Polifemus commited on
Commit
76ddc8e
·
verified ·
1 Parent(s): 88ae4b9

ablation sc80-v5-rmix14: config.json

Browse files
scaling/ablation/sc80-v5-rmix14/config.json ADDED
@@ -0,0 +1,360 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "EventTransformerPretrained",
3
+ "task": "causal next-activity + next-resource pretraining",
4
+ "corpus": "corpus v5 mix rmix14: 18 real logs repeated x14 (~9.5% of windows) + 2,500 synthetic shards; 318 seen logs + UNK",
5
+ "n_act": 777,
6
+ "n_res": 2740,
7
+ "n_logs": 321,
8
+ "n_classes": 777,
9
+ "d_model": 768,
10
+ "n_layers": 12,
11
+ "n_heads": 8,
12
+ "dropout": 0.1,
13
+ "max_len": 256,
14
+ "act_dim": 64,
15
+ "res_dim": 32,
16
+ "log_dim": 16,
17
+ "attention": "causal",
18
+ "params_millions": 88.19,
19
+ "seen_logs": [
20
+ "bpi2012",
21
+ "bpi2017",
22
+ "bpi2013_incidents",
23
+ "bpi2013_closed",
24
+ "bpi2013_open",
25
+ "bpi2020_intl",
26
+ "bpi2020_dom",
27
+ "bpi2020_prepaid",
28
+ "bpi2020_rfp",
29
+ "sepsis",
30
+ "helpdesk",
31
+ "bpi2019",
32
+ "bpi2018",
33
+ "bpi2014",
34
+ "bpi2015_2",
35
+ "bpi2015_3",
36
+ "bpi2015_4",
37
+ "bpi2015_5",
38
+ "gmx2_01",
39
+ "gmx2_02",
40
+ "gmx2_03",
41
+ "gmx2_04",
42
+ "gmx2_05",
43
+ "gmx2_06",
44
+ "gmx2_07",
45
+ "gmx2_08",
46
+ "gmx2_09",
47
+ "gmx2_10",
48
+ "gmx2_11",
49
+ "gmx2_12",
50
+ "gmx2_13",
51
+ "gmx2_14",
52
+ "gmx2_15",
53
+ "gmx2_16",
54
+ "gmx2_17",
55
+ "gmx2_18",
56
+ "gmx2_19",
57
+ "gmx2_20",
58
+ "gmx2_21",
59
+ "gmx2_22",
60
+ "gmx2_23",
61
+ "gmx2_24",
62
+ "gmx2_25",
63
+ "gmx2_26",
64
+ "gmx2_27",
65
+ "gmx2_28",
66
+ "gmx2_29",
67
+ "gmx2_30",
68
+ "gmx2_14_01",
69
+ "gmx2_14_02",
70
+ "gmx2_14_03",
71
+ "gmx2_14_04",
72
+ "gmx2_14_05",
73
+ "gmx2_14_06",
74
+ "gmx2_14_07",
75
+ "gmx2_14_08",
76
+ "gmx2_14_09",
77
+ "gmx2_14_10",
78
+ "gmx3_001",
79
+ "gmx3_002",
80
+ "gmx3_003",
81
+ "gmx3_004",
82
+ "gmx3_005",
83
+ "gmx3_006",
84
+ "gmx3_007",
85
+ "gmx3_008",
86
+ "gmx3_009",
87
+ "gmx3_010",
88
+ "gmx3_011",
89
+ "gmx3_012",
90
+ "gmx3_013",
91
+ "gmx3_014",
92
+ "gmx3_015",
93
+ "gmx3_016",
94
+ "gmx3_017",
95
+ "gmx3_018",
96
+ "gmx3_019",
97
+ "gmx3_020",
98
+ "gmx3_021",
99
+ "gmx3_022",
100
+ "gmx3_023",
101
+ "gmx3_024",
102
+ "gmx3_025",
103
+ "gmx3_026",
104
+ "gmx3_027",
105
+ "gmx3_028",
106
+ "gmx3_029",
107
+ "gmx3_030",
108
+ "gmx3_031",
109
+ "gmx3_032",
110
+ "gmx3_033",
111
+ "gmx3_034",
112
+ "gmx3_035",
113
+ "gmx3_036",
114
+ "gmx3_037",
115
+ "gmx3_038",
116
+ "gmx3_039",
117
+ "gmx3_040",
118
+ "gmx3_041",
119
+ "gmx3_042",
120
+ "gmx3_043",
121
+ "gmx3_044",
122
+ "gmx3_045",
123
+ "gmx3_046",
124
+ "gmx3_047",
125
+ "gmx3_048",
126
+ "gmx3_049",
127
+ "gmx3_050",
128
+ "gmx3_051",
129
+ "gmx3_052",
130
+ "gmx3_053",
131
+ "gmx3_054",
132
+ "gmx3_055",
133
+ "gmx3_056",
134
+ "gmx3_057",
135
+ "gmx3_058",
136
+ "gmx3_059",
137
+ "gmx3_060",
138
+ "gmx3_061",
139
+ "gmx3_062",
140
+ "gmx3_063",
141
+ "gmx3_064",
142
+ "gmx3_065",
143
+ "gmx3_066",
144
+ "gmx3_067",
145
+ "gmx3_068",
146
+ "gmx3_069",
147
+ "gmx3_070",
148
+ "gmx3_071",
149
+ "gmx3_072",
150
+ "gmx3_073",
151
+ "gmx3_074",
152
+ "gmx3_075",
153
+ "gmx3_076",
154
+ "gmx3_077",
155
+ "gmx3_078",
156
+ "gmx3_079",
157
+ "gmx3_080",
158
+ "gmx3_081",
159
+ "gmx3_082",
160
+ "gmx3_083",
161
+ "gmx3_084",
162
+ "gmx3_085",
163
+ "gmx3_086",
164
+ "gmx3_087",
165
+ "gmx3_088",
166
+ "gmx3_089",
167
+ "gmx3_090",
168
+ "gmx3_091",
169
+ "gmx3_092",
170
+ "gmx3_093",
171
+ "gmx3_094",
172
+ "gmx3_095",
173
+ "gmx3_096",
174
+ "gmx3_097",
175
+ "gmx3_098",
176
+ "gmx3_099",
177
+ "gmx3_100",
178
+ "gmx3_101",
179
+ "gmx3_102",
180
+ "gmx3_103",
181
+ "gmx3_104",
182
+ "gmx3_105",
183
+ "gmx3_106",
184
+ "gmx3_107",
185
+ "gmx3_108",
186
+ "gmx3_109",
187
+ "gmx3_110",
188
+ "gmx3_111",
189
+ "gmx3_112",
190
+ "gmx3_113",
191
+ "gmx3_114",
192
+ "gmx3_115",
193
+ "gmx3_116",
194
+ "gmx3_117",
195
+ "gmx3_118",
196
+ "gmx3_119",
197
+ "gmx3_120",
198
+ "gmx3_121",
199
+ "gmx3_122",
200
+ "gmx3_123",
201
+ "gmx3_124",
202
+ "gmx3_125",
203
+ "gmx3_126",
204
+ "gmx3_127",
205
+ "gmx3_128",
206
+ "gmx3_129",
207
+ "gmx3_130",
208
+ "gmx3_131",
209
+ "gmx3_132",
210
+ "gmx3_133",
211
+ "gmx3_134",
212
+ "gmx3_135",
213
+ "gmx3_136",
214
+ "gmx3_137",
215
+ "gmx3_138",
216
+ "gmx3_139",
217
+ "gmx3_140",
218
+ "gmx3_141",
219
+ "gmx3_142",
220
+ "gmx3_143",
221
+ "gmx3_144",
222
+ "gmx3_145",
223
+ "gmx3_146",
224
+ "gmx3_147",
225
+ "gmx3_148",
226
+ "gmx3_149",
227
+ "gmx3_150",
228
+ "gmx3_151",
229
+ "gmx3_152",
230
+ "gmx3_153",
231
+ "gmx3_154",
232
+ "gmx3_155",
233
+ "gmx3_156",
234
+ "gmx3_157",
235
+ "gmx3_158",
236
+ "gmx3_159",
237
+ "gmx3_160",
238
+ "gmx3_161",
239
+ "gmx3_162",
240
+ "gmx3_163",
241
+ "gmx3_164",
242
+ "gmx3_165",
243
+ "gmx3_166",
244
+ "gmx3_167",
245
+ "gmx3_168",
246
+ "gmx3_169",
247
+ "gmx3_170",
248
+ "gmx3_171",
249
+ "gmx3_172",
250
+ "gmx3_173",
251
+ "gmx3_174",
252
+ "gmx3_175",
253
+ "gmx3_176",
254
+ "gmx3_177",
255
+ "gmx3_178",
256
+ "gmx3_179",
257
+ "gmx3_180",
258
+ "gmx3_181",
259
+ "gmx3_182",
260
+ "gmx3_183",
261
+ "gmx3_184",
262
+ "gmx3_185",
263
+ "gmx3_186",
264
+ "gmx3_187",
265
+ "gmx3_188",
266
+ "gmx3_189",
267
+ "gmx3_190",
268
+ "gmx3_191",
269
+ "gmx3_192",
270
+ "gmx3_193",
271
+ "gmx3_194",
272
+ "gmx3_195",
273
+ "gmx3_196",
274
+ "gmx3_197",
275
+ "gmx3_198",
276
+ "gmx3_199",
277
+ "gmx3_200",
278
+ "gmx3_201",
279
+ "gmx3_202",
280
+ "gmx3_203",
281
+ "gmx3_204",
282
+ "gmx3_205",
283
+ "gmx3_206",
284
+ "gmx3_207",
285
+ "gmx3_208",
286
+ "gmx3_209",
287
+ "gmx3_210",
288
+ "gmx3_211",
289
+ "gmx3_212",
290
+ "gmx3_213",
291
+ "gmx3_214",
292
+ "gmx3_215",
293
+ "gmx3_216",
294
+ "gmx3_217",
295
+ "gmx3_218",
296
+ "gmx3_219",
297
+ "gmx3_220",
298
+ "gmx3_221",
299
+ "gmx3_222",
300
+ "gmx3_223",
301
+ "gmx3_224",
302
+ "gmx3_225",
303
+ "gmx3_226",
304
+ "gmx3_227",
305
+ "gmx3_228",
306
+ "gmx3_229",
307
+ "gmx3_230",
308
+ "gmx3_231",
309
+ "gmx3_232",
310
+ "gmx3_233",
311
+ "gmx3_234",
312
+ "gmx3_235",
313
+ "gmx3_236",
314
+ "gmx3_237",
315
+ "gmx3_238",
316
+ "gmx3_239",
317
+ "gmx3_240",
318
+ "gmx3_241",
319
+ "gmx3_242",
320
+ "gmx3_243",
321
+ "gmx3_244",
322
+ "gmx3_245",
323
+ "gmx3_246",
324
+ "gmx3_247",
325
+ "gmx3_248",
326
+ "gmx3_249",
327
+ "gmx3_250",
328
+ "gmx3_251",
329
+ "gmx3_252",
330
+ "gmx3_253",
331
+ "gmx3_254",
332
+ "gmx3_255",
333
+ "gmx3_256",
334
+ "gmx3_257",
335
+ "gmx3_258",
336
+ "gmx3_259",
337
+ "gmx3_260"
338
+ ],
339
+ "best": {
340
+ "val_loss": 1.6699742188320161,
341
+ "val_act_acc": 0.6158031940561585,
342
+ "epoch": 1
343
+ },
344
+ "id_dropout": 0.15,
345
+ "log_dropout": 0.15,
346
+ "log_unk": true,
347
+ "metrics": {
348
+ "val_act_acc": 0.6158,
349
+ "zero_shot_acc": 0.3969,
350
+ "zero_shot_top3": 0.4759,
351
+ "zero_shot_macro_f1": 0.2215,
352
+ "ft_test_acc": 0.4695,
353
+ "ft_top3": 0.6456,
354
+ "ft_macro_f1": 0.2096,
355
+ "ft_best_epoch": 2
356
+ },
357
+ "ablation": true,
358
+ "mix": "rmix14 (real shards repeated x14)",
359
+ "positions_seen": 406511104
360
+ }