quantumly commited on
Commit
22f4a87
·
verified ·
1 Parent(s): c4d4865

v0.4 multi-embedding appraiser: 2026-04-26

Browse files
Files changed (1) hide show
  1. v0_4_metadata.json +351 -0
v0_4_metadata.json ADDED
@@ -0,0 +1,351 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "trained_at": "2026-04-26T09:19:05.178649+00:00",
3
+ "data_run_date": "2026-04-25",
4
+ "version": "v0.4-multi-embedding",
5
+ "description": "v0.3 + bge-base-en-v1.5 alongside mpnet-finetuned",
6
+ "parent_version": "v0.3",
7
+ "embedders": [
8
+ "mpnet-finetuned (from v0.3)",
9
+ "BAAI/bge-base-en-v1.5"
10
+ ],
11
+ "splits": {
12
+ "train": {
13
+ "rows": 265240,
14
+ "start": "2022-01-28",
15
+ "end": "2023-09-30"
16
+ },
17
+ "val": {
18
+ "rows": 3545,
19
+ "start": "2023-10-01",
20
+ "end": "2023-12-31"
21
+ },
22
+ "test": {
23
+ "rows": 2744,
24
+ "start": "2024-01-01",
25
+ "end": "2024-05-04"
26
+ }
27
+ },
28
+ "feature_count": 154,
29
+ "feature_cols": [
30
+ "pca_000",
31
+ "pca_001",
32
+ "pca_002",
33
+ "pca_003",
34
+ "pca_004",
35
+ "pca_005",
36
+ "pca_006",
37
+ "pca_007",
38
+ "pca_008",
39
+ "pca_009",
40
+ "pca_010",
41
+ "pca_011",
42
+ "pca_012",
43
+ "pca_013",
44
+ "pca_014",
45
+ "pca_015",
46
+ "pca_016",
47
+ "pca_017",
48
+ "pca_018",
49
+ "pca_019",
50
+ "pca_020",
51
+ "pca_021",
52
+ "pca_022",
53
+ "pca_023",
54
+ "pca_024",
55
+ "pca_025",
56
+ "pca_026",
57
+ "pca_027",
58
+ "pca_028",
59
+ "pca_029",
60
+ "pca_030",
61
+ "pca_031",
62
+ "pca_032",
63
+ "pca_033",
64
+ "pca_034",
65
+ "pca_035",
66
+ "pca_036",
67
+ "pca_037",
68
+ "pca_038",
69
+ "pca_039",
70
+ "pca_040",
71
+ "pca_041",
72
+ "pca_042",
73
+ "pca_043",
74
+ "pca_044",
75
+ "pca_045",
76
+ "pca_046",
77
+ "pca_047",
78
+ "pca_048",
79
+ "pca_049",
80
+ "pca_050",
81
+ "pca_051",
82
+ "pca_052",
83
+ "pca_053",
84
+ "pca_054",
85
+ "pca_055",
86
+ "pca_056",
87
+ "pca_057",
88
+ "pca_058",
89
+ "pca_059",
90
+ "pca_060",
91
+ "pca_061",
92
+ "pca_062",
93
+ "pca_063",
94
+ "len",
95
+ "n_digits",
96
+ "n_letters",
97
+ "n_special",
98
+ "n_lower",
99
+ "n_upper",
100
+ "is_palindrome",
101
+ "is_all_digits",
102
+ "is_all_letters",
103
+ "is_ascii",
104
+ "has_unicode",
105
+ "starts_digit",
106
+ "ends_digit",
107
+ "max_char_run",
108
+ "n_unique_chars",
109
+ "in_wikipedia",
110
+ "in_geonames",
111
+ "in_us_firstname",
112
+ "in_iso3166",
113
+ "in_ticker",
114
+ "in_sec_edgar",
115
+ "in_wiktionary_en",
116
+ "wordlist_hits",
117
+ "club__prepunk_full_rankings",
118
+ "club__social_handles",
119
+ "club__fine_art",
120
+ "club__performing_arts",
121
+ "club__gamertags_double",
122
+ "club__currency_symbols",
123
+ "club__logistics",
124
+ "club__gamertags",
125
+ "club__pokemon_gen1",
126
+ "club__common_english",
127
+ "club__top_crypto_tickers",
128
+ "club__common_animals",
129
+ "club__familynames_usa",
130
+ "club__natural_wonders",
131
+ "club__historic_figures",
132
+ "club__conspiracy_theories",
133
+ "club__mythical_creatures",
134
+ "club__top_nouns",
135
+ "club__finance_terms",
136
+ "club__personas",
137
+ "club__country_codes",
138
+ "club__catholicism",
139
+ "club__crypto_terms",
140
+ "club__sports",
141
+ "club__home",
142
+ "club__top500_cities_global",
143
+ "club__firstnames_usa",
144
+ "club__paranormal",
145
+ "club__wikidata_top_fantasy_char",
146
+ "club__crayola_classic",
147
+ "club__top500_cities_usa",
148
+ "club__top_crypto_names",
149
+ "club__us_government",
150
+ "club__pokemon_gen4",
151
+ "club__pokemon_gen2",
152
+ "club__memes",
153
+ "club__currency_names",
154
+ "club__luxury",
155
+ "club__pokemon_gen3",
156
+ "club__gen_alpha",
157
+ "club__us_states",
158
+ "club__holidays",
159
+ "n_clubs",
160
+ "trademark_conflict",
161
+ "name_age_days",
162
+ "prior_transfer_count",
163
+ "fg_value",
164
+ "eth_tvl_usd",
165
+ "eth_stable_mcap",
166
+ "eth_dex_volume",
167
+ "nft_total_fee_usd",
168
+ "knnmp_count",
169
+ "knnmp_mean_log",
170
+ "knnmp_median_log",
171
+ "knnmp_p90_log",
172
+ "knnmp_max_sim",
173
+ "knnmp_min_sim",
174
+ "knnmp_log_max",
175
+ "knnmp_log_min",
176
+ "knnbg_count",
177
+ "knnbg_mean_log",
178
+ "knnbg_median_log",
179
+ "knnbg_p90_log",
180
+ "knnbg_max_sim",
181
+ "knnbg_min_sim",
182
+ "knnbg_log_max",
183
+ "knnbg_log_min"
184
+ ],
185
+ "pca_dim": 64,
186
+ "best_iteration": 757,
187
+ "xgb_params": {
188
+ "objective": "reg:squarederror",
189
+ "eval_metric": "rmse",
190
+ "tree_method": "hist",
191
+ "device": "cuda",
192
+ "max_depth": 7,
193
+ "learning_rate": 0.04,
194
+ "subsample": 0.85,
195
+ "colsample_bytree": 0.65,
196
+ "min_child_weight": 8,
197
+ "reg_alpha": 0.5,
198
+ "reg_lambda": 2.0,
199
+ "seed": 42
200
+ },
201
+ "metrics": {
202
+ "train": {
203
+ "r2_log": 0.830089807510376,
204
+ "rmse_log": 0.6656108498573303,
205
+ "mae_log": 0.44242581725120544,
206
+ "median_ape": 0.2753954231739044,
207
+ "bias_log": -0.00027043218142353
208
+ },
209
+ "val": {
210
+ "r2_log": 0.6791324615478516,
211
+ "rmse_log": 1.0376933813095093,
212
+ "mae_log": 0.7315031886100769,
213
+ "median_ape": 0.5113449096679688,
214
+ "bias_log": 0.10410919040441513
215
+ },
216
+ "test": {
217
+ "r2_log": 0.31753313541412354,
218
+ "rmse_log": 1.5363107919692993,
219
+ "mae_log": 1.2583258152008057,
220
+ "median_ape": 1.2974694967269897,
221
+ "bias_log": 0.7355491518974304
222
+ }
223
+ },
224
+ "top_features": [
225
+ {
226
+ "name": "knnmp_mean_log",
227
+ "gain": 1521.73291015625
228
+ },
229
+ {
230
+ "name": "knnmp_median_log",
231
+ "gain": 1161.393798828125
232
+ },
233
+ {
234
+ "name": "is_all_digits",
235
+ "gain": 576.3497314453125
236
+ },
237
+ {
238
+ "name": "knnmp_p90_log",
239
+ "gain": 492.0875244140625
240
+ },
241
+ {
242
+ "name": "len",
243
+ "gain": 380.18292236328125
244
+ },
245
+ {
246
+ "name": "in_wikipedia",
247
+ "gain": 365.04254150390625
248
+ },
249
+ {
250
+ "name": "n_clubs",
251
+ "gain": 194.0254669189453
252
+ },
253
+ {
254
+ "name": "knnmp_count",
255
+ "gain": 154.111572265625
256
+ },
257
+ {
258
+ "name": "ends_digit",
259
+ "gain": 137.3547821044922
260
+ },
261
+ {
262
+ "name": "trademark_conflict",
263
+ "gain": 134.1524658203125
264
+ },
265
+ {
266
+ "name": "n_letters",
267
+ "gain": 129.6276092529297
268
+ },
269
+ {
270
+ "name": "name_age_days",
271
+ "gain": 128.65174865722656
272
+ },
273
+ {
274
+ "name": "club__gamertags",
275
+ "gain": 128.5192108154297
276
+ },
277
+ {
278
+ "name": "is_palindrome",
279
+ "gain": 125.1925048828125
280
+ },
281
+ {
282
+ "name": "n_digits",
283
+ "gain": 125.02507019042969
284
+ },
285
+ {
286
+ "name": "club__social_handles",
287
+ "gain": 119.58012390136719
288
+ },
289
+ {
290
+ "name": "eth_stable_mcap",
291
+ "gain": 114.49725341796875
292
+ },
293
+ {
294
+ "name": "has_unicode",
295
+ "gain": 105.6874771118164
296
+ },
297
+ {
298
+ "name": "n_lower",
299
+ "gain": 102.0856704711914
300
+ },
301
+ {
302
+ "name": "n_special",
303
+ "gain": 89.66917419433594
304
+ },
305
+ {
306
+ "name": "eth_tvl_usd",
307
+ "gain": 89.55354309082031
308
+ },
309
+ {
310
+ "name": "club__prepunk_full_rankings",
311
+ "gain": 87.96327209472656
312
+ },
313
+ {
314
+ "name": "n_unique_chars",
315
+ "gain": 86.47875213623047
316
+ },
317
+ {
318
+ "name": "pca_045",
319
+ "gain": 85.28775787353516
320
+ },
321
+ {
322
+ "name": "is_ascii",
323
+ "gain": 83.36601257324219
324
+ },
325
+ {
326
+ "name": "starts_digit",
327
+ "gain": 73.62259674072266
328
+ },
329
+ {
330
+ "name": "knnbg_count",
331
+ "gain": 61.69795608520508
332
+ },
333
+ {
334
+ "name": "pca_001",
335
+ "gain": 60.780391693115234
336
+ },
337
+ {
338
+ "name": "fg_value",
339
+ "gain": 57.86528778076172
340
+ },
341
+ {
342
+ "name": "pca_026",
343
+ "gain": 57.13859939575195
344
+ }
345
+ ],
346
+ "knn_gain_split": {
347
+ "mpnet_ft": 3443.8879318237305,
348
+ "bge": 257.9545383453369
349
+ },
350
+ "wandb_run": "https://wandb.ai/quantumly-aletheia-research/ens-appraiser/runs/osbosu7j"
351
+ }