abdiharyadi commited on
Commit
a4db51e
1 Parent(s): ba87555

Training in progress, epoch 13

Browse files
last-checkpoint/model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5f63231d876ab3ad240434253a3c1d9b5e552e066e216bb8da212e99bc37119e
3
  size 1575259780
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b35a3bc2ac180ed070b42029c9a9dd327a1a9559e81df276f329c07eb21d04fc
3
  size 1575259780
last-checkpoint/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d1a909c257893974ba05f3eb37565085bdb2d8c625613a876b31cc2da2b2012c
3
  size 3150397656
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba79276529acf6b7fcac21ab5be5fc5756c900e697415161ce71f7759f9fa8e0
3
  size 3150397656
last-checkpoint/rng_state.pth CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7c59bbf4f77ee1d62d669bd92ebc1ac7ef8973213ecedf3290bd0e2c91e305da
3
  size 14244
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e52ca2f5b1048c2984d9cb01ff8bc5c06ec7e6e1ac850eb54ef8fe7147dcf65
3
  size 14244
last-checkpoint/scheduler.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c8ef238b253ce03385bc828f9053d9ef361aa0eb279d15d26d1e1d31cc74c139
3
  size 1064
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df05e968bcb6cb9f8c607bdaf90fbac1131121f0efd29f2e5e7bc42c79c2d577
3
  size 1064
last-checkpoint/trainer_state.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
- "best_metric": 12.8144,
3
- "best_model_checkpoint": "/kaggle/tmp/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted/checkpoint-50298",
4
- "epoch": 13.0,
5
  "eval_steps": 500,
6
- "global_step": 50298,
7
  "is_hyper_param_search": false,
8
  "is_local_process_zero": true,
9
  "is_world_process_zero": true,
@@ -12891,2342 +12891,6 @@
12891
  "eval_samples_per_second": 2.165,
12892
  "eval_steps_per_second": 1.083,
12893
  "step": 42562
12894
- },
12895
- {
12896
- "epoch": 11.0054270577594,
12897
- "learning_rate": 1.3131725657980034e-07,
12898
- "loss": 2.4441,
12899
- "step": 42580
12900
- },
12901
- {
12902
- "epoch": 11.010595684196925,
12903
- "learning_rate": 1.31284843770258e-07,
12904
- "loss": 2.5027,
12905
- "step": 42600
12906
- },
12907
- {
12908
- "epoch": 11.015764310634449,
12909
- "learning_rate": 1.3125243096071566e-07,
12910
- "loss": 2.4709,
12911
- "step": 42620
12912
- },
12913
- {
12914
- "epoch": 11.020932937071974,
12915
- "learning_rate": 1.3122001815117335e-07,
12916
- "loss": 2.4576,
12917
- "step": 42640
12918
- },
12919
- {
12920
- "epoch": 11.026101563509497,
12921
- "learning_rate": 1.31187605341631e-07,
12922
- "loss": 2.4395,
12923
- "step": 42660
12924
- },
12925
- {
12926
- "epoch": 11.031270189947021,
12927
- "learning_rate": 1.3115519253208867e-07,
12928
- "loss": 2.504,
12929
- "step": 42680
12930
- },
12931
- {
12932
- "epoch": 11.036438816384546,
12933
- "learning_rate": 1.3112277972254634e-07,
12934
- "loss": 2.5521,
12935
- "step": 42700
12936
- },
12937
- {
12938
- "epoch": 11.04160744282207,
12939
- "learning_rate": 1.31090366913004e-07,
12940
- "loss": 2.5074,
12941
- "step": 42720
12942
- },
12943
- {
12944
- "epoch": 11.046776069259595,
12945
- "learning_rate": 1.3105795410346169e-07,
12946
- "loss": 2.5079,
12947
- "step": 42740
12948
- },
12949
- {
12950
- "epoch": 11.051944695697118,
12951
- "learning_rate": 1.3102554129391935e-07,
12952
- "loss": 2.5337,
12953
- "step": 42760
12954
- },
12955
- {
12956
- "epoch": 11.057113322134644,
12957
- "learning_rate": 1.30993128484377e-07,
12958
- "loss": 2.4815,
12959
- "step": 42780
12960
- },
12961
- {
12962
- "epoch": 11.062281948572167,
12963
- "learning_rate": 1.309607156748347e-07,
12964
- "loss": 2.4897,
12965
- "step": 42800
12966
- },
12967
- {
12968
- "epoch": 11.06745057500969,
12969
- "learning_rate": 1.3092830286529236e-07,
12970
- "loss": 2.5184,
12971
- "step": 42820
12972
- },
12973
- {
12974
- "epoch": 11.072619201447216,
12975
- "learning_rate": 1.3089589005575002e-07,
12976
- "loss": 2.459,
12977
- "step": 42840
12978
- },
12979
- {
12980
- "epoch": 11.07778782788474,
12981
- "learning_rate": 1.308634772462077e-07,
12982
- "loss": 2.4623,
12983
- "step": 42860
12984
- },
12985
- {
12986
- "epoch": 11.082956454322265,
12987
- "learning_rate": 1.3083106443666535e-07,
12988
- "loss": 2.4422,
12989
- "step": 42880
12990
- },
12991
- {
12992
- "epoch": 11.088125080759788,
12993
- "learning_rate": 1.3079865162712303e-07,
12994
- "loss": 2.5212,
12995
- "step": 42900
12996
- },
12997
- {
12998
- "epoch": 11.093293707197311,
12999
- "learning_rate": 1.307662388175807e-07,
13000
- "loss": 2.4923,
13001
- "step": 42920
13002
- },
13003
- {
13004
- "epoch": 11.098462333634837,
13005
- "learning_rate": 1.3073382600803836e-07,
13006
- "loss": 2.5076,
13007
- "step": 42940
13008
- },
13009
- {
13010
- "epoch": 11.10363096007236,
13011
- "learning_rate": 1.3070141319849605e-07,
13012
- "loss": 2.5291,
13013
- "step": 42960
13014
- },
13015
- {
13016
- "epoch": 11.108799586509885,
13017
- "learning_rate": 1.306690003889537e-07,
13018
- "loss": 2.4459,
13019
- "step": 42980
13020
- },
13021
- {
13022
- "epoch": 11.113968212947409,
13023
- "learning_rate": 1.3063658757941137e-07,
13024
- "loss": 2.4522,
13025
- "step": 43000
13026
- },
13027
- {
13028
- "epoch": 11.119136839384934,
13029
- "learning_rate": 1.3060417476986906e-07,
13030
- "loss": 2.4752,
13031
- "step": 43020
13032
- },
13033
- {
13034
- "epoch": 11.124305465822458,
13035
- "learning_rate": 1.3057176196032672e-07,
13036
- "loss": 2.4877,
13037
- "step": 43040
13038
- },
13039
- {
13040
- "epoch": 11.129474092259981,
13041
- "learning_rate": 1.3053934915078438e-07,
13042
- "loss": 2.4969,
13043
- "step": 43060
13044
- },
13045
- {
13046
- "epoch": 11.134642718697506,
13047
- "learning_rate": 1.3050693634124207e-07,
13048
- "loss": 2.5331,
13049
- "step": 43080
13050
- },
13051
- {
13052
- "epoch": 11.13981134513503,
13053
- "learning_rate": 1.304745235316997e-07,
13054
- "loss": 2.4214,
13055
- "step": 43100
13056
- },
13057
- {
13058
- "epoch": 11.144979971572555,
13059
- "learning_rate": 1.304421107221574e-07,
13060
- "loss": 2.4795,
13061
- "step": 43120
13062
- },
13063
- {
13064
- "epoch": 11.150148598010079,
13065
- "learning_rate": 1.3040969791261506e-07,
13066
- "loss": 2.491,
13067
- "step": 43140
13068
- },
13069
- {
13070
- "epoch": 11.155317224447604,
13071
- "learning_rate": 1.3037728510307272e-07,
13072
- "loss": 2.4823,
13073
- "step": 43160
13074
- },
13075
- {
13076
- "epoch": 11.160485850885127,
13077
- "learning_rate": 1.303448722935304e-07,
13078
- "loss": 2.5129,
13079
- "step": 43180
13080
- },
13081
- {
13082
- "epoch": 11.16565447732265,
13083
- "learning_rate": 1.3031245948398807e-07,
13084
- "loss": 2.4543,
13085
- "step": 43200
13086
- },
13087
- {
13088
- "epoch": 11.170823103760176,
13089
- "learning_rate": 1.3028004667444573e-07,
13090
- "loss": 2.4735,
13091
- "step": 43220
13092
- },
13093
- {
13094
- "epoch": 11.1759917301977,
13095
- "learning_rate": 1.3024763386490342e-07,
13096
- "loss": 2.5127,
13097
- "step": 43240
13098
- },
13099
- {
13100
- "epoch": 11.181160356635225,
13101
- "learning_rate": 1.3021522105536105e-07,
13102
- "loss": 2.4045,
13103
- "step": 43260
13104
- },
13105
- {
13106
- "epoch": 11.186328983072748,
13107
- "learning_rate": 1.3018280824581874e-07,
13108
- "loss": 2.4599,
13109
- "step": 43280
13110
- },
13111
- {
13112
- "epoch": 11.191497609510273,
13113
- "learning_rate": 1.301503954362764e-07,
13114
- "loss": 2.4875,
13115
- "step": 43300
13116
- },
13117
- {
13118
- "epoch": 11.196666235947797,
13119
- "learning_rate": 1.3011798262673407e-07,
13120
- "loss": 2.4523,
13121
- "step": 43320
13122
- },
13123
- {
13124
- "epoch": 11.20183486238532,
13125
- "learning_rate": 1.3008556981719175e-07,
13126
- "loss": 2.47,
13127
- "step": 43340
13128
- },
13129
- {
13130
- "epoch": 11.207003488822846,
13131
- "learning_rate": 1.3005315700764942e-07,
13132
- "loss": 2.464,
13133
- "step": 43360
13134
- },
13135
- {
13136
- "epoch": 11.212172115260369,
13137
- "learning_rate": 1.3002074419810708e-07,
13138
- "loss": 2.5223,
13139
- "step": 43380
13140
- },
13141
- {
13142
- "epoch": 11.217340741697894,
13143
- "learning_rate": 1.2998833138856477e-07,
13144
- "loss": 2.4749,
13145
- "step": 43400
13146
- },
13147
- {
13148
- "epoch": 11.222509368135418,
13149
- "learning_rate": 1.2995591857902243e-07,
13150
- "loss": 2.5268,
13151
- "step": 43420
13152
- },
13153
- {
13154
- "epoch": 11.227677994572943,
13155
- "learning_rate": 1.299235057694801e-07,
13156
- "loss": 2.4837,
13157
- "step": 43440
13158
- },
13159
- {
13160
- "epoch": 11.232846621010466,
13161
- "learning_rate": 1.2989109295993778e-07,
13162
- "loss": 2.4285,
13163
- "step": 43460
13164
- },
13165
- {
13166
- "epoch": 11.23801524744799,
13167
- "learning_rate": 1.2985868015039541e-07,
13168
- "loss": 2.4918,
13169
- "step": 43480
13170
- },
13171
- {
13172
- "epoch": 11.243183873885515,
13173
- "learning_rate": 1.298262673408531e-07,
13174
- "loss": 2.4585,
13175
- "step": 43500
13176
- },
13177
- {
13178
- "epoch": 11.248352500323039,
13179
- "learning_rate": 1.2979385453131076e-07,
13180
- "loss": 2.5156,
13181
- "step": 43520
13182
- },
13183
- {
13184
- "epoch": 11.253521126760564,
13185
- "learning_rate": 1.2976144172176843e-07,
13186
- "loss": 2.5031,
13187
- "step": 43540
13188
- },
13189
- {
13190
- "epoch": 11.258689753198087,
13191
- "learning_rate": 1.2972902891222611e-07,
13192
- "loss": 2.456,
13193
- "step": 43560
13194
- },
13195
- {
13196
- "epoch": 11.263858379635613,
13197
- "learning_rate": 1.2969661610268378e-07,
13198
- "loss": 2.5083,
13199
- "step": 43580
13200
- },
13201
- {
13202
- "epoch": 11.269027006073136,
13203
- "learning_rate": 1.2966420329314144e-07,
13204
- "loss": 2.5241,
13205
- "step": 43600
13206
- },
13207
- {
13208
- "epoch": 11.27419563251066,
13209
- "learning_rate": 1.2963179048359913e-07,
13210
- "loss": 2.4751,
13211
- "step": 43620
13212
- },
13213
- {
13214
- "epoch": 11.279364258948185,
13215
- "learning_rate": 1.295993776740568e-07,
13216
- "loss": 2.4695,
13217
- "step": 43640
13218
- },
13219
- {
13220
- "epoch": 11.284532885385708,
13221
- "learning_rate": 1.2956696486451445e-07,
13222
- "loss": 2.5578,
13223
- "step": 43660
13224
- },
13225
- {
13226
- "epoch": 11.289701511823234,
13227
- "learning_rate": 1.2953455205497214e-07,
13228
- "loss": 2.4649,
13229
- "step": 43680
13230
- },
13231
- {
13232
- "epoch": 11.294870138260757,
13233
- "learning_rate": 1.2950213924542977e-07,
13234
- "loss": 2.4747,
13235
- "step": 43700
13236
- },
13237
- {
13238
- "epoch": 11.300038764698282,
13239
- "learning_rate": 1.2946972643588746e-07,
13240
- "loss": 2.4679,
13241
- "step": 43720
13242
- },
13243
- {
13244
- "epoch": 11.305207391135806,
13245
- "learning_rate": 1.2943731362634512e-07,
13246
- "loss": 2.4683,
13247
- "step": 43740
13248
- },
13249
- {
13250
- "epoch": 11.31037601757333,
13251
- "learning_rate": 1.2940490081680279e-07,
13252
- "loss": 2.4857,
13253
- "step": 43760
13254
- },
13255
- {
13256
- "epoch": 11.315544644010854,
13257
- "learning_rate": 1.2937248800726047e-07,
13258
- "loss": 2.426,
13259
- "step": 43780
13260
- },
13261
- {
13262
- "epoch": 11.320713270448378,
13263
- "learning_rate": 1.2934007519771814e-07,
13264
- "loss": 2.4733,
13265
- "step": 43800
13266
- },
13267
- {
13268
- "epoch": 11.325881896885903,
13269
- "learning_rate": 1.293076623881758e-07,
13270
- "loss": 2.4812,
13271
- "step": 43820
13272
- },
13273
- {
13274
- "epoch": 11.331050523323427,
13275
- "learning_rate": 1.2927524957863349e-07,
13276
- "loss": 2.4528,
13277
- "step": 43840
13278
- },
13279
- {
13280
- "epoch": 11.33621914976095,
13281
- "learning_rate": 1.2924283676909112e-07,
13282
- "loss": 2.494,
13283
- "step": 43860
13284
- },
13285
- {
13286
- "epoch": 11.341387776198475,
13287
- "learning_rate": 1.292104239595488e-07,
13288
- "loss": 2.4795,
13289
- "step": 43880
13290
- },
13291
- {
13292
- "epoch": 11.346556402635999,
13293
- "learning_rate": 1.2917801115000647e-07,
13294
- "loss": 2.4816,
13295
- "step": 43900
13296
- },
13297
- {
13298
- "epoch": 11.351725029073524,
13299
- "learning_rate": 1.2914559834046413e-07,
13300
- "loss": 2.5282,
13301
- "step": 43920
13302
- },
13303
- {
13304
- "epoch": 11.356893655511048,
13305
- "learning_rate": 1.2911318553092182e-07,
13306
- "loss": 2.4849,
13307
- "step": 43940
13308
- },
13309
- {
13310
- "epoch": 11.362062281948573,
13311
- "learning_rate": 1.2908077272137948e-07,
13312
- "loss": 2.4246,
13313
- "step": 43960
13314
- },
13315
- {
13316
- "epoch": 11.367230908386096,
13317
- "learning_rate": 1.2904835991183715e-07,
13318
- "loss": 2.5491,
13319
- "step": 43980
13320
- },
13321
- {
13322
- "epoch": 11.372399534823622,
13323
- "learning_rate": 1.2901594710229483e-07,
13324
- "loss": 2.4584,
13325
- "step": 44000
13326
- },
13327
- {
13328
- "epoch": 11.377568161261145,
13329
- "learning_rate": 1.289835342927525e-07,
13330
- "loss": 2.5351,
13331
- "step": 44020
13332
- },
13333
- {
13334
- "epoch": 11.382736787698668,
13335
- "learning_rate": 1.2895112148321016e-07,
13336
- "loss": 2.5108,
13337
- "step": 44040
13338
- },
13339
- {
13340
- "epoch": 11.387905414136194,
13341
- "learning_rate": 1.2891870867366782e-07,
13342
- "loss": 2.5328,
13343
- "step": 44060
13344
- },
13345
- {
13346
- "epoch": 11.393074040573717,
13347
- "learning_rate": 1.2888629586412548e-07,
13348
- "loss": 2.4501,
13349
- "step": 44080
13350
- },
13351
- {
13352
- "epoch": 11.398242667011242,
13353
- "learning_rate": 1.2885388305458317e-07,
13354
- "loss": 2.434,
13355
- "step": 44100
13356
- },
13357
- {
13358
- "epoch": 11.403411293448766,
13359
- "learning_rate": 1.2882147024504083e-07,
13360
- "loss": 2.4713,
13361
- "step": 44120
13362
- },
13363
- {
13364
- "epoch": 11.40857991988629,
13365
- "learning_rate": 1.287890574354985e-07,
13366
- "loss": 2.5093,
13367
- "step": 44140
13368
- },
13369
- {
13370
- "epoch": 11.413748546323815,
13371
- "learning_rate": 1.2875664462595618e-07,
13372
- "loss": 2.5031,
13373
- "step": 44160
13374
- },
13375
- {
13376
- "epoch": 11.418917172761338,
13377
- "learning_rate": 1.2872423181641384e-07,
13378
- "loss": 2.4712,
13379
- "step": 44180
13380
- },
13381
- {
13382
- "epoch": 11.424085799198863,
13383
- "learning_rate": 1.286918190068715e-07,
13384
- "loss": 2.4963,
13385
- "step": 44200
13386
- },
13387
- {
13388
- "epoch": 11.429254425636387,
13389
- "learning_rate": 1.2865940619732917e-07,
13390
- "loss": 2.4866,
13391
- "step": 44220
13392
- },
13393
- {
13394
- "epoch": 11.434423052073912,
13395
- "learning_rate": 1.2862699338778686e-07,
13396
- "loss": 2.4665,
13397
- "step": 44240
13398
- },
13399
- {
13400
- "epoch": 11.439591678511436,
13401
- "learning_rate": 1.2859458057824452e-07,
13402
- "loss": 2.4525,
13403
- "step": 44260
13404
- },
13405
- {
13406
- "epoch": 11.444760304948959,
13407
- "learning_rate": 1.2856216776870218e-07,
13408
- "loss": 2.5271,
13409
- "step": 44280
13410
- },
13411
- {
13412
- "epoch": 11.449928931386484,
13413
- "learning_rate": 1.2852975495915984e-07,
13414
- "loss": 2.4676,
13415
- "step": 44300
13416
- },
13417
- {
13418
- "epoch": 11.455097557824008,
13419
- "learning_rate": 1.2849734214961753e-07,
13420
- "loss": 2.4996,
13421
- "step": 44320
13422
- },
13423
- {
13424
- "epoch": 11.460266184261533,
13425
- "learning_rate": 1.284649293400752e-07,
13426
- "loss": 2.4901,
13427
- "step": 44340
13428
- },
13429
- {
13430
- "epoch": 11.465434810699056,
13431
- "learning_rate": 1.2843251653053285e-07,
13432
- "loss": 2.411,
13433
- "step": 44360
13434
- },
13435
- {
13436
- "epoch": 11.470603437136582,
13437
- "learning_rate": 1.2840010372099054e-07,
13438
- "loss": 2.4688,
13439
- "step": 44380
13440
- },
13441
- {
13442
- "epoch": 11.475772063574105,
13443
- "learning_rate": 1.283676909114482e-07,
13444
- "loss": 2.4458,
13445
- "step": 44400
13446
- },
13447
- {
13448
- "epoch": 11.480940690011629,
13449
- "learning_rate": 1.2833527810190587e-07,
13450
- "loss": 2.4871,
13451
- "step": 44420
13452
- },
13453
- {
13454
- "epoch": 11.486109316449154,
13455
- "learning_rate": 1.2830286529236353e-07,
13456
- "loss": 2.4476,
13457
- "step": 44440
13458
- },
13459
- {
13460
- "epoch": 11.491277942886677,
13461
- "learning_rate": 1.282704524828212e-07,
13462
- "loss": 2.4647,
13463
- "step": 44460
13464
- },
13465
- {
13466
- "epoch": 11.496446569324203,
13467
- "learning_rate": 1.2823803967327888e-07,
13468
- "loss": 2.4887,
13469
- "step": 44480
13470
- },
13471
- {
13472
- "epoch": 11.501615195761726,
13473
- "learning_rate": 1.2820562686373654e-07,
13474
- "loss": 2.4835,
13475
- "step": 44500
13476
- },
13477
- {
13478
- "epoch": 11.506783822199251,
13479
- "learning_rate": 1.281732140541942e-07,
13480
- "loss": 2.4435,
13481
- "step": 44520
13482
- },
13483
- {
13484
- "epoch": 11.511952448636775,
13485
- "learning_rate": 1.281408012446519e-07,
13486
- "loss": 2.4716,
13487
- "step": 44540
13488
- },
13489
- {
13490
- "epoch": 11.517121075074298,
13491
- "learning_rate": 1.2810838843510955e-07,
13492
- "loss": 2.457,
13493
- "step": 44560
13494
- },
13495
- {
13496
- "epoch": 11.522289701511824,
13497
- "learning_rate": 1.2807597562556721e-07,
13498
- "loss": 2.4465,
13499
- "step": 44580
13500
- },
13501
- {
13502
- "epoch": 11.527458327949347,
13503
- "learning_rate": 1.2804356281602488e-07,
13504
- "loss": 2.4804,
13505
- "step": 44600
13506
- },
13507
- {
13508
- "epoch": 11.532626954386872,
13509
- "learning_rate": 1.2801115000648256e-07,
13510
- "loss": 2.5168,
13511
- "step": 44620
13512
- },
13513
- {
13514
- "epoch": 11.537795580824396,
13515
- "learning_rate": 1.2797873719694023e-07,
13516
- "loss": 2.4759,
13517
- "step": 44640
13518
- },
13519
- {
13520
- "epoch": 11.542964207261921,
13521
- "learning_rate": 1.279463243873979e-07,
13522
- "loss": 2.4929,
13523
- "step": 44660
13524
- },
13525
- {
13526
- "epoch": 11.548132833699444,
13527
- "learning_rate": 1.2791391157785555e-07,
13528
- "loss": 2.4136,
13529
- "step": 44680
13530
- },
13531
- {
13532
- "epoch": 11.553301460136968,
13533
- "learning_rate": 1.2788149876831324e-07,
13534
- "loss": 2.4497,
13535
- "step": 44700
13536
- },
13537
- {
13538
- "epoch": 11.558470086574493,
13539
- "learning_rate": 1.278490859587709e-07,
13540
- "loss": 2.4912,
13541
- "step": 44720
13542
- },
13543
- {
13544
- "epoch": 11.563638713012017,
13545
- "learning_rate": 1.2781667314922856e-07,
13546
- "loss": 2.5089,
13547
- "step": 44740
13548
- },
13549
- {
13550
- "epoch": 11.568807339449542,
13551
- "learning_rate": 1.2778426033968622e-07,
13552
- "loss": 2.4516,
13553
- "step": 44760
13554
- },
13555
- {
13556
- "epoch": 11.573975965887065,
13557
- "learning_rate": 1.277518475301439e-07,
13558
- "loss": 2.5023,
13559
- "step": 44780
13560
- },
13561
- {
13562
- "epoch": 11.579144592324589,
13563
- "learning_rate": 1.2771943472060157e-07,
13564
- "loss": 2.4883,
13565
- "step": 44800
13566
- },
13567
- {
13568
- "epoch": 11.584313218762114,
13569
- "learning_rate": 1.2768702191105924e-07,
13570
- "loss": 2.4885,
13571
- "step": 44820
13572
- },
13573
- {
13574
- "epoch": 11.589481845199638,
13575
- "learning_rate": 1.2765460910151692e-07,
13576
- "loss": 2.5421,
13577
- "step": 44840
13578
- },
13579
- {
13580
- "epoch": 11.594650471637163,
13581
- "learning_rate": 1.2762219629197459e-07,
13582
- "loss": 2.4938,
13583
- "step": 44860
13584
- },
13585
- {
13586
- "epoch": 11.599819098074686,
13587
- "learning_rate": 1.2758978348243225e-07,
13588
- "loss": 2.4712,
13589
- "step": 44880
13590
- },
13591
- {
13592
- "epoch": 11.604987724512211,
13593
- "learning_rate": 1.275573706728899e-07,
13594
- "loss": 2.5227,
13595
- "step": 44900
13596
- },
13597
- {
13598
- "epoch": 11.610156350949735,
13599
- "learning_rate": 1.275249578633476e-07,
13600
- "loss": 2.4463,
13601
- "step": 44920
13602
- },
13603
- {
13604
- "epoch": 11.61532497738726,
13605
- "learning_rate": 1.2749254505380526e-07,
13606
- "loss": 2.4526,
13607
- "step": 44940
13608
- },
13609
- {
13610
- "epoch": 11.620493603824784,
13611
- "learning_rate": 1.2746013224426292e-07,
13612
- "loss": 2.4763,
13613
- "step": 44960
13614
- },
13615
- {
13616
- "epoch": 11.625662230262307,
13617
- "learning_rate": 1.2742771943472058e-07,
13618
- "loss": 2.5575,
13619
- "step": 44980
13620
- },
13621
- {
13622
- "epoch": 11.630830856699832,
13623
- "learning_rate": 1.2739530662517827e-07,
13624
- "loss": 2.5079,
13625
- "step": 45000
13626
- },
13627
- {
13628
- "epoch": 11.635999483137356,
13629
- "learning_rate": 1.2736289381563593e-07,
13630
- "loss": 2.4868,
13631
- "step": 45020
13632
- },
13633
- {
13634
- "epoch": 11.641168109574881,
13635
- "learning_rate": 1.273304810060936e-07,
13636
- "loss": 2.4547,
13637
- "step": 45040
13638
- },
13639
- {
13640
- "epoch": 11.646336736012405,
13641
- "learning_rate": 1.2729806819655126e-07,
13642
- "loss": 2.4594,
13643
- "step": 45060
13644
- },
13645
- {
13646
- "epoch": 11.651505362449928,
13647
- "learning_rate": 1.2726565538700895e-07,
13648
- "loss": 2.495,
13649
- "step": 45080
13650
- },
13651
- {
13652
- "epoch": 11.656673988887453,
13653
- "learning_rate": 1.272332425774666e-07,
13654
- "loss": 2.5174,
13655
- "step": 45100
13656
- },
13657
- {
13658
- "epoch": 11.661842615324977,
13659
- "learning_rate": 1.2720082976792427e-07,
13660
- "loss": 2.4223,
13661
- "step": 45120
13662
- },
13663
- {
13664
- "epoch": 11.667011241762502,
13665
- "learning_rate": 1.2716841695838196e-07,
13666
- "loss": 2.4523,
13667
- "step": 45140
13668
- },
13669
- {
13670
- "epoch": 11.672179868200026,
13671
- "learning_rate": 1.2713600414883962e-07,
13672
- "loss": 2.4271,
13673
- "step": 45160
13674
- },
13675
- {
13676
- "epoch": 11.67734849463755,
13677
- "learning_rate": 1.2710359133929728e-07,
13678
- "loss": 2.5235,
13679
- "step": 45180
13680
- },
13681
- {
13682
- "epoch": 11.682517121075074,
13683
- "learning_rate": 1.2707117852975494e-07,
13684
- "loss": 2.4666,
13685
- "step": 45200
13686
- },
13687
- {
13688
- "epoch": 11.687685747512598,
13689
- "learning_rate": 1.2703876572021263e-07,
13690
- "loss": 2.4413,
13691
- "step": 45220
13692
- },
13693
- {
13694
- "epoch": 11.692854373950123,
13695
- "learning_rate": 1.270063529106703e-07,
13696
- "loss": 2.422,
13697
- "step": 45240
13698
- },
13699
- {
13700
- "epoch": 11.698023000387646,
13701
- "learning_rate": 1.2697394010112796e-07,
13702
- "loss": 2.5057,
13703
- "step": 45260
13704
- },
13705
- {
13706
- "epoch": 11.703191626825172,
13707
- "learning_rate": 1.2694152729158562e-07,
13708
- "loss": 2.5153,
13709
- "step": 45280
13710
- },
13711
- {
13712
- "epoch": 11.708360253262695,
13713
- "learning_rate": 1.269091144820433e-07,
13714
- "loss": 2.4894,
13715
- "step": 45300
13716
- },
13717
- {
13718
- "epoch": 11.71352887970022,
13719
- "learning_rate": 1.2687670167250097e-07,
13720
- "loss": 2.5211,
13721
- "step": 45320
13722
- },
13723
- {
13724
- "epoch": 11.718697506137744,
13725
- "learning_rate": 1.2684428886295863e-07,
13726
- "loss": 2.4449,
13727
- "step": 45340
13728
- },
13729
- {
13730
- "epoch": 11.723866132575267,
13731
- "learning_rate": 1.268118760534163e-07,
13732
- "loss": 2.4859,
13733
- "step": 45360
13734
- },
13735
- {
13736
- "epoch": 11.729034759012793,
13737
- "learning_rate": 1.2677946324387398e-07,
13738
- "loss": 2.4578,
13739
- "step": 45380
13740
- },
13741
- {
13742
- "epoch": 11.734203385450316,
13743
- "learning_rate": 1.2674705043433164e-07,
13744
- "loss": 2.4715,
13745
- "step": 45400
13746
- },
13747
- {
13748
- "epoch": 11.739372011887841,
13749
- "learning_rate": 1.267146376247893e-07,
13750
- "loss": 2.4258,
13751
- "step": 45420
13752
- },
13753
- {
13754
- "epoch": 11.744540638325365,
13755
- "learning_rate": 1.26682224815247e-07,
13756
- "loss": 2.4635,
13757
- "step": 45440
13758
- },
13759
- {
13760
- "epoch": 11.74970926476289,
13761
- "learning_rate": 1.2664981200570465e-07,
13762
- "loss": 2.4803,
13763
- "step": 45460
13764
- },
13765
- {
13766
- "epoch": 11.754877891200413,
13767
- "learning_rate": 1.2661739919616232e-07,
13768
- "loss": 2.4333,
13769
- "step": 45480
13770
- },
13771
- {
13772
- "epoch": 11.760046517637937,
13773
- "learning_rate": 1.2658498638661998e-07,
13774
- "loss": 2.5486,
13775
- "step": 45500
13776
- },
13777
- {
13778
- "epoch": 11.765215144075462,
13779
- "learning_rate": 1.2655257357707767e-07,
13780
- "loss": 2.4803,
13781
- "step": 45520
13782
- },
13783
- {
13784
- "epoch": 11.770383770512986,
13785
- "learning_rate": 1.2652016076753533e-07,
13786
- "loss": 2.4958,
13787
- "step": 45540
13788
- },
13789
- {
13790
- "epoch": 11.775552396950511,
13791
- "learning_rate": 1.26487747957993e-07,
13792
- "loss": 2.4918,
13793
- "step": 45560
13794
- },
13795
- {
13796
- "epoch": 11.780721023388034,
13797
- "learning_rate": 1.2645533514845065e-07,
13798
- "loss": 2.513,
13799
- "step": 45580
13800
- },
13801
- {
13802
- "epoch": 11.78588964982556,
13803
- "learning_rate": 1.2642292233890834e-07,
13804
- "loss": 2.4897,
13805
- "step": 45600
13806
- },
13807
- {
13808
- "epoch": 11.791058276263083,
13809
- "learning_rate": 1.26390509529366e-07,
13810
- "loss": 2.4468,
13811
- "step": 45620
13812
- },
13813
- {
13814
- "epoch": 11.796226902700607,
13815
- "learning_rate": 1.2635809671982366e-07,
13816
- "loss": 2.4928,
13817
- "step": 45640
13818
- },
13819
- {
13820
- "epoch": 11.801395529138132,
13821
- "learning_rate": 1.2632568391028132e-07,
13822
- "loss": 2.4793,
13823
- "step": 45660
13824
- },
13825
- {
13826
- "epoch": 11.806564155575655,
13827
- "learning_rate": 1.26293271100739e-07,
13828
- "loss": 2.4788,
13829
- "step": 45680
13830
- },
13831
- {
13832
- "epoch": 11.81173278201318,
13833
- "learning_rate": 1.2626085829119668e-07,
13834
- "loss": 2.4913,
13835
- "step": 45700
13836
- },
13837
- {
13838
- "epoch": 11.816901408450704,
13839
- "learning_rate": 1.2622844548165434e-07,
13840
- "loss": 2.4966,
13841
- "step": 45720
13842
- },
13843
- {
13844
- "epoch": 11.82207003488823,
13845
- "learning_rate": 1.2619603267211203e-07,
13846
- "loss": 2.4954,
13847
- "step": 45740
13848
- },
13849
- {
13850
- "epoch": 11.827238661325753,
13851
- "learning_rate": 1.261636198625697e-07,
13852
- "loss": 2.4608,
13853
- "step": 45760
13854
- },
13855
- {
13856
- "epoch": 11.832407287763276,
13857
- "learning_rate": 1.2613120705302735e-07,
13858
- "loss": 2.4901,
13859
- "step": 45780
13860
- },
13861
- {
13862
- "epoch": 11.837575914200801,
13863
- "learning_rate": 1.26098794243485e-07,
13864
- "loss": 2.5462,
13865
- "step": 45800
13866
- },
13867
- {
13868
- "epoch": 11.842744540638325,
13869
- "learning_rate": 1.260663814339427e-07,
13870
- "loss": 2.5217,
13871
- "step": 45820
13872
- },
13873
- {
13874
- "epoch": 11.84791316707585,
13875
- "learning_rate": 1.2603396862440036e-07,
13876
- "loss": 2.4889,
13877
- "step": 45840
13878
- },
13879
- {
13880
- "epoch": 11.853081793513374,
13881
- "learning_rate": 1.2600155581485802e-07,
13882
- "loss": 2.5021,
13883
- "step": 45860
13884
- },
13885
- {
13886
- "epoch": 11.858250419950899,
13887
- "learning_rate": 1.2596914300531568e-07,
13888
- "loss": 2.5027,
13889
- "step": 45880
13890
- },
13891
- {
13892
- "epoch": 11.863419046388422,
13893
- "learning_rate": 1.2593673019577337e-07,
13894
- "loss": 2.4681,
13895
- "step": 45900
13896
- },
13897
- {
13898
- "epoch": 11.868587672825946,
13899
- "learning_rate": 1.2590431738623104e-07,
13900
- "loss": 2.409,
13901
- "step": 45920
13902
- },
13903
- {
13904
- "epoch": 11.873756299263471,
13905
- "learning_rate": 1.258719045766887e-07,
13906
- "loss": 2.4426,
13907
- "step": 45940
13908
- },
13909
- {
13910
- "epoch": 11.878924925700995,
13911
- "learning_rate": 1.2583949176714636e-07,
13912
- "loss": 2.4285,
13913
- "step": 45960
13914
- },
13915
- {
13916
- "epoch": 11.88409355213852,
13917
- "learning_rate": 1.2580707895760405e-07,
13918
- "loss": 2.4452,
13919
- "step": 45980
13920
- },
13921
- {
13922
- "epoch": 11.889262178576043,
13923
- "learning_rate": 1.257746661480617e-07,
13924
- "loss": 2.4816,
13925
- "step": 46000
13926
- },
13927
- {
13928
- "epoch": 11.894430805013567,
13929
- "learning_rate": 1.2574225333851937e-07,
13930
- "loss": 2.4585,
13931
- "step": 46020
13932
- },
13933
- {
13934
- "epoch": 11.899599431451092,
13935
- "learning_rate": 1.2570984052897706e-07,
13936
- "loss": 2.4171,
13937
- "step": 46040
13938
- },
13939
- {
13940
- "epoch": 11.904768057888615,
13941
- "learning_rate": 1.2567742771943472e-07,
13942
- "loss": 2.4597,
13943
- "step": 46060
13944
- },
13945
- {
13946
- "epoch": 11.90993668432614,
13947
- "learning_rate": 1.2564501490989238e-07,
13948
- "loss": 2.4195,
13949
- "step": 46080
13950
- },
13951
- {
13952
- "epoch": 11.915105310763664,
13953
- "learning_rate": 1.2561260210035004e-07,
13954
- "loss": 2.5034,
13955
- "step": 46100
13956
- },
13957
- {
13958
- "epoch": 11.92027393720119,
13959
- "learning_rate": 1.2558018929080773e-07,
13960
- "loss": 2.4407,
13961
- "step": 46120
13962
- },
13963
- {
13964
- "epoch": 11.925442563638713,
13965
- "learning_rate": 1.255477764812654e-07,
13966
- "loss": 2.4813,
13967
- "step": 46140
13968
- },
13969
- {
13970
- "epoch": 11.930611190076238,
13971
- "learning_rate": 1.2551536367172306e-07,
13972
- "loss": 2.4599,
13973
- "step": 46160
13974
- },
13975
- {
13976
- "epoch": 11.935779816513762,
13977
- "learning_rate": 1.2548295086218072e-07,
13978
- "loss": 2.4641,
13979
- "step": 46180
13980
- },
13981
- {
13982
- "epoch": 11.940948442951285,
13983
- "learning_rate": 1.254505380526384e-07,
13984
- "loss": 2.5262,
13985
- "step": 46200
13986
- },
13987
- {
13988
- "epoch": 11.94611706938881,
13989
- "learning_rate": 1.2541812524309607e-07,
13990
- "loss": 2.4518,
13991
- "step": 46220
13992
- },
13993
- {
13994
- "epoch": 11.951285695826334,
13995
- "learning_rate": 1.2538571243355373e-07,
13996
- "loss": 2.5018,
13997
- "step": 46240
13998
- },
13999
- {
14000
- "epoch": 11.956454322263859,
14001
- "learning_rate": 1.253532996240114e-07,
14002
- "loss": 2.4634,
14003
- "step": 46260
14004
- },
14005
- {
14006
- "epoch": 11.961622948701383,
14007
- "learning_rate": 1.2532088681446908e-07,
14008
- "loss": 2.4889,
14009
- "step": 46280
14010
- },
14011
- {
14012
- "epoch": 11.966791575138906,
14013
- "learning_rate": 1.2528847400492674e-07,
14014
- "loss": 2.4577,
14015
- "step": 46300
14016
- },
14017
- {
14018
- "epoch": 11.971960201576431,
14019
- "learning_rate": 1.252560611953844e-07,
14020
- "loss": 2.4812,
14021
- "step": 46320
14022
- },
14023
- {
14024
- "epoch": 11.977128828013955,
14025
- "learning_rate": 1.252236483858421e-07,
14026
- "loss": 2.4448,
14027
- "step": 46340
14028
- },
14029
- {
14030
- "epoch": 11.98229745445148,
14031
- "learning_rate": 1.2519123557629975e-07,
14032
- "loss": 2.4594,
14033
- "step": 46360
14034
- },
14035
- {
14036
- "epoch": 11.987466080889003,
14037
- "learning_rate": 1.2515882276675742e-07,
14038
- "loss": 2.4473,
14039
- "step": 46380
14040
- },
14041
- {
14042
- "epoch": 11.992634707326529,
14043
- "learning_rate": 1.2512640995721508e-07,
14044
- "loss": 2.4831,
14045
- "step": 46400
14046
- },
14047
- {
14048
- "epoch": 11.997803333764052,
14049
- "learning_rate": 1.2509399714767277e-07,
14050
- "loss": 2.4706,
14051
- "step": 46420
14052
- },
14053
- {
14054
- "epoch": 11.999870784339063,
14055
- "eval_bleu": 12.51,
14056
- "eval_gen_len": 37.4146,
14057
- "eval_loss": 2.4478776454925537,
14058
- "eval_runtime": 807.7318,
14059
- "eval_samples_per_second": 2.132,
14060
- "eval_steps_per_second": 1.066,
14061
- "step": 46428
14062
- },
14063
- {
14064
- "epoch": 12.002971960201576,
14065
- "learning_rate": 1.2506158433813043e-07,
14066
- "loss": 2.4527,
14067
- "step": 46440
14068
- },
14069
- {
14070
- "epoch": 12.008140586639101,
14071
- "learning_rate": 1.250291715285881e-07,
14072
- "loss": 2.436,
14073
- "step": 46460
14074
- },
14075
- {
14076
- "epoch": 12.013309213076624,
14077
- "learning_rate": 1.2499675871904575e-07,
14078
- "loss": 2.4514,
14079
- "step": 46480
14080
- },
14081
- {
14082
- "epoch": 12.01847783951415,
14083
- "learning_rate": 1.2496434590950344e-07,
14084
- "loss": 2.4921,
14085
- "step": 46500
14086
- },
14087
- {
14088
- "epoch": 12.023646465951673,
14089
- "learning_rate": 1.249319330999611e-07,
14090
- "loss": 2.4637,
14091
- "step": 46520
14092
- },
14093
- {
14094
- "epoch": 12.028815092389198,
14095
- "learning_rate": 1.2489952029041876e-07,
14096
- "loss": 2.4924,
14097
- "step": 46540
14098
- },
14099
- {
14100
- "epoch": 12.033983718826722,
14101
- "learning_rate": 1.2486710748087643e-07,
14102
- "loss": 2.4788,
14103
- "step": 46560
14104
- },
14105
- {
14106
- "epoch": 12.039152345264245,
14107
- "learning_rate": 1.2483469467133411e-07,
14108
- "loss": 2.5291,
14109
- "step": 46580
14110
- },
14111
- {
14112
- "epoch": 12.04432097170177,
14113
- "learning_rate": 1.2480228186179178e-07,
14114
- "loss": 2.4419,
14115
- "step": 46600
14116
- },
14117
- {
14118
- "epoch": 12.049489598139294,
14119
- "learning_rate": 1.2476986905224944e-07,
14120
- "loss": 2.4768,
14121
- "step": 46620
14122
- },
14123
- {
14124
- "epoch": 12.05465822457682,
14125
- "learning_rate": 1.2473745624270713e-07,
14126
- "loss": 2.444,
14127
- "step": 46640
14128
- },
14129
- {
14130
- "epoch": 12.059826851014343,
14131
- "learning_rate": 1.247050434331648e-07,
14132
- "loss": 2.5193,
14133
- "step": 46660
14134
- },
14135
- {
14136
- "epoch": 12.064995477451868,
14137
- "learning_rate": 1.2467263062362245e-07,
14138
- "loss": 2.5019,
14139
- "step": 46680
14140
- },
14141
- {
14142
- "epoch": 12.070164103889391,
14143
- "learning_rate": 1.246402178140801e-07,
14144
- "loss": 2.465,
14145
- "step": 46700
14146
- },
14147
- {
14148
- "epoch": 12.075332730326915,
14149
- "learning_rate": 1.246078050045378e-07,
14150
- "loss": 2.4508,
14151
- "step": 46720
14152
- },
14153
- {
14154
- "epoch": 12.08050135676444,
14155
- "learning_rate": 1.2457539219499546e-07,
14156
- "loss": 2.4784,
14157
- "step": 46740
14158
- },
14159
- {
14160
- "epoch": 12.085669983201964,
14161
- "learning_rate": 1.2454297938545312e-07,
14162
- "loss": 2.5013,
14163
- "step": 46760
14164
- },
14165
- {
14166
- "epoch": 12.090838609639489,
14167
- "learning_rate": 1.2451056657591079e-07,
14168
- "loss": 2.3997,
14169
- "step": 46780
14170
- },
14171
- {
14172
- "epoch": 12.096007236077012,
14173
- "learning_rate": 1.2447815376636847e-07,
14174
- "loss": 2.4475,
14175
- "step": 46800
14176
- },
14177
- {
14178
- "epoch": 12.101175862514538,
14179
- "learning_rate": 1.2444574095682614e-07,
14180
- "loss": 2.4564,
14181
- "step": 46820
14182
- },
14183
- {
14184
- "epoch": 12.106344488952061,
14185
- "learning_rate": 1.244133281472838e-07,
14186
- "loss": 2.5028,
14187
- "step": 46840
14188
- },
14189
- {
14190
- "epoch": 12.111513115389585,
14191
- "learning_rate": 1.2438091533774146e-07,
14192
- "loss": 2.4969,
14193
- "step": 46860
14194
- },
14195
- {
14196
- "epoch": 12.11668174182711,
14197
- "learning_rate": 1.2434850252819915e-07,
14198
- "loss": 2.4488,
14199
- "step": 46880
14200
- },
14201
- {
14202
- "epoch": 12.121850368264633,
14203
- "learning_rate": 1.243160897186568e-07,
14204
- "loss": 2.4779,
14205
- "step": 46900
14206
- },
14207
- {
14208
- "epoch": 12.127018994702158,
14209
- "learning_rate": 1.2428367690911447e-07,
14210
- "loss": 2.4605,
14211
- "step": 46920
14212
- },
14213
- {
14214
- "epoch": 12.132187621139682,
14215
- "learning_rate": 1.2425126409957216e-07,
14216
- "loss": 2.4592,
14217
- "step": 46940
14218
- },
14219
- {
14220
- "epoch": 12.137356247577207,
14221
- "learning_rate": 1.2421885129002982e-07,
14222
- "loss": 2.5115,
14223
- "step": 46960
14224
- },
14225
- {
14226
- "epoch": 12.14252487401473,
14227
- "learning_rate": 1.2418643848048748e-07,
14228
- "loss": 2.494,
14229
- "step": 46980
14230
- },
14231
- {
14232
- "epoch": 12.147693500452254,
14233
- "learning_rate": 1.2415402567094515e-07,
14234
- "loss": 2.4545,
14235
- "step": 47000
14236
- },
14237
- {
14238
- "epoch": 12.15286212688978,
14239
- "learning_rate": 1.2412161286140283e-07,
14240
- "loss": 2.5211,
14241
- "step": 47020
14242
- },
14243
- {
14244
- "epoch": 12.158030753327303,
14245
- "learning_rate": 1.240892000518605e-07,
14246
- "loss": 2.4661,
14247
- "step": 47040
14248
- },
14249
- {
14250
- "epoch": 12.163199379764828,
14251
- "learning_rate": 1.2405678724231816e-07,
14252
- "loss": 2.4593,
14253
- "step": 47060
14254
- },
14255
- {
14256
- "epoch": 12.168368006202352,
14257
- "learning_rate": 1.2402437443277582e-07,
14258
- "loss": 2.555,
14259
- "step": 47080
14260
- },
14261
- {
14262
- "epoch": 12.173536632639877,
14263
- "learning_rate": 1.239919616232335e-07,
14264
- "loss": 2.5006,
14265
- "step": 47100
14266
- },
14267
- {
14268
- "epoch": 12.1787052590774,
14269
- "learning_rate": 1.2395954881369117e-07,
14270
- "loss": 2.4607,
14271
- "step": 47120
14272
- },
14273
- {
14274
- "epoch": 12.183873885514924,
14275
- "learning_rate": 1.2392713600414883e-07,
14276
- "loss": 2.4595,
14277
- "step": 47140
14278
- },
14279
- {
14280
- "epoch": 12.189042511952449,
14281
- "learning_rate": 1.238947231946065e-07,
14282
- "loss": 2.444,
14283
- "step": 47160
14284
- },
14285
- {
14286
- "epoch": 12.194211138389973,
14287
- "learning_rate": 1.2386231038506418e-07,
14288
- "loss": 2.4657,
14289
- "step": 47180
14290
- },
14291
- {
14292
- "epoch": 12.199379764827498,
14293
- "learning_rate": 1.2382989757552184e-07,
14294
- "loss": 2.454,
14295
- "step": 47200
14296
- },
14297
- {
14298
- "epoch": 12.204548391265021,
14299
- "learning_rate": 1.237974847659795e-07,
14300
- "loss": 2.4939,
14301
- "step": 47220
14302
- },
14303
- {
14304
- "epoch": 12.209717017702545,
14305
- "learning_rate": 1.237650719564372e-07,
14306
- "loss": 2.4465,
14307
- "step": 47240
14308
- },
14309
- {
14310
- "epoch": 12.21488564414007,
14311
- "learning_rate": 1.2373265914689486e-07,
14312
- "loss": 2.5464,
14313
- "step": 47260
14314
- },
14315
- {
14316
- "epoch": 12.220054270577593,
14317
- "learning_rate": 1.2370024633735252e-07,
14318
- "loss": 2.4521,
14319
- "step": 47280
14320
- },
14321
- {
14322
- "epoch": 12.225222897015119,
14323
- "learning_rate": 1.2366783352781018e-07,
14324
- "loss": 2.5132,
14325
- "step": 47300
14326
- },
14327
- {
14328
- "epoch": 12.230391523452642,
14329
- "learning_rate": 1.2363542071826787e-07,
14330
- "loss": 2.43,
14331
- "step": 47320
14332
- },
14333
- {
14334
- "epoch": 12.235560149890167,
14335
- "learning_rate": 1.2360300790872553e-07,
14336
- "loss": 2.4295,
14337
- "step": 47340
14338
- },
14339
- {
14340
- "epoch": 12.24072877632769,
14341
- "learning_rate": 1.235705950991832e-07,
14342
- "loss": 2.5134,
14343
- "step": 47360
14344
- },
14345
- {
14346
- "epoch": 12.245897402765214,
14347
- "learning_rate": 1.2353818228964085e-07,
14348
- "loss": 2.461,
14349
- "step": 47380
14350
- },
14351
- {
14352
- "epoch": 12.25106602920274,
14353
- "learning_rate": 1.2350576948009854e-07,
14354
- "loss": 2.4942,
14355
- "step": 47400
14356
- },
14357
- {
14358
- "epoch": 12.256234655640263,
14359
- "learning_rate": 1.234733566705562e-07,
14360
- "loss": 2.4607,
14361
- "step": 47420
14362
- },
14363
- {
14364
- "epoch": 12.261403282077788,
14365
- "learning_rate": 1.2344094386101387e-07,
14366
- "loss": 2.4634,
14367
- "step": 47440
14368
- },
14369
- {
14370
- "epoch": 12.266571908515312,
14371
- "learning_rate": 1.2340853105147153e-07,
14372
- "loss": 2.4448,
14373
- "step": 47460
14374
- },
14375
- {
14376
- "epoch": 12.271740534952837,
14377
- "learning_rate": 1.2337611824192922e-07,
14378
- "loss": 2.4803,
14379
- "step": 47480
14380
- },
14381
- {
14382
- "epoch": 12.27690916139036,
14383
- "learning_rate": 1.2334370543238688e-07,
14384
- "loss": 2.4154,
14385
- "step": 47500
14386
- },
14387
- {
14388
- "epoch": 12.282077787827884,
14389
- "learning_rate": 1.2331129262284454e-07,
14390
- "loss": 2.5181,
14391
- "step": 47520
14392
- },
14393
- {
14394
- "epoch": 12.28724641426541,
14395
- "learning_rate": 1.2327887981330223e-07,
14396
- "loss": 2.5086,
14397
- "step": 47540
14398
- },
14399
- {
14400
- "epoch": 12.292415040702933,
14401
- "learning_rate": 1.232464670037599e-07,
14402
- "loss": 2.4487,
14403
- "step": 47560
14404
- },
14405
- {
14406
- "epoch": 12.297583667140458,
14407
- "learning_rate": 1.2321405419421755e-07,
14408
- "loss": 2.4109,
14409
- "step": 47580
14410
- },
14411
- {
14412
- "epoch": 12.302752293577981,
14413
- "learning_rate": 1.2318164138467521e-07,
14414
- "loss": 2.4675,
14415
- "step": 47600
14416
- },
14417
- {
14418
- "epoch": 12.307920920015507,
14419
- "learning_rate": 1.2314922857513288e-07,
14420
- "loss": 2.4326,
14421
- "step": 47620
14422
- },
14423
- {
14424
- "epoch": 12.31308954645303,
14425
- "learning_rate": 1.2311681576559056e-07,
14426
- "loss": 2.441,
14427
- "step": 47640
14428
- },
14429
- {
14430
- "epoch": 12.318258172890554,
14431
- "learning_rate": 1.2308440295604823e-07,
14432
- "loss": 2.4154,
14433
- "step": 47660
14434
- },
14435
- {
14436
- "epoch": 12.323426799328079,
14437
- "learning_rate": 1.230519901465059e-07,
14438
- "loss": 2.4491,
14439
- "step": 47680
14440
- },
14441
- {
14442
- "epoch": 12.328595425765602,
14443
- "learning_rate": 1.2301957733696358e-07,
14444
- "loss": 2.4764,
14445
- "step": 47700
14446
- },
14447
- {
14448
- "epoch": 12.333764052203128,
14449
- "learning_rate": 1.2298716452742124e-07,
14450
- "loss": 2.4786,
14451
- "step": 47720
14452
- },
14453
- {
14454
- "epoch": 12.338932678640651,
14455
- "learning_rate": 1.229547517178789e-07,
14456
- "loss": 2.4033,
14457
- "step": 47740
14458
- },
14459
- {
14460
- "epoch": 12.344101305078176,
14461
- "learning_rate": 1.2292233890833656e-07,
14462
- "loss": 2.4742,
14463
- "step": 47760
14464
- },
14465
- {
14466
- "epoch": 12.3492699315157,
14467
- "learning_rate": 1.2288992609879422e-07,
14468
- "loss": 2.5316,
14469
- "step": 47780
14470
- },
14471
- {
14472
- "epoch": 12.354438557953223,
14473
- "learning_rate": 1.228575132892519e-07,
14474
- "loss": 2.4717,
14475
- "step": 47800
14476
- },
14477
- {
14478
- "epoch": 12.359607184390748,
14479
- "learning_rate": 1.2282510047970957e-07,
14480
- "loss": 2.4518,
14481
- "step": 47820
14482
- },
14483
- {
14484
- "epoch": 12.364775810828272,
14485
- "learning_rate": 1.2279268767016724e-07,
14486
- "loss": 2.4747,
14487
- "step": 47840
14488
- },
14489
- {
14490
- "epoch": 12.369944437265797,
14491
- "learning_rate": 1.2276027486062492e-07,
14492
- "loss": 2.4391,
14493
- "step": 47860
14494
- },
14495
- {
14496
- "epoch": 12.37511306370332,
14497
- "learning_rate": 1.2272786205108259e-07,
14498
- "loss": 2.4076,
14499
- "step": 47880
14500
- },
14501
- {
14502
- "epoch": 12.380281690140846,
14503
- "learning_rate": 1.2269544924154025e-07,
14504
- "loss": 2.4275,
14505
- "step": 47900
14506
- },
14507
- {
14508
- "epoch": 12.38545031657837,
14509
- "learning_rate": 1.2266303643199794e-07,
14510
- "loss": 2.4593,
14511
- "step": 47920
14512
- },
14513
- {
14514
- "epoch": 12.390618943015893,
14515
- "learning_rate": 1.226306236224556e-07,
14516
- "loss": 2.4666,
14517
- "step": 47940
14518
- },
14519
- {
14520
- "epoch": 12.395787569453418,
14521
- "learning_rate": 1.2259821081291326e-07,
14522
- "loss": 2.4939,
14523
- "step": 47960
14524
- },
14525
- {
14526
- "epoch": 12.400956195890942,
14527
- "learning_rate": 1.2256579800337092e-07,
14528
- "loss": 2.4933,
14529
- "step": 47980
14530
- },
14531
- {
14532
- "epoch": 12.406124822328467,
14533
- "learning_rate": 1.2253338519382858e-07,
14534
- "loss": 2.4298,
14535
- "step": 48000
14536
- },
14537
- {
14538
- "epoch": 12.41129344876599,
14539
- "learning_rate": 1.2250097238428627e-07,
14540
- "loss": 2.4923,
14541
- "step": 48020
14542
- },
14543
- {
14544
- "epoch": 12.416462075203516,
14545
- "learning_rate": 1.2246855957474393e-07,
14546
- "loss": 2.4368,
14547
- "step": 48040
14548
- },
14549
- {
14550
- "epoch": 12.421630701641039,
14551
- "learning_rate": 1.224361467652016e-07,
14552
- "loss": 2.4882,
14553
- "step": 48060
14554
- },
14555
- {
14556
- "epoch": 12.426799328078562,
14557
- "learning_rate": 1.2240373395565928e-07,
14558
- "loss": 2.4417,
14559
- "step": 48080
14560
- },
14561
- {
14562
- "epoch": 12.431967954516088,
14563
- "learning_rate": 1.2237132114611695e-07,
14564
- "loss": 2.4881,
14565
- "step": 48100
14566
- },
14567
- {
14568
- "epoch": 12.437136580953611,
14569
- "learning_rate": 1.223389083365746e-07,
14570
- "loss": 2.4633,
14571
- "step": 48120
14572
- },
14573
- {
14574
- "epoch": 12.442305207391136,
14575
- "learning_rate": 1.223064955270323e-07,
14576
- "loss": 2.4914,
14577
- "step": 48140
14578
- },
14579
- {
14580
- "epoch": 12.44747383382866,
14581
- "learning_rate": 1.2227408271748993e-07,
14582
- "loss": 2.4236,
14583
- "step": 48160
14584
- },
14585
- {
14586
- "epoch": 12.452642460266183,
14587
- "learning_rate": 1.2224166990794762e-07,
14588
- "loss": 2.4603,
14589
- "step": 48180
14590
- },
14591
- {
14592
- "epoch": 12.457811086703709,
14593
- "learning_rate": 1.2220925709840528e-07,
14594
- "loss": 2.5019,
14595
- "step": 48200
14596
- },
14597
- {
14598
- "epoch": 12.462979713141232,
14599
- "learning_rate": 1.2217684428886294e-07,
14600
- "loss": 2.4368,
14601
- "step": 48220
14602
- },
14603
- {
14604
- "epoch": 12.468148339578757,
14605
- "learning_rate": 1.2214443147932063e-07,
14606
- "loss": 2.5019,
14607
- "step": 48240
14608
- },
14609
- {
14610
- "epoch": 12.47331696601628,
14611
- "learning_rate": 1.221120186697783e-07,
14612
- "loss": 2.4857,
14613
- "step": 48260
14614
- },
14615
- {
14616
- "epoch": 12.478485592453806,
14617
- "learning_rate": 1.2207960586023596e-07,
14618
- "loss": 2.4496,
14619
- "step": 48280
14620
- },
14621
- {
14622
- "epoch": 12.48365421889133,
14623
- "learning_rate": 1.2204719305069364e-07,
14624
- "loss": 2.5045,
14625
- "step": 48300
14626
- },
14627
- {
14628
- "epoch": 12.488822845328853,
14629
- "learning_rate": 1.2201478024115128e-07,
14630
- "loss": 2.4589,
14631
- "step": 48320
14632
- },
14633
- {
14634
- "epoch": 12.493991471766378,
14635
- "learning_rate": 1.2198236743160897e-07,
14636
- "loss": 2.4778,
14637
- "step": 48340
14638
- },
14639
- {
14640
- "epoch": 12.499160098203902,
14641
- "learning_rate": 1.2194995462206663e-07,
14642
- "loss": 2.4297,
14643
- "step": 48360
14644
- },
14645
- {
14646
- "epoch": 12.504328724641427,
14647
- "learning_rate": 1.219175418125243e-07,
14648
- "loss": 2.395,
14649
- "step": 48380
14650
- },
14651
- {
14652
- "epoch": 12.50949735107895,
14653
- "learning_rate": 1.2188512900298198e-07,
14654
- "loss": 2.4515,
14655
- "step": 48400
14656
- },
14657
- {
14658
- "epoch": 12.514665977516476,
14659
- "learning_rate": 1.2185271619343964e-07,
14660
- "loss": 2.4912,
14661
- "step": 48420
14662
- },
14663
- {
14664
- "epoch": 12.519834603954,
14665
- "learning_rate": 1.218203033838973e-07,
14666
- "loss": 2.4681,
14667
- "step": 48440
14668
- },
14669
- {
14670
- "epoch": 12.525003230391523,
14671
- "learning_rate": 1.21787890574355e-07,
14672
- "loss": 2.4598,
14673
- "step": 48460
14674
- },
14675
- {
14676
- "epoch": 12.530171856829048,
14677
- "learning_rate": 1.2175547776481265e-07,
14678
- "loss": 2.4792,
14679
- "step": 48480
14680
- },
14681
- {
14682
- "epoch": 12.535340483266571,
14683
- "learning_rate": 1.2172306495527032e-07,
14684
- "loss": 2.4331,
14685
- "step": 48500
14686
- },
14687
- {
14688
- "epoch": 12.540509109704097,
14689
- "learning_rate": 1.21690652145728e-07,
14690
- "loss": 2.4537,
14691
- "step": 48520
14692
- },
14693
- {
14694
- "epoch": 12.54567773614162,
14695
- "learning_rate": 1.2165823933618564e-07,
14696
- "loss": 2.5384,
14697
- "step": 48540
14698
- },
14699
- {
14700
- "epoch": 12.550846362579145,
14701
- "learning_rate": 1.2162582652664333e-07,
14702
- "loss": 2.5251,
14703
- "step": 48560
14704
- },
14705
- {
14706
- "epoch": 12.556014989016669,
14707
- "learning_rate": 1.21593413717101e-07,
14708
- "loss": 2.4403,
14709
- "step": 48580
14710
- },
14711
- {
14712
- "epoch": 12.561183615454192,
14713
- "learning_rate": 1.2156100090755865e-07,
14714
- "loss": 2.4652,
14715
- "step": 48600
14716
- },
14717
- {
14718
- "epoch": 12.566352241891718,
14719
- "learning_rate": 1.2152858809801634e-07,
14720
- "loss": 2.5051,
14721
- "step": 48620
14722
- },
14723
- {
14724
- "epoch": 12.571520868329241,
14725
- "learning_rate": 1.21496175288474e-07,
14726
- "loss": 2.4572,
14727
- "step": 48640
14728
- },
14729
- {
14730
- "epoch": 12.576689494766766,
14731
- "learning_rate": 1.2146376247893166e-07,
14732
- "loss": 2.4891,
14733
- "step": 48660
14734
- },
14735
- {
14736
- "epoch": 12.58185812120429,
14737
- "learning_rate": 1.2143134966938935e-07,
14738
- "loss": 2.5186,
14739
- "step": 48680
14740
- },
14741
- {
14742
- "epoch": 12.587026747641815,
14743
- "learning_rate": 1.21398936859847e-07,
14744
- "loss": 2.4751,
14745
- "step": 48700
14746
- },
14747
- {
14748
- "epoch": 12.592195374079338,
14749
- "learning_rate": 1.2136652405030468e-07,
14750
- "loss": 2.4716,
14751
- "step": 48720
14752
- },
14753
- {
14754
- "epoch": 12.597364000516862,
14755
- "learning_rate": 1.2133411124076234e-07,
14756
- "loss": 2.4849,
14757
- "step": 48740
14758
- },
14759
- {
14760
- "epoch": 12.602532626954387,
14761
- "learning_rate": 1.2130169843122e-07,
14762
- "loss": 2.4891,
14763
- "step": 48760
14764
- },
14765
- {
14766
- "epoch": 12.60770125339191,
14767
- "learning_rate": 1.212692856216777e-07,
14768
- "loss": 2.4776,
14769
- "step": 48780
14770
- },
14771
- {
14772
- "epoch": 12.612869879829436,
14773
- "learning_rate": 1.2123687281213535e-07,
14774
- "loss": 2.5003,
14775
- "step": 48800
14776
- },
14777
- {
14778
- "epoch": 12.61803850626696,
14779
- "learning_rate": 1.21204460002593e-07,
14780
- "loss": 2.4541,
14781
- "step": 48820
14782
- },
14783
- {
14784
- "epoch": 12.623207132704485,
14785
- "learning_rate": 1.211720471930507e-07,
14786
- "loss": 2.3995,
14787
- "step": 48840
14788
- },
14789
- {
14790
- "epoch": 12.628375759142008,
14791
- "learning_rate": 1.2113963438350836e-07,
14792
- "loss": 2.4991,
14793
- "step": 48860
14794
- },
14795
- {
14796
- "epoch": 12.633544385579532,
14797
- "learning_rate": 1.2110722157396602e-07,
14798
- "loss": 2.4393,
14799
- "step": 48880
14800
- },
14801
- {
14802
- "epoch": 12.638713012017057,
14803
- "learning_rate": 1.2107480876442369e-07,
14804
- "loss": 2.4266,
14805
- "step": 48900
14806
- },
14807
- {
14808
- "epoch": 12.64388163845458,
14809
- "learning_rate": 1.2104239595488135e-07,
14810
- "loss": 2.4504,
14811
- "step": 48920
14812
- },
14813
- {
14814
- "epoch": 12.649050264892105,
14815
- "learning_rate": 1.2100998314533904e-07,
14816
- "loss": 2.4426,
14817
- "step": 48940
14818
- },
14819
- {
14820
- "epoch": 12.654218891329629,
14821
- "learning_rate": 1.209775703357967e-07,
14822
- "loss": 2.4279,
14823
- "step": 48960
14824
- },
14825
- {
14826
- "epoch": 12.659387517767154,
14827
- "learning_rate": 1.2094515752625436e-07,
14828
- "loss": 2.4973,
14829
- "step": 48980
14830
- },
14831
- {
14832
- "epoch": 12.664556144204678,
14833
- "learning_rate": 1.2091274471671205e-07,
14834
- "loss": 2.4447,
14835
- "step": 49000
14836
- },
14837
- {
14838
- "epoch": 12.669724770642201,
14839
- "learning_rate": 1.208803319071697e-07,
14840
- "loss": 2.497,
14841
- "step": 49020
14842
- },
14843
- {
14844
- "epoch": 12.674893397079726,
14845
- "learning_rate": 1.2084791909762737e-07,
14846
- "loss": 2.4223,
14847
- "step": 49040
14848
- },
14849
- {
14850
- "epoch": 12.68006202351725,
14851
- "learning_rate": 1.2081550628808503e-07,
14852
- "loss": 2.447,
14853
- "step": 49060
14854
- },
14855
- {
14856
- "epoch": 12.685230649954775,
14857
- "learning_rate": 1.2078309347854272e-07,
14858
- "loss": 2.4652,
14859
- "step": 49080
14860
- },
14861
- {
14862
- "epoch": 12.690399276392299,
14863
- "learning_rate": 1.2075068066900038e-07,
14864
- "loss": 2.4167,
14865
- "step": 49100
14866
- },
14867
- {
14868
- "epoch": 12.695567902829822,
14869
- "learning_rate": 1.2071826785945805e-07,
14870
- "loss": 2.4103,
14871
- "step": 49120
14872
- },
14873
- {
14874
- "epoch": 12.700736529267347,
14875
- "learning_rate": 1.206858550499157e-07,
14876
- "loss": 2.4225,
14877
- "step": 49140
14878
- },
14879
- {
14880
- "epoch": 12.70590515570487,
14881
- "learning_rate": 1.206534422403734e-07,
14882
- "loss": 2.4668,
14883
- "step": 49160
14884
- },
14885
- {
14886
- "epoch": 12.711073782142396,
14887
- "learning_rate": 1.2062102943083106e-07,
14888
- "loss": 2.4861,
14889
- "step": 49180
14890
- },
14891
- {
14892
- "epoch": 12.71624240857992,
14893
- "learning_rate": 1.2058861662128872e-07,
14894
- "loss": 2.433,
14895
- "step": 49200
14896
- },
14897
- {
14898
- "epoch": 12.721411035017445,
14899
- "learning_rate": 1.205562038117464e-07,
14900
- "loss": 2.4395,
14901
- "step": 49220
14902
- },
14903
- {
14904
- "epoch": 12.726579661454968,
14905
- "learning_rate": 1.2052379100220407e-07,
14906
- "loss": 2.4702,
14907
- "step": 49240
14908
- },
14909
- {
14910
- "epoch": 12.731748287892493,
14911
- "learning_rate": 1.2049137819266173e-07,
14912
- "loss": 2.4585,
14913
- "step": 49260
14914
- },
14915
- {
14916
- "epoch": 12.736916914330017,
14917
- "learning_rate": 1.204589653831194e-07,
14918
- "loss": 2.414,
14919
- "step": 49280
14920
- },
14921
- {
14922
- "epoch": 12.74208554076754,
14923
- "learning_rate": 1.2042655257357706e-07,
14924
- "loss": 2.4743,
14925
- "step": 49300
14926
- },
14927
- {
14928
- "epoch": 12.747254167205066,
14929
- "learning_rate": 1.2039413976403474e-07,
14930
- "loss": 2.4963,
14931
- "step": 49320
14932
- },
14933
- {
14934
- "epoch": 12.75242279364259,
14935
- "learning_rate": 1.203617269544924e-07,
14936
- "loss": 2.4157,
14937
- "step": 49340
14938
- },
14939
- {
14940
- "epoch": 12.757591420080114,
14941
- "learning_rate": 1.2032931414495007e-07,
14942
- "loss": 2.4355,
14943
- "step": 49360
14944
- },
14945
- {
14946
- "epoch": 12.762760046517638,
14947
- "learning_rate": 1.2029690133540776e-07,
14948
- "loss": 2.4174,
14949
- "step": 49380
14950
- },
14951
- {
14952
- "epoch": 12.767928672955161,
14953
- "learning_rate": 1.2026448852586542e-07,
14954
- "loss": 2.4903,
14955
- "step": 49400
14956
- },
14957
- {
14958
- "epoch": 12.773097299392687,
14959
- "learning_rate": 1.2023207571632308e-07,
14960
- "loss": 2.4285,
14961
- "step": 49420
14962
- },
14963
- {
14964
- "epoch": 12.77826592583021,
14965
- "learning_rate": 1.2019966290678074e-07,
14966
- "loss": 2.5139,
14967
- "step": 49440
14968
- },
14969
- {
14970
- "epoch": 12.783434552267735,
14971
- "learning_rate": 1.2016725009723843e-07,
14972
- "loss": 2.4306,
14973
- "step": 49460
14974
- },
14975
- {
14976
- "epoch": 12.788603178705259,
14977
- "learning_rate": 1.201348372876961e-07,
14978
- "loss": 2.4208,
14979
- "step": 49480
14980
- },
14981
- {
14982
- "epoch": 12.793771805142784,
14983
- "learning_rate": 1.2010242447815375e-07,
14984
- "loss": 2.4154,
14985
- "step": 49500
14986
- },
14987
- {
14988
- "epoch": 12.798940431580307,
14989
- "learning_rate": 1.2007001166861142e-07,
14990
- "loss": 2.4862,
14991
- "step": 49520
14992
- },
14993
- {
14994
- "epoch": 12.804109058017831,
14995
- "learning_rate": 1.200375988590691e-07,
14996
- "loss": 2.4184,
14997
- "step": 49540
14998
- },
14999
- {
15000
- "epoch": 12.809277684455356,
15001
- "learning_rate": 1.2000518604952677e-07,
15002
- "loss": 2.374,
15003
- "step": 49560
15004
- },
15005
- {
15006
- "epoch": 12.81444631089288,
15007
- "learning_rate": 1.1997277323998443e-07,
15008
- "loss": 2.4885,
15009
- "step": 49580
15010
- },
15011
- {
15012
- "epoch": 12.819614937330405,
15013
- "learning_rate": 1.199403604304421e-07,
15014
- "loss": 2.4429,
15015
- "step": 49600
15016
- },
15017
- {
15018
- "epoch": 12.824783563767928,
15019
- "learning_rate": 1.1990794762089978e-07,
15020
- "loss": 2.3834,
15021
- "step": 49620
15022
- },
15023
- {
15024
- "epoch": 12.829952190205454,
15025
- "learning_rate": 1.1987553481135744e-07,
15026
- "loss": 2.415,
15027
- "step": 49640
15028
- },
15029
- {
15030
- "epoch": 12.835120816642977,
15031
- "learning_rate": 1.198431220018151e-07,
15032
- "loss": 2.5206,
15033
- "step": 49660
15034
- },
15035
- {
15036
- "epoch": 12.8402894430805,
15037
- "learning_rate": 1.198107091922728e-07,
15038
- "loss": 2.4883,
15039
- "step": 49680
15040
- },
15041
- {
15042
- "epoch": 12.845458069518026,
15043
- "learning_rate": 1.1977829638273045e-07,
15044
- "loss": 2.4729,
15045
- "step": 49700
15046
- },
15047
- {
15048
- "epoch": 12.85062669595555,
15049
- "learning_rate": 1.1974588357318811e-07,
15050
- "loss": 2.3815,
15051
- "step": 49720
15052
- },
15053
- {
15054
- "epoch": 12.855795322393075,
15055
- "learning_rate": 1.1971347076364578e-07,
15056
- "loss": 2.453,
15057
- "step": 49740
15058
- },
15059
- {
15060
- "epoch": 12.860963948830598,
15061
- "learning_rate": 1.1968105795410346e-07,
15062
- "loss": 2.4414,
15063
- "step": 49760
15064
- },
15065
- {
15066
- "epoch": 12.866132575268123,
15067
- "learning_rate": 1.1964864514456113e-07,
15068
- "loss": 2.5048,
15069
- "step": 49780
15070
- },
15071
- {
15072
- "epoch": 12.871301201705647,
15073
- "learning_rate": 1.196162323350188e-07,
15074
- "loss": 2.4836,
15075
- "step": 49800
15076
- },
15077
- {
15078
- "epoch": 12.87646982814317,
15079
- "learning_rate": 1.1958381952547645e-07,
15080
- "loss": 2.4018,
15081
- "step": 49820
15082
- },
15083
- {
15084
- "epoch": 12.881638454580695,
15085
- "learning_rate": 1.1955140671593414e-07,
15086
- "loss": 2.4562,
15087
- "step": 49840
15088
- },
15089
- {
15090
- "epoch": 12.886807081018219,
15091
- "learning_rate": 1.195189939063918e-07,
15092
- "loss": 2.4431,
15093
- "step": 49860
15094
- },
15095
- {
15096
- "epoch": 12.891975707455744,
15097
- "learning_rate": 1.1948658109684946e-07,
15098
- "loss": 2.4554,
15099
- "step": 49880
15100
- },
15101
- {
15102
- "epoch": 12.897144333893268,
15103
- "learning_rate": 1.1945416828730712e-07,
15104
- "loss": 2.4341,
15105
- "step": 49900
15106
- },
15107
- {
15108
- "epoch": 12.902312960330793,
15109
- "learning_rate": 1.194217554777648e-07,
15110
- "loss": 2.4189,
15111
- "step": 49920
15112
- },
15113
- {
15114
- "epoch": 12.907481586768316,
15115
- "learning_rate": 1.1938934266822247e-07,
15116
- "loss": 2.4739,
15117
- "step": 49940
15118
- },
15119
- {
15120
- "epoch": 12.91265021320584,
15121
- "learning_rate": 1.1935692985868014e-07,
15122
- "loss": 2.3941,
15123
- "step": 49960
15124
- },
15125
- {
15126
- "epoch": 12.917818839643365,
15127
- "learning_rate": 1.1932451704913782e-07,
15128
- "loss": 2.4357,
15129
- "step": 49980
15130
- },
15131
- {
15132
- "epoch": 12.922987466080889,
15133
- "learning_rate": 1.1929210423959549e-07,
15134
- "loss": 2.4676,
15135
- "step": 50000
15136
- },
15137
- {
15138
- "epoch": 12.928156092518414,
15139
- "learning_rate": 1.1925969143005315e-07,
15140
- "loss": 2.4551,
15141
- "step": 50020
15142
- },
15143
- {
15144
- "epoch": 12.933324718955937,
15145
- "learning_rate": 1.192272786205108e-07,
15146
- "loss": 2.4811,
15147
- "step": 50040
15148
- },
15149
- {
15150
- "epoch": 12.938493345393463,
15151
- "learning_rate": 1.191948658109685e-07,
15152
- "loss": 2.4513,
15153
- "step": 50060
15154
- },
15155
- {
15156
- "epoch": 12.943661971830986,
15157
- "learning_rate": 1.1916245300142616e-07,
15158
- "loss": 2.4607,
15159
- "step": 50080
15160
- },
15161
- {
15162
- "epoch": 12.94883059826851,
15163
- "learning_rate": 1.1913004019188382e-07,
15164
- "loss": 2.482,
15165
- "step": 50100
15166
- },
15167
- {
15168
- "epoch": 12.953999224706035,
15169
- "learning_rate": 1.1909762738234151e-07,
15170
- "loss": 2.4465,
15171
- "step": 50120
15172
- },
15173
- {
15174
- "epoch": 12.959167851143558,
15175
- "learning_rate": 1.1906521457279917e-07,
15176
- "loss": 2.4672,
15177
- "step": 50140
15178
- },
15179
- {
15180
- "epoch": 12.964336477581083,
15181
- "learning_rate": 1.1903280176325683e-07,
15182
- "loss": 2.421,
15183
- "step": 50160
15184
- },
15185
- {
15186
- "epoch": 12.969505104018607,
15187
- "learning_rate": 1.190003889537145e-07,
15188
- "loss": 2.4271,
15189
- "step": 50180
15190
- },
15191
- {
15192
- "epoch": 12.974673730456132,
15193
- "learning_rate": 1.1896797614417218e-07,
15194
- "loss": 2.4226,
15195
- "step": 50200
15196
- },
15197
- {
15198
- "epoch": 12.979842356893656,
15199
- "learning_rate": 1.1893556333462985e-07,
15200
- "loss": 2.4162,
15201
- "step": 50220
15202
- },
15203
- {
15204
- "epoch": 12.985010983331179,
15205
- "learning_rate": 1.1890315052508751e-07,
15206
- "loss": 2.509,
15207
- "step": 50240
15208
- },
15209
- {
15210
- "epoch": 12.990179609768704,
15211
- "learning_rate": 1.1887073771554518e-07,
15212
- "loss": 2.4553,
15213
- "step": 50260
15214
- },
15215
- {
15216
- "epoch": 12.995348236206228,
15217
- "learning_rate": 1.1883832490600286e-07,
15218
- "loss": 2.4558,
15219
- "step": 50280
15220
- },
15221
- {
15222
- "epoch": 13.0,
15223
- "eval_bleu": 12.8144,
15224
- "eval_gen_len": 37.2979,
15225
- "eval_loss": 2.433004140853882,
15226
- "eval_runtime": 806.1399,
15227
- "eval_samples_per_second": 2.136,
15228
- "eval_steps_per_second": 1.068,
15229
- "step": 50298
15230
  }
15231
  ],
15232
  "logging_steps": 20,
@@ -15246,7 +12910,7 @@
15246
  "attributes": {}
15247
  }
15248
  },
15249
- "total_flos": 3.301586329527091e+17,
15250
  "train_batch_size": 2,
15251
  "trial_name": null,
15252
  "trial_params": null
 
1
  {
2
+ "best_metric": 12.1716,
3
+ "best_model_checkpoint": "/kaggle/tmp/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted/checkpoint-42562",
4
+ "epoch": 11.0,
5
  "eval_steps": 500,
6
+ "global_step": 42562,
7
  "is_hyper_param_search": false,
8
  "is_local_process_zero": true,
9
  "is_world_process_zero": true,
 
12891
  "eval_samples_per_second": 2.165,
12892
  "eval_steps_per_second": 1.083,
12893
  "step": 42562
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
12894
  }
12895
  ],
12896
  "logging_steps": 20,
 
12910
  "attributes": {}
12911
  }
12912
  },
12913
+ "total_flos": 2.793935715793797e+17,
12914
  "train_batch_size": 2,
12915
  "trial_name": null,
12916
  "trial_params": null
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5f63231d876ab3ad240434253a3c1d9b5e552e066e216bb8da212e99bc37119e
3
  size 1575259780
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b02f536083a5028b848f016fb701948c3bb3daae9d3cbf2cf6bec6fb93e75a3
3
  size 1575259780
run-2024-10-28T15:22:08+00:00.log CHANGED
The diff for this file is too large to render. See raw diff
 
val_outputs/val_generated_predictions_95652971.txt ADDED
The diff for this file is too large to render. See raw diff