Upload folder using huggingface_hub
Browse files
qwen3_06B.blocks.12.hook_resid_post.sae.sparsity100.mse0.001.kl0.01.physics.exp4/history.json
CHANGED
|
@@ -1,83 +1,326 @@
|
|
| 1 |
{
|
| 2 |
"loss": [
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
],
|
| 5 |
"recon_loss": [
|
| 6 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
],
|
| 8 |
"l1_loss": [
|
| 9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
],
|
| 11 |
"sparsity": [
|
| 12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
],
|
| 14 |
"recon_contribution": [
|
| 15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
],
|
| 17 |
"l1_contribution": [
|
| 18 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
],
|
| 20 |
"dead_features": [
|
| 21 |
-
0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
],
|
| 23 |
"dead_feature_percentage": [
|
| 24 |
-
0.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
],
|
| 26 |
"total_post_layer_mse": [
|
| 27 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
],
|
| 29 |
"block_mse_contribution": [
|
| 30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
],
|
| 32 |
"13_mse": [
|
| 33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
],
|
| 35 |
"14_mse": [
|
| 36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
],
|
| 38 |
"15_mse": [
|
| 39 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
],
|
| 41 |
"16_mse": [
|
| 42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
],
|
| 44 |
"17_mse": [
|
| 45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
],
|
| 47 |
"18_mse": [
|
| 48 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
],
|
| 50 |
"19_mse": [
|
| 51 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
],
|
| 53 |
"20_mse": [
|
| 54 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
],
|
| 56 |
"21_mse": [
|
| 57 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
],
|
| 59 |
"22_mse": [
|
| 60 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
],
|
| 62 |
"23_mse": [
|
| 63 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
],
|
| 65 |
"24_mse": [
|
| 66 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
],
|
| 68 |
"25_mse": [
|
| 69 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
],
|
| 71 |
"26_mse": [
|
| 72 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 73 |
],
|
| 74 |
"27_mse": [
|
| 75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 76 |
],
|
| 77 |
"logit_kl": [
|
| 78 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
],
|
| 80 |
"kl_contribution": [
|
| 81 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 82 |
]
|
| 83 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"loss": [
|
| 3 |
+
31.34565018699791,
|
| 4 |
+
5.406206299182682,
|
| 5 |
+
5.307205936809937,
|
| 6 |
+
5.217020622426425,
|
| 7 |
+
5.146753770907275,
|
| 8 |
+
5.106208472609653,
|
| 9 |
+
5.038848593947868,
|
| 10 |
+
4.967890664951798,
|
| 11 |
+
4.911847650137086,
|
| 12 |
+
4.935633503709902
|
| 13 |
],
|
| 14 |
"recon_loss": [
|
| 15 |
+
7.363610733262902,
|
| 16 |
+
1.0907883467893185,
|
| 17 |
+
1.019061345269138,
|
| 18 |
+
0.9891309939066213,
|
| 19 |
+
0.9592718267761179,
|
| 20 |
+
0.9471277262168663,
|
| 21 |
+
0.9267459577835955,
|
| 22 |
+
0.9187925020764683,
|
| 23 |
+
0.9031714983242757,
|
| 24 |
+
0.8953778161714219
|
| 25 |
],
|
| 26 |
"l1_loss": [
|
| 27 |
+
0.22866084941841186,
|
| 28 |
+
0.0390433030638923,
|
| 29 |
+
0.03939214351087442,
|
| 30 |
+
0.038972768060704495,
|
| 31 |
+
0.03880955467015205,
|
| 32 |
+
0.03855483874695299,
|
| 33 |
+
0.03821273812604912,
|
| 34 |
+
0.037600100013320016,
|
| 35 |
+
0.03725809549072221,
|
| 36 |
+
0.037608155624029485
|
| 37 |
],
|
| 38 |
"sparsity": [
|
| 39 |
+
61.31447249110308,
|
| 40 |
+
18.32951434328735,
|
| 41 |
+
15.62452454155797,
|
| 42 |
+
13.748240198560417,
|
| 43 |
+
12.145425100209064,
|
| 44 |
+
10.875871384931438,
|
| 45 |
+
9.955307994150502,
|
| 46 |
+
9.252135027963398,
|
| 47 |
+
8.656200779512242,
|
| 48 |
+
8.173118343011529
|
| 49 |
],
|
| 50 |
"recon_contribution": [
|
| 51 |
+
7.363610733262902,
|
| 52 |
+
1.0907883467893185,
|
| 53 |
+
1.019061345269138,
|
| 54 |
+
0.9891309939066213,
|
| 55 |
+
0.9592718267761179,
|
| 56 |
+
0.9471277262168663,
|
| 57 |
+
0.9267459577835955,
|
| 58 |
+
0.9187925020764683,
|
| 59 |
+
0.9031714983242757,
|
| 60 |
+
0.8953778161714219
|
| 61 |
],
|
| 62 |
"l1_contribution": [
|
| 63 |
+
22.866084938487173,
|
| 64 |
+
3.9043303037944592,
|
| 65 |
+
3.9392143545332305,
|
| 66 |
+
3.897276803333842,
|
| 67 |
+
3.8809554718376407,
|
| 68 |
+
3.855483877431906,
|
| 69 |
+
3.821273808783673,
|
| 70 |
+
3.7600100029756347,
|
| 71 |
+
3.7258095508994074,
|
| 72 |
+
3.760815557680632
|
| 73 |
],
|
| 74 |
"dead_features": [
|
| 75 |
+
0,
|
| 76 |
+
172,
|
| 77 |
+
163,
|
| 78 |
+
91,
|
| 79 |
+
66,
|
| 80 |
+
57,
|
| 81 |
+
53,
|
| 82 |
+
82,
|
| 83 |
+
91,
|
| 84 |
+
128
|
| 85 |
],
|
| 86 |
"dead_feature_percentage": [
|
| 87 |
+
0.0,
|
| 88 |
+
4.19921875,
|
| 89 |
+
3.9794921875,
|
| 90 |
+
2.2216796875,
|
| 91 |
+
1.611328125,
|
| 92 |
+
1.3916015625,
|
| 93 |
+
1.2939453125,
|
| 94 |
+
2.001953125,
|
| 95 |
+
2.2216796875,
|
| 96 |
+
3.125
|
| 97 |
],
|
| 98 |
"total_post_layer_mse": [
|
| 99 |
+
148.70505004105016,
|
| 100 |
+
52.953618915590965,
|
| 101 |
+
47.59579880320345,
|
| 102 |
+
45.40587117121642,
|
| 103 |
+
43.90041130157094,
|
| 104 |
+
43.04117819446599,
|
| 105 |
+
42.58426681207649,
|
| 106 |
+
42.081879781523654,
|
| 107 |
+
41.968728798167305,
|
| 108 |
+
41.544887778273086
|
| 109 |
],
|
| 110 |
"block_mse_contribution": [
|
| 111 |
+
0.14870505717561555,
|
| 112 |
+
0.05295362142570398,
|
| 113 |
+
0.04759580106815036,
|
| 114 |
+
0.04540587332978853,
|
| 115 |
+
0.04390041339679825,
|
| 116 |
+
0.04304118026969784,
|
| 117 |
+
0.042584268839643696,
|
| 118 |
+
0.04208188177977811,
|
| 119 |
+
0.041968730814175625,
|
| 120 |
+
0.04154488974576959
|
| 121 |
],
|
| 122 |
"13_mse": [
|
| 123 |
+
2.2772952729016374,
|
| 124 |
+
0.3212174427969859,
|
| 125 |
+
0.300773655246914,
|
| 126 |
+
0.29224107453697606,
|
| 127 |
+
0.2873397515909402,
|
| 128 |
+
0.28486264569628544,
|
| 129 |
+
0.28206844365877104,
|
| 130 |
+
0.280317462947374,
|
| 131 |
+
0.2795116462038719,
|
| 132 |
+
0.2780818284592463
|
| 133 |
],
|
| 134 |
"14_mse": [
|
| 135 |
+
2.2520010681428673,
|
| 136 |
+
0.34081664202528333,
|
| 137 |
+
0.3177218272164047,
|
| 138 |
+
0.30777032830213247,
|
| 139 |
+
0.3016610434748157,
|
| 140 |
+
0.2982806815011771,
|
| 141 |
+
0.2951081657463065,
|
| 142 |
+
0.29278522958480496,
|
| 143 |
+
0.29149628146097883,
|
| 144 |
+
0.2897012154973234
|
| 145 |
],
|
| 146 |
"15_mse": [
|
| 147 |
+
2.3888719536756216,
|
| 148 |
+
0.4142598775773406,
|
| 149 |
+
0.38422218377285816,
|
| 150 |
+
0.3708618829592078,
|
| 151 |
+
0.3633045323051237,
|
| 152 |
+
0.35886498945290674,
|
| 153 |
+
0.3555400279775045,
|
| 154 |
+
0.35256703012734486,
|
| 155 |
+
0.35170833960025333,
|
| 156 |
+
0.3496244555424736
|
| 157 |
],
|
| 158 |
"16_mse": [
|
| 159 |
+
2.7679619446713875,
|
| 160 |
+
0.6436231954634389,
|
| 161 |
+
0.5921369913510753,
|
| 162 |
+
0.5689708225404028,
|
| 163 |
+
0.5550883400346771,
|
| 164 |
+
0.5466277638381433,
|
| 165 |
+
0.5416478285730753,
|
| 166 |
+
0.5363458237649346,
|
| 167 |
+
0.534718270164321,
|
| 168 |
+
0.5304997910944639
|
| 169 |
],
|
| 170 |
"17_mse": [
|
| 171 |
+
3.1861002010701083,
|
| 172 |
+
0.8477333854187242,
|
| 173 |
+
0.7744848554638972,
|
| 174 |
+
0.7425822650999131,
|
| 175 |
+
0.7233719172963636,
|
| 176 |
+
0.7116526685329728,
|
| 177 |
+
0.7054496887082615,
|
| 178 |
+
0.6982922257995285,
|
| 179 |
+
0.6963646468296564,
|
| 180 |
+
0.6902589849034723
|
| 181 |
],
|
| 182 |
"18_mse": [
|
| 183 |
+
3.7555036417313947,
|
| 184 |
+
1.1326086423031294,
|
| 185 |
+
1.0296482221209855,
|
| 186 |
+
0.984571643632577,
|
| 187 |
+
0.9569441505936198,
|
| 188 |
+
0.9402139862667673,
|
| 189 |
+
0.931800523403644,
|
| 190 |
+
0.9222762093431872,
|
| 191 |
+
0.9197688159790445,
|
| 192 |
+
0.9107675966196604
|
| 193 |
],
|
| 194 |
"19_mse": [
|
| 195 |
+
4.658164522949552,
|
| 196 |
+
1.6065882342927,
|
| 197 |
+
1.4536406816987681,
|
| 198 |
+
1.3873231445688416,
|
| 199 |
+
1.3469748173956106,
|
| 200 |
+
1.3220691324481773,
|
| 201 |
+
1.3102280548278322,
|
| 202 |
+
1.2960434516626937,
|
| 203 |
+
1.2931566882507264,
|
| 204 |
+
1.2800830798880651
|
| 205 |
],
|
| 206 |
"20_mse": [
|
| 207 |
+
6.004892664019765,
|
| 208 |
+
2.2889097544540884,
|
| 209 |
+
2.064345257146094,
|
| 210 |
+
1.9672186479456348,
|
| 211 |
+
1.9083819135706475,
|
| 212 |
+
1.8707357317309363,
|
| 213 |
+
1.8544755548672607,
|
| 214 |
+
1.832564960869002,
|
| 215 |
+
1.8290262126175045,
|
| 216 |
+
1.8100585000378822
|
| 217 |
],
|
| 218 |
"21_mse": [
|
| 219 |
+
7.594265379804246,
|
| 220 |
+
3.046438876485237,
|
| 221 |
+
2.7379585747489203,
|
| 222 |
+
2.6038963830751376,
|
| 223 |
+
2.521404554301921,
|
| 224 |
+
2.4692950052483487,
|
| 225 |
+
2.4479584069161238,
|
| 226 |
+
2.4181223828741847,
|
| 227 |
+
2.413917049582034,
|
| 228 |
+
2.3880339142723552
|
| 229 |
],
|
| 230 |
"22_mse": [
|
| 231 |
+
9.418807795707215,
|
| 232 |
+
3.8846809654460106,
|
| 233 |
+
3.4818888906668977,
|
| 234 |
+
3.310109221921923,
|
| 235 |
+
3.2005052570533112,
|
| 236 |
+
3.1332413839199242,
|
| 237 |
+
3.1046694344929593,
|
| 238 |
+
3.0659138459507855,
|
| 239 |
+
3.060757096110106,
|
| 240 |
+
3.028429642636725
|
| 241 |
],
|
| 242 |
"23_mse": [
|
| 243 |
+
11.834317743444496,
|
| 244 |
+
4.997612617050547,
|
| 245 |
+
4.4771716028685535,
|
| 246 |
+
4.257016977697845,
|
| 247 |
+
4.114696562757417,
|
| 248 |
+
4.027720311454269,
|
| 249 |
+
3.991431857143244,
|
| 250 |
+
3.940902676187197,
|
| 251 |
+
3.9357787141073723,
|
| 252 |
+
3.8935535536100723
|
| 253 |
],
|
| 254 |
"24_mse": [
|
| 255 |
+
14.68451576083398,
|
| 256 |
+
6.1915769609064695,
|
| 257 |
+
5.545960285094925,
|
| 258 |
+
5.2792783787376,
|
| 259 |
+
5.099575045928442,
|
| 260 |
+
4.992917687468406,
|
| 261 |
+
4.945194986072126,
|
| 262 |
+
4.882690692935785,
|
| 263 |
+
4.874595674128175,
|
| 264 |
+
4.822797090449232
|
| 265 |
],
|
| 266 |
"25_mse": [
|
| 267 |
+
18.612011243621552,
|
| 268 |
+
7.7885244760374475,
|
| 269 |
+
6.980879172499209,
|
| 270 |
+
6.645965072102135,
|
| 271 |
+
6.420509336776989,
|
| 272 |
+
6.283283990548146,
|
| 273 |
+
6.221309304904831,
|
| 274 |
+
6.13913771446715,
|
| 275 |
+
6.127566409671694,
|
| 276 |
+
6.061285808932741
|
| 277 |
],
|
| 278 |
"26_mse": [
|
| 279 |
+
20.381727528865767,
|
| 280 |
+
8.219243945626904,
|
| 281 |
+
7.374626742765591,
|
| 282 |
+
7.031916140442066,
|
| 283 |
+
6.795138808407415,
|
| 284 |
+
6.653138730186898,
|
| 285 |
+
6.587366769722301,
|
| 286 |
+
6.502645790510188,
|
| 287 |
+
6.48968698116059,
|
| 288 |
+
6.422333440481615
|
| 289 |
],
|
| 290 |
"27_mse": [
|
| 291 |
+
38.88861331961056,
|
| 292 |
+
11.229783899706662,
|
| 293 |
+
10.080339860542358,
|
| 294 |
+
9.656149187654023,
|
| 295 |
+
9.305515270083642,
|
| 296 |
+
9.148273486172632,
|
| 297 |
+
9.010017765062253,
|
| 298 |
+
8.92127428449949,
|
| 299 |
+
8.870675972300974,
|
| 300 |
+
8.78937887584776
|
| 301 |
],
|
| 302 |
"logit_kl": [
|
| 303 |
+
96.72496022462578,
|
| 304 |
+
35.81340313491736,
|
| 305 |
+
30.1334463538141,
|
| 306 |
+
28.520700067582084,
|
| 307 |
+
26.26260372665935,
|
| 308 |
+
26.055573190046186,
|
| 309 |
+
24.824461100216276,
|
| 310 |
+
24.70062872170067,
|
| 311 |
+
24.089785100631456,
|
| 312 |
+
23.78952288387337
|
| 313 |
],
|
| 314 |
"kl_contribution": [
|
| 315 |
+
0.9672495801698454,
|
| 316 |
+
0.3581340230275509,
|
| 317 |
+
0.30133445653125795,
|
| 318 |
+
0.28520699394376653,
|
| 319 |
+
0.2626260318120677,
|
| 320 |
+
0.2605557263581643,
|
| 321 |
+
0.2482446057665655,
|
| 322 |
+
0.24700628154288481,
|
| 323 |
+
0.24089784570630693,
|
| 324 |
+
0.23789522302736427
|
| 325 |
]
|
| 326 |
}
|
qwen3_06B.blocks.12.hook_resid_post.sae.sparsity100.mse0.001.kl0.01.physics.exp4/weights.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 33577309
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3efa500ab6d183493196c395823b56fd8b4a5d5754bd65b21ee281906be91db2
|
| 3 |
size 33577309
|