534068557608,
|
| 88 |
+
"scores": [
|
| 89 |
+
0.46666666865348816,
|
| 90 |
+
0.4166666567325592,
|
| 91 |
+
0.44333332777023315
|
| 92 |
+
],
|
| 93 |
+
"params": 3018,
|
| 94 |
+
"hidden": 24
|
| 95 |
+
},
|
| 96 |
+
"v12": {
|
| 97 |
+
"mean": 0.8677777647972107,
|
| 98 |
+
"std": 0.01911628554646968,
|
| 99 |
+
"scores": [
|
| 100 |
+
0.8700000047683716,
|
| 101 |
+
0.8899999856948853,
|
| 102 |
+
0.8433333039283752
|
| 103 |
+
],
|
| 104 |
+
"params": 3005,
|
| 105 |
+
"hidden": 21
|
| 106 |
+
}
|
| 107 |
+
},
|
| 108 |
+
"Checker": {
|
| 109 |
+
"Vanilla": {
|
| 110 |
+
"mean": 0.5788888931274414,
|
| 111 |
+
"std": 0.022825147689377753,
|
| 112 |
+
"scores": [
|
| 113 |
+
0.596666693687439,
|
| 114 |
+
0.54666668176651,
|
| 115 |
+
0.5933333039283752
|
| 116 |
+
],
|
| 117 |
+
"params": 2999,
|
| 118 |
+
"hidden": 37
|
| 119 |
+
},
|
| 120 |
+
"SinGLU": {
|
| 121 |
+
"mean": 0.9377777775128683,
|
| 122 |
+
"std": 0.014989694108291244,
|
| 123 |
+
"scores": [
|
| 124 |
+
0.9166666865348816,
|
| 125 |
+
0.9466666579246521,
|
| 126 |
+
0.949999988079071
|
| 127 |
+
],
|
| 128 |
+
"params": 3018,
|
| 129 |
+
"hidden": 24
|
| 130 |
+
},
|
| 131 |
+
"v12": {
|
| 132 |
+
"mean": 0.9477777679761251,
|
| 133 |
+
"std": 0.013698684720279257,
|
| 134 |
+
"scores": [
|
| 135 |
+
0.9300000071525574,
|
| 136 |
+
0.949999988079071,
|
| 137 |
+
0.9633333086967468
|
| 138 |
+
],
|
| 139 |
+
"params": 3005,
|
| 140 |
+
"hidden": 21
|
| 141 |
+
}
|
| 142 |
+
},
|
| 143 |
+
"High-Freq": {
|
| 144 |
+
"Vanilla": {
|
| 145 |
+
"mean": 1.099808136622111,
|
| 146 |
+
"std": 0.01846947085143113,
|
| 147 |
+
"scores": [
|
| 148 |
+
1.0741740465164185,
|
| 149 |
+
1.108283281326294,
|
| 150 |
+
1.1169670820236206
|
| 151 |
+
],
|
| 152 |
+
"params": 7999,
|
| 153 |
+
"hidden": 62
|
| 154 |
+
},
|
| 155 |
+
"SinGLU": {
|
| 156 |
+
"mean": 1.0171122153600056,
|
| 157 |
+
"std": 0.06186627648533991,
|
| 158 |
+
"scores": [
|
| 159 |
+
1.0643826723098755,
|
| 160 |
+
0.9297174215316772,
|
| 161 |
+
1.0572365522384644
|
| 162 |
+
],
|
| 163 |
+
"params": 8214,
|
| 164 |
+
"hidden": 41
|
| 165 |
+
},
|
| 166 |
+
"v12": {
|
| 167 |
+
"mean": 1.952195127805074,
|
| 168 |
+
"std": 0.3974538576816994,
|
| 169 |
+
"scores": [
|
| 170 |
+
2.416501998901367,
|
| 171 |
+
1.4456902742385864,
|
| 172 |
+
1.9943931102752686
|
| 173 |
+
],
|
| 174 |
+
"params": 7734,
|
| 175 |
+
"hidden": 35
|
| 176 |
+
}
|
| 177 |
+
},
|
| 178 |
+
"Memorize": {
|
| 179 |
+
"Vanilla": {
|
| 180 |
+
"mean": 0.15677066644032797,
|
| 181 |
+
"std": 0.010408440580230162,
|
| 182 |
+
"scores": [
|
| 183 |
+
0.1653069108724594,
|
| 184 |
+
0.14211732149124146,
|
| 185 |
+
0.16288776695728302
|
| 186 |
+
],
|
| 187 |
+
"params": 4926,
|
| 188 |
+
"hidden": 46
|
| 189 |
+
},
|
| 190 |
+
"SinGLU": {
|
| 191 |
+
"mean": 9.313488410119438e-07,
|
| 192 |
+
"std": 1.0663730164468857e-06,
|
| 193 |
+
"scores": [
|
| 194 |
+
1.7445064770527097e-07,
|
| 195 |
+
1.801714688554057e-07,
|
| 196 |
+
2.4394244064751547e-06
|
| 197 |
+
],
|
| 198 |
+
"params": 5067,
|
| 199 |
+
"hidden": 31
|
| 200 |
+
},
|
| 201 |
+
"v12": {
|
| 202 |
+
"mean": 2.1600329309497585e-05,
|
| 203 |
+
"std": 2.097512212913466e-05,
|
| 204 |
+
"scores": [
|
| 205 |
+
5.010054155718535e-05,
|
| 206 |
+
2.272288099902653e-07,
|
| 207 |
+
1.4473217561317142e-05
|
| 208 |
+
],
|
| 209 |
+
"params": 5215,
|
| 210 |
+
"hidden": 27
|
| 211 |
+
}
|
| 212 |
+
},
|
| 213 |
+
"OOD": {
|
| 214 |
+
"Vanilla": {
|
| 215 |
+
"mean": 1.5319531361262004,
|
| 216 |
+
"std": 0.2293290940636717,
|
| 217 |
+
"scores": [
|
| 218 |
+
1.5319531361262004
|
| 219 |
+
],
|
| 220 |
+
"params": 0,
|
| 221 |
+
"hidden": 0
|
| 222 |
+
},
|
| 223 |
+
"SinGLU": {
|
| 224 |
+
"mean": 5.897943178812663,
|
| 225 |
+
"std": 0.8500770525046415,
|
| 226 |
+
"scores": [
|
| 227 |
+
5.897943178812663
|
| 228 |
+
],
|
| 229 |
+
"params": 0,
|
| 230 |
+
"hidden": 0
|
| 231 |
+
},
|
| 232 |
+
"v12": {
|
| 233 |
+
"mean": 4.693021059036255,
|
| 234 |
+
"std": 0.9724661958608424,
|
| 235 |
+
"scores": [
|
| 236 |
+
4.693021059036255
|
| 237 |
+
],
|
| 238 |
+
"params": 0,
|
| 239 |
+
"hidden": 0
|
| 240 |
+
}
|
| 241 |
+
}
|
| 242 |
+
},
|
| 243 |
+
"ood": {
|
| 244 |
+
"Vanilla": {
|
| 245 |
+
"id": 0.21685118973255157,
|
| 246 |
+
"ood": 1.5319531361262004,
|
| 247 |
+
"deg": 7.064536459383043,
|
| 248 |
+
"is": 0.011753120779597635,
|
| 249 |
+
"os": 0.2293290940636717
|
| 250 |
+
},
|
| 251 |
+
"SinGLU": {
|
| 252 |
+
"id": 0.24567786852518717,
|
| 253 |
+
"ood": 5.897943178812663,
|
| 254 |
+
"deg": 24.006815160918734,
|
| 255 |
+
"is": 0.013796241102558409,
|
| 256 |
+
"os": 0.8500770525046415
|
| 257 |
+
},
|
| 258 |
+
"v12": {
|
| 259 |
+
"id": 0.10941666861375172,
|
| 260 |
+
"ood": 4.693021059036255,
|
| 261 |
+
"deg": 42.89128081209398,
|
| 262 |
+
"is": 0.02017077570723175,
|
| 263 |
+
"os": 0.9724661958608424
|
| 264 |
+
}
|
| 265 |
+
},
|
| 266 |
+
"phi": {
|
| 267 |
+
"Complex Fn": {
|
| 268 |
+
"m": -0.20280706882476807,
|
| 269 |
+
"s": 0.1989799588918686,
|
| 270 |
+
"mn": -0.8377503156661987,
|
| 271 |
+
"mx": 0.2197563499212265
|
| 272 |
+
},
|
| 273 |
+
"Nested Fn": {
|
| 274 |
+
"m": -0.029095139354467392,
|
| 275 |
+
"s": 0.13748733699321747,
|
| 276 |
+
"mn": -0.49569252133369446,
|
| 277 |
+
"mx": 0.5955169796943665
|
| 278 |
+
},
|
| 279 |
+
"Spiral": {
|
| 280 |
+
"m": 0.015783855691552162,
|
| 281 |
+
"s": 0.21821874380111694,
|
| 282 |
+
"mn": -0.7240511178970337,
|
| 283 |
+
"mx": 0.7039177417755127
|
| 284 |
+
},
|
| 285 |
+
"Checker": {
|
| 286 |
+
"m": -0.0024248864501714706,
|
| 287 |
+
"s": 0.20414583384990692,
|
| 288 |
+
"mn": -0.7574748992919922,
|
| 289 |
+
"mx": 0.690476655960083
|
| 290 |
+
},
|
| 291 |
+
"High-Freq": {
|
| 292 |
+
"m": -0.1338253617286682,
|
| 293 |
+
"s": 0.22809551656246185,
|
| 294 |
+
"mn": -0.7467907071113586,
|
| 295 |
+
"mx": 0.5918763279914856
|
| 296 |
+
},
|
| 297 |
+
"Memorize": {
|
| 298 |
+
"m": 0.013628730550408363,
|
| 299 |
+
"s": 0.17024029791355133,
|
| 300 |
+
"mn": -0.6205101609230042,
|
| 301 |
+
"mx": 0.6283335089683533
|
| 302 |
+
}
|
| 303 |
+
}
|
| 304 |
+
}
|