diff --git "a/log/debug_0.log" "b/log/debug_0.log"
--- "a/log/debug_0.log"
+++ "b/log/debug_0.log"
@@ -46578,3 +46578,1009 @@ Use FP16 precision: False
 02/26/2022 13:08:55 - INFO - codeparrot_training - Step 45998: {'lr': 8.52701533001332e-06, 'samples': 23551488, 'steps': 45998, 'loss/train': 1.7301279306411743}
 02/26/2022 13:09:00 - INFO - codeparrot_training - Step 45999: {'lr': 8.522778861667273e-06, 'samples': 23552000, 'steps': 45999, 'loss/train': 1.087491512298584}
 02/26/2022 13:09:00 - INFO - codeparrot_training - Evaluating and saving model checkpoint
+02/26/2022 13:09:18 - WARNING - huggingface_hub.repository - Several commits (46) will be pushed upstream.
+02/26/2022 13:09:18 - WARNING - huggingface_hub.repository - The progress bars may be unreliable.
+02/26/2022 13:09:53 - WARNING - huggingface_hub.repository - To https://huggingface.co/ncoop57/multi-code-clippy
+   9a19140..572893f  floral-grass-11 -> floral-grass-11
+
+02/26/2022 13:09:58 - INFO - codeparrot_training - Step 46000: {'lr': 8.51854342773295e-06, 'samples': 23552512, 'steps': 46000, 'loss/train': 1.445226788520813}
+02/26/2022 13:10:03 - INFO - codeparrot_training - Step 46001: {'lr': 8.514309028228473e-06, 'samples': 23553024, 'steps': 46001, 'loss/train': 2.392515182495117}
+02/26/2022 13:10:07 - INFO - codeparrot_training - Step 46002: {'lr': 8.51007566317205e-06, 'samples': 23553536, 'steps': 46002, 'loss/train': 1.8956668376922607}
+02/26/2022 13:10:12 - INFO - codeparrot_training - Step 46003: {'lr': 8.505843332581781e-06, 'samples': 23554048, 'steps': 46003, 'loss/train': 1.446045160293579}
+02/26/2022 13:10:16 - INFO - codeparrot_training - Step 46004: {'lr': 8.501612036475815e-06, 'samples': 23554560, 'steps': 46004, 'loss/train': 0.8679484128952026}
+02/26/2022 13:10:21 - INFO - codeparrot_training - Step 46005: {'lr': 8.49738177487222e-06, 'samples': 23555072, 'steps': 46005, 'loss/train': 1.8368754386901855}
+02/26/2022 13:10:25 - INFO - codeparrot_training - Step 46006: {'lr': 8.493152547789152e-06, 'samples': 23555584, 'steps': 46006, 'loss/train': 1.892811894416809}
+02/26/2022 13:10:31 - INFO - codeparrot_training - Step 46007: {'lr': 8.48892435524476e-06, 'samples': 23556096, 'steps': 46007, 'loss/train': 1.4583792686462402}
+02/26/2022 13:10:34 - INFO - codeparrot_training - Step 46008: {'lr': 8.484697197257142e-06, 'samples': 23556608, 'steps': 46008, 'loss/train': 0.33918002247810364}
+02/26/2022 13:10:40 - INFO - codeparrot_training - Step 46009: {'lr': 8.480471073844365e-06, 'samples': 23557120, 'steps': 46009, 'loss/train': 1.828680157661438}
+02/26/2022 13:10:43 - INFO - codeparrot_training - Step 46010: {'lr': 8.476245985024528e-06, 'samples': 23557632, 'steps': 46010, 'loss/train': 2.216860055923462}
+02/26/2022 13:10:50 - INFO - codeparrot_training - Step 46011: {'lr': 8.47202193081581e-06, 'samples': 23558144, 'steps': 46011, 'loss/train': 1.7501882314682007}
+02/26/2022 13:10:54 - INFO - codeparrot_training - Step 46012: {'lr': 8.467798911236224e-06, 'samples': 23558656, 'steps': 46012, 'loss/train': 1.0804424285888672}
+02/26/2022 13:10:59 - INFO - codeparrot_training - Step 46013: {'lr': 8.463576926303922e-06, 'samples': 23559168, 'steps': 46013, 'loss/train': 0.7688332200050354}
+02/26/2022 13:11:03 - INFO - codeparrot_training - Step 46014: {'lr': 8.459355976036947e-06, 'samples': 23559680, 'steps': 46014, 'loss/train': 0.9228298664093018}
+02/26/2022 13:11:08 - INFO - codeparrot_training - Step 46015: {'lr': 8.455136060453395e-06, 'samples': 23560192, 'steps': 46015, 'loss/train': 2.0492424964904785}
+02/26/2022 13:11:12 - INFO - codeparrot_training - Step 46016: {'lr': 8.450917179571306e-06, 'samples': 23560704, 'steps': 46016, 'loss/train': 0.7214325070381165}
+02/26/2022 13:11:17 - INFO - codeparrot_training - Step 46017: {'lr': 8.446699333408859e-06, 'samples': 23561216, 'steps': 46017, 'loss/train': 2.676889657974243}
+02/26/2022 13:11:21 - INFO - codeparrot_training - Step 46018: {'lr': 8.442482521983986e-06, 'samples': 23561728, 'steps': 46018, 'loss/train': 1.5434142351150513}
+02/26/2022 13:11:27 - INFO - codeparrot_training - Step 46019: {'lr': 8.438266745314838e-06, 'samples': 23562240, 'steps': 46019, 'loss/train': 1.4020642042160034}
+02/26/2022 13:11:30 - INFO - codeparrot_training - Step 46020: {'lr': 8.434052003419457e-06, 'samples': 23562752, 'steps': 46020, 'loss/train': 1.0194350481033325}
+02/26/2022 13:11:36 - INFO - codeparrot_training - Step 46021: {'lr': 8.429838296315883e-06, 'samples': 23563264, 'steps': 46021, 'loss/train': 1.6184508800506592}
+02/26/2022 13:11:40 - INFO - codeparrot_training - Step 46022: {'lr': 8.425625624022187e-06, 'samples': 23563776, 'steps': 46022, 'loss/train': 1.6239887475967407}
+02/26/2022 13:11:45 - INFO - codeparrot_training - Step 46023: {'lr': 8.421413986556408e-06, 'samples': 23564288, 'steps': 46023, 'loss/train': 0.8545622229576111}
+02/26/2022 13:11:49 - INFO - codeparrot_training - Step 46024: {'lr': 8.417203383936534e-06, 'samples': 23564800, 'steps': 46024, 'loss/train': 1.424389362335205}
+02/26/2022 13:11:54 - INFO - codeparrot_training - Step 46025: {'lr': 8.412993816180687e-06, 'samples': 23565312, 'steps': 46025, 'loss/train': 2.0163636207580566}
+02/26/2022 13:11:58 - INFO - codeparrot_training - Step 46026: {'lr': 8.40878528330688e-06, 'samples': 23565824, 'steps': 46026, 'loss/train': 1.8338297605514526}
+02/26/2022 13:12:03 - INFO - codeparrot_training - Step 46027: {'lr': 8.4045777853331e-06, 'samples': 23566336, 'steps': 46027, 'loss/train': 2.553800582885742}
+02/26/2022 13:12:07 - INFO - codeparrot_training - Step 46028: {'lr': 8.400371322277362e-06, 'samples': 23566848, 'steps': 46028, 'loss/train': 1.0743021965026855}
+02/26/2022 13:12:12 - INFO - codeparrot_training - Step 46029: {'lr': 8.39616589415776e-06, 'samples': 23567360, 'steps': 46029, 'loss/train': 1.1663625240325928}
+02/26/2022 13:12:16 - INFO - codeparrot_training - Step 46030: {'lr': 8.391961500992252e-06, 'samples': 23567872, 'steps': 46030, 'loss/train': 1.4144514799118042}
+02/26/2022 13:12:21 - INFO - codeparrot_training - Step 46031: {'lr': 8.387758142798879e-06, 'samples': 23568384, 'steps': 46031, 'loss/train': 2.026470184326172}
+02/26/2022 13:12:25 - INFO - codeparrot_training - Step 46032: {'lr': 8.383555819595601e-06, 'samples': 23568896, 'steps': 46032, 'loss/train': 1.2396329641342163}
+02/26/2022 13:12:31 - INFO - codeparrot_training - Step 46033: {'lr': 8.379354531400457e-06, 'samples': 23569408, 'steps': 46033, 'loss/train': 2.0409650802612305}
+02/26/2022 13:12:34 - INFO - codeparrot_training - Step 46034: {'lr': 8.375154278231434e-06, 'samples': 23569920, 'steps': 46034, 'loss/train': 1.0311235189437866}
+02/26/2022 13:12:40 - INFO - codeparrot_training - Step 46035: {'lr': 8.370955060106573e-06, 'samples': 23570432, 'steps': 46035, 'loss/train': 1.600880742073059}
+02/26/2022 13:12:43 - INFO - codeparrot_training - Step 46036: {'lr': 8.366756877043747e-06, 'samples': 23570944, 'steps': 46036, 'loss/train': 1.2611210346221924}
+02/26/2022 13:12:49 - INFO - codeparrot_training - Step 46037: {'lr': 8.362559729061025e-06, 'samples': 23571456, 'steps': 46037, 'loss/train': 1.92186439037323}
+02/26/2022 13:12:52 - INFO - codeparrot_training - Step 46038: {'lr': 8.358363616176396e-06, 'samples': 23571968, 'steps': 46038, 'loss/train': 1.3052363395690918}
+02/26/2022 13:12:58 - INFO - codeparrot_training - Step 46039: {'lr': 8.354168538407787e-06, 'samples': 23572480, 'steps': 46039, 'loss/train': 1.0481584072113037}
+02/26/2022 13:13:01 - INFO - codeparrot_training - Step 46040: {'lr': 8.349974495773182e-06, 'samples': 23572992, 'steps': 46040, 'loss/train': 1.6523873805999756}
+02/26/2022 13:13:07 - INFO - codeparrot_training - Step 46041: {'lr': 8.34578148829057e-06, 'samples': 23573504, 'steps': 46041, 'loss/train': 2.1606557369232178}
+02/26/2022 13:13:10 - INFO - codeparrot_training - Step 46042: {'lr': 8.341589515977882e-06, 'samples': 23574016, 'steps': 46042, 'loss/train': 1.7149804830551147}
+02/26/2022 13:13:16 - INFO - codeparrot_training - Step 46043: {'lr': 8.3373985788531e-06, 'samples': 23574528, 'steps': 46043, 'loss/train': 2.0021350383758545}
+02/26/2022 13:13:20 - INFO - codeparrot_training - Step 46044: {'lr': 8.333208676934156e-06, 'samples': 23575040, 'steps': 46044, 'loss/train': 1.8283710479736328}
+02/26/2022 13:13:25 - INFO - codeparrot_training - Step 46045: {'lr': 8.329019810239008e-06, 'samples': 23575552, 'steps': 46045, 'loss/train': 1.4947823286056519}
+02/26/2022 13:13:29 - INFO - codeparrot_training - Step 46046: {'lr': 8.324831978785614e-06, 'samples': 23576064, 'steps': 46046, 'loss/train': 4.915385723114014}
+02/26/2022 13:13:34 - INFO - codeparrot_training - Step 46047: {'lr': 8.320645182591875e-06, 'samples': 23576576, 'steps': 46047, 'loss/train': 0.7544673681259155}
+02/26/2022 13:13:38 - INFO - codeparrot_training - Step 46048: {'lr': 8.316459421675776e-06, 'samples': 23577088, 'steps': 46048, 'loss/train': 1.2011456489562988}
+02/26/2022 13:13:43 - INFO - codeparrot_training - Step 46049: {'lr': 8.312274696055195e-06, 'samples': 23577600, 'steps': 46049, 'loss/train': 1.645440697669983}
+02/26/2022 13:13:47 - INFO - codeparrot_training - Step 46050: {'lr': 8.308091005748114e-06, 'samples': 23578112, 'steps': 46050, 'loss/train': 1.871660828590393}
+02/26/2022 13:13:52 - INFO - codeparrot_training - Step 46051: {'lr': 8.30390835077241e-06, 'samples': 23578624, 'steps': 46051, 'loss/train': 2.5569796562194824}
+02/26/2022 13:13:56 - INFO - codeparrot_training - Step 46052: {'lr': 8.29972673114604e-06, 'samples': 23579136, 'steps': 46052, 'loss/train': 1.9057695865631104}
+02/26/2022 13:14:02 - INFO - codeparrot_training - Step 46053: {'lr': 8.295546146886851e-06, 'samples': 23579648, 'steps': 46053, 'loss/train': 1.3864293098449707}
+02/26/2022 13:14:05 - INFO - codeparrot_training - Step 46054: {'lr': 8.291366598012828e-06, 'samples': 23580160, 'steps': 46054, 'loss/train': 1.3967887163162231}
+02/26/2022 13:14:11 - INFO - codeparrot_training - Step 46055: {'lr': 8.28718808454179e-06, 'samples': 23580672, 'steps': 46055, 'loss/train': 2.090911388397217}
+02/26/2022 13:14:14 - INFO - codeparrot_training - Step 46056: {'lr': 8.283010606491752e-06, 'samples': 23581184, 'steps': 46056, 'loss/train': 1.7933493852615356}
+02/26/2022 13:14:21 - INFO - codeparrot_training - Step 46057: {'lr': 8.278834163880505e-06, 'samples': 23581696, 'steps': 46057, 'loss/train': 2.135993719100952}
+02/26/2022 13:14:24 - INFO - codeparrot_training - Step 46058: {'lr': 8.274658756725978e-06, 'samples': 23582208, 'steps': 46058, 'loss/train': 1.618568778038025}
+02/26/2022 13:14:30 - INFO - codeparrot_training - Step 46059: {'lr': 8.270484385046046e-06, 'samples': 23582720, 'steps': 46059, 'loss/train': 1.7424323558807373}
+02/26/2022 13:14:33 - INFO - codeparrot_training - Step 46060: {'lr': 8.26631104885861e-06, 'samples': 23583232, 'steps': 46060, 'loss/train': 2.4552671909332275}
+02/26/2022 13:14:39 - INFO - codeparrot_training - Step 46061: {'lr': 8.262138748181547e-06, 'samples': 23583744, 'steps': 46061, 'loss/train': 1.168294906616211}
+02/26/2022 13:14:42 - INFO - codeparrot_training - Step 46062: {'lr': 8.25796748303273e-06, 'samples': 23584256, 'steps': 46062, 'loss/train': 1.5917696952819824}
+02/26/2022 13:14:48 - INFO - codeparrot_training - Step 46063: {'lr': 8.253797253429978e-06, 'samples': 23584768, 'steps': 46063, 'loss/train': 2.327592134475708}
+02/26/2022 13:14:51 - INFO - codeparrot_training - Step 46064: {'lr': 8.249628059391251e-06, 'samples': 23585280, 'steps': 46064, 'loss/train': 0.6519719958305359}
+02/26/2022 13:14:57 - INFO - codeparrot_training - Step 46065: {'lr': 8.24545990093431e-06, 'samples': 23585792, 'steps': 46065, 'loss/train': 1.8860714435577393}
+02/26/2022 13:15:01 - INFO - codeparrot_training - Step 46066: {'lr': 8.241292778077086e-06, 'samples': 23586304, 'steps': 46066, 'loss/train': 2.0867278575897217}
+02/26/2022 13:15:06 - INFO - codeparrot_training - Step 46067: {'lr': 8.23712669083737e-06, 'samples': 23586816, 'steps': 46067, 'loss/train': 0.8177598118782043}
+02/26/2022 13:15:10 - INFO - codeparrot_training - Step 46068: {'lr': 8.232961639233038e-06, 'samples': 23587328, 'steps': 46068, 'loss/train': 2.596421718597412}
+02/26/2022 13:15:16 - INFO - codeparrot_training - Step 46069: {'lr': 8.228797623281908e-06, 'samples': 23587840, 'steps': 46069, 'loss/train': 3.172916889190674}
+02/26/2022 13:15:20 - INFO - codeparrot_training - Step 46070: {'lr': 8.22463464300191e-06, 'samples': 23588352, 'steps': 46070, 'loss/train': 1.8660672903060913}
+02/26/2022 13:15:25 - INFO - codeparrot_training - Step 46071: {'lr': 8.220472698410753e-06, 'samples': 23588864, 'steps': 46071, 'loss/train': 1.1300395727157593}
+02/26/2022 13:15:29 - INFO - codeparrot_training - Step 46072: {'lr': 8.21631178952631e-06, 'samples': 23589376, 'steps': 46072, 'loss/train': 3.1932754516601562}
+02/26/2022 13:15:35 - INFO - codeparrot_training - Step 46073: {'lr': 8.212151916366428e-06, 'samples': 23589888, 'steps': 46073, 'loss/train': 2.163679599761963}
+02/26/2022 13:15:38 - INFO - codeparrot_training - Step 46074: {'lr': 8.207993078948928e-06, 'samples': 23590400, 'steps': 46074, 'loss/train': 1.4854207038879395}
+02/26/2022 13:15:44 - INFO - codeparrot_training - Step 46075: {'lr': 8.203835277291572e-06, 'samples': 23590912, 'steps': 46075, 'loss/train': 0.3289591073989868}
+02/26/2022 13:15:47 - INFO - codeparrot_training - Step 46076: {'lr': 8.199678511412234e-06, 'samples': 23591424, 'steps': 46076, 'loss/train': 1.9764361381530762}
+02/26/2022 13:15:53 - INFO - codeparrot_training - Step 46077: {'lr': 8.195522781328652e-06, 'samples': 23591936, 'steps': 46077, 'loss/train': 1.4759719371795654}
+02/26/2022 13:15:56 - INFO - codeparrot_training - Step 46078: {'lr': 8.191368087058754e-06, 'samples': 23592448, 'steps': 46078, 'loss/train': 1.5361700057983398}
+02/26/2022 13:16:02 - INFO - codeparrot_training - Step 46079: {'lr': 8.187214428620166e-06, 'samples': 23592960, 'steps': 46079, 'loss/train': 0.8220481276512146}
+02/26/2022 13:16:06 - INFO - codeparrot_training - Step 46080: {'lr': 8.183061806030818e-06, 'samples': 23593472, 'steps': 46080, 'loss/train': 3.4895219802856445}
+02/26/2022 13:16:11 - INFO - codeparrot_training - Step 46081: {'lr': 8.178910219308388e-06, 'samples': 23593984, 'steps': 46081, 'loss/train': 0.33559557795524597}
+02/26/2022 13:16:15 - INFO - codeparrot_training - Step 46082: {'lr': 8.174759668470783e-06, 'samples': 23594496, 'steps': 46082, 'loss/train': 1.1306363344192505}
+02/26/2022 13:16:20 - INFO - codeparrot_training - Step 46083: {'lr': 8.170610153535708e-06, 'samples': 23595008, 'steps': 46083, 'loss/train': 0.44970691204071045}
+02/26/2022 13:16:24 - INFO - codeparrot_training - Step 46084: {'lr': 8.166461674520953e-06, 'samples': 23595520, 'steps': 46084, 'loss/train': 0.8695681691169739}
+02/26/2022 13:16:29 - INFO - codeparrot_training - Step 46085: {'lr': 8.162314231444256e-06, 'samples': 23596032, 'steps': 46085, 'loss/train': 3.0058653354644775}
+02/26/2022 13:16:33 - INFO - codeparrot_training - Step 46086: {'lr': 8.158167824323408e-06, 'samples': 23596544, 'steps': 46086, 'loss/train': 1.5750559568405151}
+02/26/2022 13:16:39 - INFO - codeparrot_training - Step 46087: {'lr': 8.154022453176202e-06, 'samples': 23597056, 'steps': 46087, 'loss/train': 1.1741547584533691}
+02/26/2022 13:16:42 - INFO - codeparrot_training - Step 46088: {'lr': 8.149878118020371e-06, 'samples': 23597568, 'steps': 46088, 'loss/train': 2.2009966373443604}
+02/26/2022 13:16:48 - INFO - codeparrot_training - Step 46089: {'lr': 8.14573481887368e-06, 'samples': 23598080, 'steps': 46089, 'loss/train': 1.8772943019866943}
+02/26/2022 13:16:52 - INFO - codeparrot_training - Step 46090: {'lr': 8.141592555753807e-06, 'samples': 23598592, 'steps': 46090, 'loss/train': 2.1989333629608154}
+02/26/2022 13:16:57 - INFO - codeparrot_training - Step 46091: {'lr': 8.137451328678603e-06, 'samples': 23599104, 'steps': 46091, 'loss/train': 0.9876439571380615}
+02/26/2022 13:17:01 - INFO - codeparrot_training - Step 46092: {'lr': 8.133311137665745e-06, 'samples': 23599616, 'steps': 46092, 'loss/train': 0.758363664150238}
+02/26/2022 13:17:06 - INFO - codeparrot_training - Step 46093: {'lr': 8.129171982732996e-06, 'samples': 23600128, 'steps': 46093, 'loss/train': 0.9829350113868713}
+02/26/2022 13:17:10 - INFO - codeparrot_training - Step 46094: {'lr': 8.125033863898013e-06, 'samples': 23600640, 'steps': 46094, 'loss/train': 1.7901220321655273}
+02/26/2022 13:17:15 - INFO - codeparrot_training - Step 46095: {'lr': 8.120896781178639e-06, 'samples': 23601152, 'steps': 46095, 'loss/train': 2.7484991550445557}
+02/26/2022 13:17:19 - INFO - codeparrot_training - Step 46096: {'lr': 8.116760734592526e-06, 'samples': 23601664, 'steps': 46096, 'loss/train': 1.8779226541519165}
+02/26/2022 13:17:24 - INFO - codeparrot_training - Step 46097: {'lr': 8.112625724157412e-06, 'samples': 23602176, 'steps': 46097, 'loss/train': 2.378857374191284}
+02/26/2022 13:17:28 - INFO - codeparrot_training - Step 46098: {'lr': 8.10849174989095e-06, 'samples': 23602688, 'steps': 46098, 'loss/train': 1.6186782121658325}
+02/26/2022 13:17:33 - INFO - codeparrot_training - Step 46099: {'lr': 8.104358811810957e-06, 'samples': 23603200, 'steps': 46099, 'loss/train': 2.0097146034240723}
+02/26/2022 13:17:36 - INFO - codeparrot_training - Step 46100: {'lr': 8.10022690993506e-06, 'samples': 23603712, 'steps': 46100, 'loss/train': 1.5551592111587524}
+02/26/2022 13:17:42 - INFO - codeparrot_training - Step 46101: {'lr': 8.096096044280993e-06, 'samples': 23604224, 'steps': 46101, 'loss/train': 2.4417436122894287}
+02/26/2022 13:17:45 - INFO - codeparrot_training - Step 46102: {'lr': 8.091966214866409e-06, 'samples': 23604736, 'steps': 46102, 'loss/train': 1.0414965152740479}
+02/26/2022 13:17:51 - INFO - codeparrot_training - Step 46103: {'lr': 8.087837421709043e-06, 'samples': 23605248, 'steps': 46103, 'loss/train': 2.6877174377441406}
+02/26/2022 13:17:54 - INFO - codeparrot_training - Step 46104: {'lr': 8.083709664826577e-06, 'samples': 23605760, 'steps': 46104, 'loss/train': 2.2320430278778076}
+02/26/2022 13:18:01 - INFO - codeparrot_training - Step 46105: {'lr': 8.079582944236663e-06, 'samples': 23606272, 'steps': 46105, 'loss/train': 2.077634811401367}
+02/26/2022 13:18:04 - INFO - codeparrot_training - Step 46106: {'lr': 8.075457259956981e-06, 'samples': 23606784, 'steps': 46106, 'loss/train': 2.1648178100585938}
+02/26/2022 13:18:10 - INFO - codeparrot_training - Step 46107: {'lr': 8.071332612005266e-06, 'samples': 23607296, 'steps': 46107, 'loss/train': 1.6990742683410645}
+02/26/2022 13:18:13 - INFO - codeparrot_training - Step 46108: {'lr': 8.067209000399089e-06, 'samples': 23607808, 'steps': 46108, 'loss/train': 0.46222829818725586}
+02/26/2022 13:18:18 - INFO - codeparrot_training - Step 46109: {'lr': 8.06308642515624e-06, 'samples': 23608320, 'steps': 46109, 'loss/train': 0.5889194011688232}
+02/26/2022 13:18:22 - INFO - codeparrot_training - Step 46110: {'lr': 8.058964886294234e-06, 'samples': 23608832, 'steps': 46110, 'loss/train': 1.3687952756881714}
+02/26/2022 13:18:27 - INFO - codeparrot_training - Step 46111: {'lr': 8.054844383830834e-06, 'samples': 23609344, 'steps': 46111, 'loss/train': 1.207642674446106}
+02/26/2022 13:18:33 - INFO - codeparrot_training - Step 46112: {'lr': 8.050724917783635e-06, 'samples': 23609856, 'steps': 46112, 'loss/train': 1.4088826179504395}
+02/26/2022 13:18:36 - INFO - codeparrot_training - Step 46113: {'lr': 8.04660648817035e-06, 'samples': 23610368, 'steps': 46113, 'loss/train': 0.6511171460151672}
+02/26/2022 13:18:43 - INFO - codeparrot_training - Step 46114: {'lr': 8.042489095008516e-06, 'samples': 23610880, 'steps': 46114, 'loss/train': 0.82984459400177}
+02/26/2022 13:18:46 - INFO - codeparrot_training - Step 46115: {'lr': 8.03837273831584e-06, 'samples': 23611392, 'steps': 46115, 'loss/train': 1.4841128587722778}
+02/26/2022 13:18:52 - INFO - codeparrot_training - Step 46116: {'lr': 8.034257418109924e-06, 'samples': 23611904, 'steps': 46116, 'loss/train': 2.2533814907073975}
+02/26/2022 13:18:55 - INFO - codeparrot_training - Step 46117: {'lr': 8.030143134408474e-06, 'samples': 23612416, 'steps': 46117, 'loss/train': 2.6550283432006836}
+02/26/2022 13:18:59 - INFO - codeparrot_training - Step 46118: {'lr': 8.02602988722903e-06, 'samples': 23612928, 'steps': 46118, 'loss/train': 1.1456936597824097}
+02/26/2022 13:19:04 - INFO - codeparrot_training - Step 46119: {'lr': 8.021917676589247e-06, 'samples': 23613440, 'steps': 46119, 'loss/train': 3.120382308959961}
+02/26/2022 13:19:08 - INFO - codeparrot_training - Step 46120: {'lr': 8.017806502506692e-06, 'samples': 23613952, 'steps': 46120, 'loss/train': 1.411298155784607}
+02/26/2022 13:19:13 - INFO - codeparrot_training - Step 46121: {'lr': 8.013696364999047e-06, 'samples': 23614464, 'steps': 46121, 'loss/train': 1.811667561531067}
+02/26/2022 13:19:19 - INFO - codeparrot_training - Step 46122: {'lr': 8.009587264083907e-06, 'samples': 23614976, 'steps': 46122, 'loss/train': 0.86668860912323}
+02/26/2022 13:19:22 - INFO - codeparrot_training - Step 46123: {'lr': 8.005479199778842e-06, 'samples': 23615488, 'steps': 46123, 'loss/train': 1.4036000967025757}
+02/26/2022 13:19:29 - INFO - codeparrot_training - Step 46124: {'lr': 8.001372172101422e-06, 'samples': 23616000, 'steps': 46124, 'loss/train': 1.817915678024292}
+02/26/2022 13:19:32 - INFO - codeparrot_training - Step 46125: {'lr': 7.99726618106933e-06, 'samples': 23616512, 'steps': 46125, 'loss/train': 1.6431174278259277}
+02/26/2022 13:19:38 - INFO - codeparrot_training - Step 46126: {'lr': 7.993161226700074e-06, 'samples': 23617024, 'steps': 46126, 'loss/train': 1.3536540269851685}
+02/26/2022 13:19:41 - INFO - codeparrot_training - Step 46127: {'lr': 7.98905730901131e-06, 'samples': 23617536, 'steps': 46127, 'loss/train': 2.411864757537842}
+02/26/2022 13:19:47 - INFO - codeparrot_training - Step 46128: {'lr': 7.984954428020524e-06, 'samples': 23618048, 'steps': 46128, 'loss/train': 1.6283249855041504}
+02/26/2022 13:19:50 - INFO - codeparrot_training - Step 46129: {'lr': 7.980852583745368e-06, 'samples': 23618560, 'steps': 46129, 'loss/train': 2.173253059387207}
+02/26/2022 13:19:56 - INFO - codeparrot_training - Step 46130: {'lr': 7.97675177620341e-06, 'samples': 23619072, 'steps': 46130, 'loss/train': 0.7825809717178345}
+02/26/2022 13:19:59 - INFO - codeparrot_training - Step 46131: {'lr': 7.972652005412196e-06, 'samples': 23619584, 'steps': 46131, 'loss/train': 1.8228380680084229}
+02/26/2022 13:20:05 - INFO - codeparrot_training - Step 46132: {'lr': 7.968553271389261e-06, 'samples': 23620096, 'steps': 46132, 'loss/train': 1.6632410287857056}
+02/26/2022 13:20:08 - INFO - codeparrot_training - Step 46133: {'lr': 7.96445557415218e-06, 'samples': 23620608, 'steps': 46133, 'loss/train': 2.0599172115325928}
+02/26/2022 13:20:15 - INFO - codeparrot_training - Step 46134: {'lr': 7.960358913718546e-06, 'samples': 23621120, 'steps': 46134, 'loss/train': 1.736124873161316}
+02/26/2022 13:20:18 - INFO - codeparrot_training - Step 46135: {'lr': 7.956263290105875e-06, 'samples': 23621632, 'steps': 46135, 'loss/train': 2.062497854232788}
+02/26/2022 13:20:24 - INFO - codeparrot_training - Step 46136: {'lr': 7.952168703331708e-06, 'samples': 23622144, 'steps': 46136, 'loss/train': 1.4588398933410645}
+02/26/2022 13:20:28 - INFO - codeparrot_training - Step 46137: {'lr': 7.948075153413587e-06, 'samples': 23622656, 'steps': 46137, 'loss/train': 1.1146643161773682}
+02/26/2022 13:20:33 - INFO - codeparrot_training - Step 46138: {'lr': 7.943982640369052e-06, 'samples': 23623168, 'steps': 46138, 'loss/train': 1.966606616973877}
+02/26/2022 13:20:37 - INFO - codeparrot_training - Step 46139: {'lr': 7.939891164215645e-06, 'samples': 23623680, 'steps': 46139, 'loss/train': 1.8707178831100464}
+02/26/2022 13:20:42 - INFO - codeparrot_training - Step 46140: {'lr': 7.935800724970883e-06, 'samples': 23624192, 'steps': 46140, 'loss/train': 1.5159659385681152}
+02/26/2022 13:20:46 - INFO - codeparrot_training - Step 46141: {'lr': 7.931711322652274e-06, 'samples': 23624704, 'steps': 46141, 'loss/train': 0.6920977234840393}
+02/26/2022 13:20:51 - INFO - codeparrot_training - Step 46142: {'lr': 7.927622957277364e-06, 'samples': 23625216, 'steps': 46142, 'loss/train': 1.78933584690094}
+02/26/2022 13:20:54 - INFO - codeparrot_training - Step 46143: {'lr': 7.92353562886361e-06, 'samples': 23625728, 'steps': 46143, 'loss/train': 2.1544649600982666}
+02/26/2022 13:21:00 - INFO - codeparrot_training - Step 46144: {'lr': 7.919449337428636e-06, 'samples': 23626240, 'steps': 46144, 'loss/train': 1.3264826536178589}
+02/26/2022 13:21:03 - INFO - codeparrot_training - Step 46145: {'lr': 7.91536408298979e-06, 'samples': 23626752, 'steps': 46145, 'loss/train': 0.9895395040512085}
+02/26/2022 13:21:09 - INFO - codeparrot_training - Step 46146: {'lr': 7.911279865564697e-06, 'samples': 23627264, 'steps': 46146, 'loss/train': 1.5277233123779297}
+02/26/2022 13:21:12 - INFO - codeparrot_training - Step 46147: {'lr': 7.907196685170787e-06, 'samples': 23627776, 'steps': 46147, 'loss/train': 1.7551686763763428}
+02/26/2022 13:21:18 - INFO - codeparrot_training - Step 46148: {'lr': 7.903114541825628e-06, 'samples': 23628288, 'steps': 46148, 'loss/train': 1.9766554832458496}
+02/26/2022 13:21:22 - INFO - codeparrot_training - Step 46149: {'lr': 7.899033435546598e-06, 'samples': 23628800, 'steps': 46149, 'loss/train': 1.8132158517837524}
+02/26/2022 13:21:28 - INFO - codeparrot_training - Step 46150: {'lr': 7.894953366351265e-06, 'samples': 23629312, 'steps': 46150, 'loss/train': 1.2718414068222046}
+02/26/2022 13:21:31 - INFO - codeparrot_training - Step 46151: {'lr': 7.890874334257031e-06, 'samples': 23629824, 'steps': 46151, 'loss/train': 2.383247137069702}
+02/26/2022 13:21:37 - INFO - codeparrot_training - Step 46152: {'lr': 7.886796339281466e-06, 'samples': 23630336, 'steps': 46152, 'loss/train': 2.277276039123535}
+02/26/2022 13:21:40 - INFO - codeparrot_training - Step 46153: {'lr': 7.882719381441944e-06, 'samples': 23630848, 'steps': 46153, 'loss/train': 1.1779136657714844}
+02/26/2022 13:21:46 - INFO - codeparrot_training - Step 46154: {'lr': 7.87864346075598e-06, 'samples': 23631360, 'steps': 46154, 'loss/train': 1.4246450662612915}
+02/26/2022 13:21:49 - INFO - codeparrot_training - Step 46155: {'lr': 7.87456857724103e-06, 'samples': 23631872, 'steps': 46155, 'loss/train': 0.8637968897819519}
+02/26/2022 13:21:55 - INFO - codeparrot_training - Step 46156: {'lr': 7.870494730914556e-06, 'samples': 23632384, 'steps': 46156, 'loss/train': 1.0921019315719604}
+02/26/2022 13:21:58 - INFO - codeparrot_training - Step 46157: {'lr': 7.866421921793987e-06, 'samples': 23632896, 'steps': 46157, 'loss/train': 0.554743766784668}
+02/26/2022 13:22:04 - INFO - codeparrot_training - Step 46158: {'lr': 7.862350149896779e-06, 'samples': 23633408, 'steps': 46158, 'loss/train': 1.5484811067581177}
+02/26/2022 13:22:07 - INFO - codeparrot_training - Step 46159: {'lr': 7.858279415240365e-06, 'samples': 23633920, 'steps': 46159, 'loss/train': 1.6888887882232666}
+02/26/2022 13:22:14 - INFO - codeparrot_training - Step 46160: {'lr': 7.854209717842232e-06, 'samples': 23634432, 'steps': 46160, 'loss/train': 1.7405825853347778}
+02/26/2022 13:22:17 - INFO - codeparrot_training - Step 46161: {'lr': 7.85014105771975e-06, 'samples': 23634944, 'steps': 46161, 'loss/train': 1.9780075550079346}
+02/26/2022 13:22:22 - INFO - codeparrot_training - Step 46162: {'lr': 7.846073434890383e-06, 'samples': 23635456, 'steps': 46162, 'loss/train': 1.694478988647461}
+02/26/2022 13:22:26 - INFO - codeparrot_training - Step 46163: {'lr': 7.842006849371502e-06, 'samples': 23635968, 'steps': 46163, 'loss/train': 1.0776216983795166}
+02/26/2022 13:22:31 - INFO - codeparrot_training - Step 46164: {'lr': 7.837941301180624e-06, 'samples': 23636480, 'steps': 46164, 'loss/train': 1.905545949935913}
+02/26/2022 13:22:35 - INFO - codeparrot_training - Step 46165: {'lr': 7.833876790335093e-06, 'samples': 23636992, 'steps': 46165, 'loss/train': 2.163379192352295}
+02/26/2022 13:22:41 - INFO - codeparrot_training - Step 46166: {'lr': 7.829813316852342e-06, 'samples': 23637504, 'steps': 46166, 'loss/train': 0.5601224303245544}
+02/26/2022 13:22:44 - INFO - codeparrot_training - Step 46167: {'lr': 7.825750880749743e-06, 'samples': 23638016, 'steps': 46167, 'loss/train': 1.049468994140625}
+02/26/2022 13:22:49 - INFO - codeparrot_training - Step 46168: {'lr': 7.821689482044758e-06, 'samples': 23638528, 'steps': 46168, 'loss/train': 0.643880307674408}
+02/26/2022 13:22:53 - INFO - codeparrot_training - Step 46169: {'lr': 7.817629120754732e-06, 'samples': 23639040, 'steps': 46169, 'loss/train': 1.6984556913375854}
+02/26/2022 13:22:59 - INFO - codeparrot_training - Step 46170: {'lr': 7.813569796897124e-06, 'samples': 23639552, 'steps': 46170, 'loss/train': 0.9831675887107849}
+02/26/2022 13:23:03 - INFO - codeparrot_training - Step 46171: {'lr': 7.809511510489226e-06, 'samples': 23640064, 'steps': 46171, 'loss/train': 1.606002688407898}
+02/26/2022 13:23:08 - INFO - codeparrot_training - Step 46172: {'lr': 7.805454261548495e-06, 'samples': 23640576, 'steps': 46172, 'loss/train': 2.493654251098633}
+02/26/2022 13:23:12 - INFO - codeparrot_training - Step 46173: {'lr': 7.801398050092306e-06, 'samples': 23641088, 'steps': 46173, 'loss/train': 1.751905083656311}
+02/26/2022 13:23:17 - INFO - codeparrot_training - Step 46174: {'lr': 7.797342876138037e-06, 'samples': 23641600, 'steps': 46174, 'loss/train': 2.019636869430542}
+02/26/2022 13:23:21 - INFO - codeparrot_training - Step 46175: {'lr': 7.793288739703003e-06, 'samples': 23642112, 'steps': 46175, 'loss/train': 1.7807376384735107}
+02/26/2022 13:23:27 - INFO - codeparrot_training - Step 46176: {'lr': 7.789235640804637e-06, 'samples': 23642624, 'steps': 46176, 'loss/train': 2.1343982219696045}
+02/26/2022 13:23:30 - INFO - codeparrot_training - Step 46177: {'lr': 7.785183579460258e-06, 'samples': 23643136, 'steps': 46177, 'loss/train': 1.0486819744110107}
+02/26/2022 13:23:34 - INFO - codeparrot_training - Step 46178: {'lr': 7.781132555687242e-06, 'samples': 23643648, 'steps': 46178, 'loss/train': 1.5688691139221191}
+02/26/2022 13:23:39 - INFO - codeparrot_training - Step 46179: {'lr': 7.777082569502964e-06, 'samples': 23644160, 'steps': 46179, 'loss/train': 1.3218811750411987}
+02/26/2022 13:23:43 - INFO - codeparrot_training - Step 46180: {'lr': 7.773033620924741e-06, 'samples': 23644672, 'steps': 46180, 'loss/train': 1.9818297624588013}
+02/26/2022 13:23:49 - INFO - codeparrot_training - Step 46181: {'lr': 7.768985709969922e-06, 'samples': 23645184, 'steps': 46181, 'loss/train': 1.5427573919296265}
+02/26/2022 13:23:52 - INFO - codeparrot_training - Step 46182: {'lr': 7.764938836655828e-06, 'samples': 23645696, 'steps': 46182, 'loss/train': 2.053532838821411}
+02/26/2022 13:23:58 - INFO - codeparrot_training - Step 46183: {'lr': 7.760893000999858e-06, 'samples': 23646208, 'steps': 46183, 'loss/train': 0.8363004326820374}
+02/26/2022 13:24:01 - INFO - codeparrot_training - Step 46184: {'lr': 7.756848203019279e-06, 'samples': 23646720, 'steps': 46184, 'loss/train': 2.4321486949920654}
+02/26/2022 13:24:07 - INFO - codeparrot_training - Step 46185: {'lr': 7.752804442731437e-06, 'samples': 23647232, 'steps': 46185, 'loss/train': 1.4808934926986694}
+02/26/2022 13:24:10 - INFO - codeparrot_training - Step 46186: {'lr': 7.748761720153652e-06, 'samples': 23647744, 'steps': 46186, 'loss/train': 2.048483371734619}
+02/26/2022 13:24:16 - INFO - codeparrot_training - Step 46187: {'lr': 7.744720035303298e-06, 'samples': 23648256, 'steps': 46187, 'loss/train': 2.595313787460327}
+02/26/2022 13:24:19 - INFO - codeparrot_training - Step 46188: {'lr': 7.740679388197586e-06, 'samples': 23648768, 'steps': 46188, 'loss/train': 0.47151291370391846}
+02/26/2022 13:24:25 - INFO - codeparrot_training - Step 46189: {'lr': 7.736639778853888e-06, 'samples': 23649280, 'steps': 46189, 'loss/train': 0.3350508511066437}
+02/26/2022 13:24:30 - INFO - codeparrot_training - Step 46190: {'lr': 7.732601207289497e-06, 'samples': 23649792, 'steps': 46190, 'loss/train': 2.1283323764801025}
+02/26/2022 13:24:34 - INFO - codeparrot_training - Step 46191: {'lr': 7.728563673521732e-06, 'samples': 23650304, 'steps': 46191, 'loss/train': 2.0503880977630615}
+02/26/2022 13:24:39 - INFO - codeparrot_training - Step 46192: {'lr': 7.724527177567858e-06, 'samples': 23650816, 'steps': 46192, 'loss/train': 2.233004093170166}
+02/26/2022 13:24:43 - INFO - codeparrot_training - Step 46193: {'lr': 7.720491719445195e-06, 'samples': 23651328, 'steps': 46193, 'loss/train': 0.49077585339546204}
+02/26/2022 13:24:48 - INFO - codeparrot_training - Step 46194: {'lr': 7.716457299170975e-06, 'samples': 23651840, 'steps': 46194, 'loss/train': 1.1692137718200684}
+02/26/2022 13:24:52 - INFO - codeparrot_training - Step 46195: {'lr': 7.712423916762552e-06, 'samples': 23652352, 'steps': 46195, 'loss/train': 1.0794163942337036}
+02/26/2022 13:24:58 - INFO - codeparrot_training - Step 46196: {'lr': 7.708391572237183e-06, 'samples': 23652864, 'steps': 46196, 'loss/train': 1.959424614906311}
+02/26/2022 13:25:02 - INFO - codeparrot_training - Step 46197: {'lr': 7.70436026561211e-06, 'samples': 23653376, 'steps': 46197, 'loss/train': 1.2503868341445923}
+02/26/2022 13:25:07 - INFO - codeparrot_training - Step 46198: {'lr': 7.700329996904593e-06, 'samples': 23653888, 'steps': 46198, 'loss/train': 2.1610403060913086}
+02/26/2022 13:25:11 - INFO - codeparrot_training - Step 46199: {'lr': 7.696300766131952e-06, 'samples': 23654400, 'steps': 46199, 'loss/train': 1.25043523311615}
+02/26/2022 13:25:16 - INFO - codeparrot_training - Step 46200: {'lr': 7.692272573311426e-06, 'samples': 23654912, 'steps': 46200, 'loss/train': 1.3419629335403442}
+02/26/2022 13:25:20 - INFO - codeparrot_training - Step 46201: {'lr': 7.68824541846025e-06, 'samples': 23655424, 'steps': 46201, 'loss/train': 2.131420373916626}
+02/26/2022 13:25:25 - INFO - codeparrot_training - Step 46202: {'lr': 7.684219301595658e-06, 'samples': 23655936, 'steps': 46202, 'loss/train': 2.698681116104126}
+02/26/2022 13:25:28 - INFO - codeparrot_training - Step 46203: {'lr': 7.68019422273497e-06, 'samples': 23656448, 'steps': 46203, 'loss/train': 1.2755769491195679}
+02/26/2022 13:25:34 - INFO - codeparrot_training - Step 46204: {'lr': 7.676170181895342e-06, 'samples': 23656960, 'steps': 46204, 'loss/train': 2.321380376815796}
+02/26/2022 13:25:37 - INFO - codeparrot_training - Step 46205: {'lr': 7.67214717909412e-06, 'samples': 23657472, 'steps': 46205, 'loss/train': 1.3493386507034302}
+02/26/2022 13:25:44 - INFO - codeparrot_training - Step 46206: {'lr': 7.668125214348427e-06, 'samples': 23657984, 'steps': 46206, 'loss/train': 1.7210662364959717}
+02/26/2022 13:25:47 - INFO - codeparrot_training - Step 46207: {'lr': 7.664104287675555e-06, 'samples': 23658496, 'steps': 46207, 'loss/train': 1.959551215171814}
+02/26/2022 13:25:52 - INFO - codeparrot_training - Step 46208: {'lr': 7.660084399092659e-06, 'samples': 23659008, 'steps': 46208, 'loss/train': 1.615695595741272}
+02/26/2022 13:25:56 - INFO - codeparrot_training - Step 46209: {'lr': 7.656065548617086e-06, 'samples': 23659520, 'steps': 46209, 'loss/train': 1.3801510334014893}
+02/26/2022 13:26:02 - INFO - codeparrot_training - Step 46210: {'lr': 7.652047736265933e-06, 'samples': 23660032, 'steps': 46210, 'loss/train': 1.4071534872055054}
+02/26/2022 13:26:05 - INFO - codeparrot_training - Step 46211: {'lr': 7.648030962056463e-06, 'samples': 23660544, 'steps': 46211, 'loss/train': 1.8339293003082275}
+02/26/2022 13:26:11 - INFO - codeparrot_training - Step 46212: {'lr': 7.644015226005857e-06, 'samples': 23661056, 'steps': 46212, 'loss/train': 0.5704971551895142}
+02/26/2022 13:26:14 - INFO - codeparrot_training - Step 46213: {'lr': 7.64000052813138e-06, 'samples': 23661568, 'steps': 46213, 'loss/train': 1.8081496953964233}
+02/26/2022 13:26:20 - INFO - codeparrot_training - Step 46214: {'lr': 7.635986868450129e-06, 'samples': 23662080, 'steps': 46214, 'loss/train': 1.4197722673416138}
+02/26/2022 13:26:23 - INFO - codeparrot_training - Step 46215: {'lr': 7.631974246979367e-06, 'samples': 23662592, 'steps': 46215, 'loss/train': 1.6627098321914673}
+02/26/2022 13:26:28 - INFO - codeparrot_training - Step 46216: {'lr': 7.627962663736277e-06, 'samples': 23663104, 'steps': 46216, 'loss/train': 2.077439308166504}
+02/26/2022 13:26:32 - INFO - codeparrot_training - Step 46217: {'lr': 7.623952118738009e-06, 'samples': 23663616, 'steps': 46217, 'loss/train': 2.4660086631774902}
+02/26/2022 13:26:38 - INFO - codeparrot_training - Step 46218: {'lr': 7.619942612001801e-06, 'samples': 23664128, 'steps': 46218, 'loss/train': 1.0642417669296265}
+02/26/2022 13:26:42 - INFO - codeparrot_training - Step 46219: {'lr': 7.615934143544778e-06, 'samples': 23664640, 'steps': 46219, 'loss/train': 2.13999342918396}
+02/26/2022 13:26:47 - INFO - codeparrot_training - Step 46220: {'lr': 7.611926713384121e-06, 'samples': 23665152, 'steps': 46220, 'loss/train': 1.95249342918396}
+02/26/2022 13:26:51 - INFO - codeparrot_training - Step 46221: {'lr': 7.6079203215370095e-06, 'samples': 23665664, 'steps': 46221, 'loss/train': 1.711698055267334}
+02/26/2022 13:26:56 - INFO - codeparrot_training - Step 46222: {'lr': 7.6039149680205975e-06, 'samples': 23666176, 'steps': 46222, 'loss/train': 1.4104186296463013}
+02/26/2022 13:27:00 - INFO - codeparrot_training - Step 46223: {'lr': 7.599910652852066e-06, 'samples': 23666688, 'steps': 46223, 'loss/train': 2.186832904815674}
+02/26/2022 13:27:05 - INFO - codeparrot_training - Step 46224: {'lr': 7.5959073760485105e-06, 'samples': 23667200, 'steps': 46224, 'loss/train': 2.075455665588379}
+02/26/2022 13:27:09 - INFO - codeparrot_training - Step 46225: {'lr': 7.591905137627114e-06, 'samples': 23667712, 'steps': 46225, 'loss/train': 1.6174030303955078}
+02/26/2022 13:27:14 - INFO - codeparrot_training - Step 46226: {'lr': 7.587903937605056e-06, 'samples': 23668224, 'steps': 46226, 'loss/train': 1.9038349390029907}
+02/26/2022 13:27:18 - INFO - codeparrot_training - Step 46227: {'lr': 7.583903775999435e-06, 'samples': 23668736, 'steps': 46227, 'loss/train': 2.4301764965057373}
+02/26/2022 13:27:24 - INFO - codeparrot_training - Step 46228: {'lr': 7.579904652827402e-06, 'samples': 23669248, 'steps': 46228, 'loss/train': 1.7941884994506836}
+02/26/2022 13:27:27 - INFO - codeparrot_training - Step 46229: {'lr': 7.575906568106028e-06, 'samples': 23669760, 'steps': 46229, 'loss/train': 2.422224283218384}
+02/26/2022 13:27:33 - INFO - codeparrot_training - Step 46230: {'lr': 7.571909521852549e-06, 'samples': 23670272, 'steps': 46230, 'loss/train': 1.4480489492416382}
+02/26/2022 13:27:36 - INFO - codeparrot_training - Step 46231: {'lr': 7.5679135140840076e-06, 'samples': 23670784, 'steps': 46231, 'loss/train': 0.734429657459259}
+02/26/2022 13:27:42 - INFO - codeparrot_training - Step 46232: {'lr': 7.563918544817555e-06, 'samples': 23671296, 'steps': 46232, 'loss/train': 3.3537869453430176}
+02/26/2022 13:27:45 - INFO - codeparrot_training - Step 46233: {'lr': 7.559924614070263e-06, 'samples': 23671808, 'steps': 46233, 'loss/train': 0.9917415976524353}
+02/26/2022 13:27:51 - INFO - codeparrot_training - Step 46234: {'lr': 7.555931721859311e-06, 'samples': 23672320, 'steps': 46234, 'loss/train': 1.5242668390274048}
+02/26/2022 13:27:54 - INFO - codeparrot_training - Step 46235: {'lr': 7.55193986820174e-06, 'samples': 23672832, 'steps': 46235, 'loss/train': 1.3739936351776123}
+02/26/2022 13:28:00 - INFO - codeparrot_training - Step 46236: {'lr': 7.547949053114705e-06, 'samples': 23673344, 'steps': 46236, 'loss/train': 2.2427783012390137}
+02/26/2022 13:28:03 - INFO - codeparrot_training - Step 46237: {'lr': 7.54395927661522e-06, 'samples': 23673856, 'steps': 46237, 'loss/train': 2.017179250717163}
+02/26/2022 13:28:09 - INFO - codeparrot_training - Step 46238: {'lr': 7.5399705387204365e-06, 'samples': 23674368, 'steps': 46238, 'loss/train': 0.41155701875686646}
+02/26/2022 13:28:12 - INFO - codeparrot_training - Step 46239: {'lr': 7.535982839447425e-06, 'samples': 23674880, 'steps': 46239, 'loss/train': 0.9758650660514832}
+02/26/2022 13:28:18 - INFO - codeparrot_training - Step 46240: {'lr': 7.53199617881331e-06, 'samples': 23675392, 'steps': 46240, 'loss/train': 1.9534239768981934}
+02/26/2022 13:28:21 - INFO - codeparrot_training - Step 46241: {'lr': 7.528010556835108e-06, 'samples': 23675904, 'steps': 46241, 'loss/train': 1.3885554075241089}
+02/26/2022 13:28:27 - INFO - codeparrot_training - Step 46242: {'lr': 7.5240259735299135e-06, 'samples': 23676416, 'steps': 46242, 'loss/train': 1.9461522102355957}
+02/26/2022 13:28:30 - INFO - codeparrot_training - Step 46243: {'lr': 7.5200424289147695e-06, 'samples': 23676928, 'steps': 46243, 'loss/train': 2.7684545516967773}
+02/26/2022 13:28:37 - INFO - codeparrot_training - Step 46244: {'lr': 7.516059923006829e-06, 'samples': 23677440, 'steps': 46244, 'loss/train': 2.2178447246551514}
+02/26/2022 13:28:40 - INFO - codeparrot_training - Step 46245: {'lr': 7.512078455823052e-06, 'samples': 23677952, 'steps': 46245, 'loss/train': 1.61265230178833}
+02/26/2022 13:28:46 - INFO - codeparrot_training - Step 46246: {'lr': 7.508098027380534e-06, 'samples': 23678464, 'steps': 46246, 'loss/train': 1.355047345161438}
+02/26/2022 13:28:49 - INFO - codeparrot_training - Step 46247: {'lr': 7.504118637696317e-06, 'samples': 23678976, 'steps': 46247, 'loss/train': 1.4046990871429443}
+02/26/2022 13:28:55 - INFO - codeparrot_training - Step 46248: {'lr': 7.5001402867874995e-06, 'samples': 23679488, 'steps': 46248, 'loss/train': 1.0004087686538696}
+02/26/2022 13:28:58 - INFO - codeparrot_training - Step 46249: {'lr': 7.496162974671039e-06, 'samples': 23680000, 'steps': 46249, 'loss/train': 2.097203016281128}
+02/26/2022 13:29:04 - INFO - codeparrot_training - Step 46250: {'lr': 7.4921867013640064e-06, 'samples': 23680512, 'steps': 46250, 'loss/train': 0.35356733202934265}
+02/26/2022 13:29:07 - INFO - codeparrot_training - Step 46251: {'lr': 7.488211466883443e-06, 'samples': 23681024, 'steps': 46251, 'loss/train': 1.7939902544021606}
+02/26/2022 13:29:13 - INFO - codeparrot_training - Step 46252: {'lr': 7.484237271246391e-06, 'samples': 23681536, 'steps': 46252, 'loss/train': 1.8993935585021973}
+02/26/2022 13:29:16 - INFO - codeparrot_training - Step 46253: {'lr': 7.480264114469865e-06, 'samples': 23682048, 'steps': 46253, 'loss/train': 1.6327611207962036}
+02/26/2022 13:29:22 - INFO - codeparrot_training - Step 46254: {'lr': 7.476291996570877e-06, 'samples': 23682560, 'steps': 46254, 'loss/train': 0.8919408321380615}
+02/26/2022 13:29:26 - INFO - codeparrot_training - Step 46255: {'lr': 7.472320917566416e-06, 'samples': 23683072, 'steps': 46255, 'loss/train': 1.2883343696594238}
+02/26/2022 13:29:31 - INFO - codeparrot_training - Step 46256: {'lr': 7.468350877473551e-06, 'samples': 23683584, 'steps': 46256, 'loss/train': 0.9894641637802124}
+02/26/2022 13:29:35 - INFO - codeparrot_training - Step 46257: {'lr': 7.46438187630924e-06, 'samples': 23684096, 'steps': 46257, 'loss/train': 1.46388840675354}
+02/26/2022 13:29:40 - INFO - codeparrot_training - Step 46258: {'lr': 7.460413914090525e-06, 'samples': 23684608, 'steps': 46258, 'loss/train': 0.6810938119888306}
+02/26/2022 13:29:44 - INFO - codeparrot_training - Step 46259: {'lr': 7.456446990834337e-06, 'samples': 23685120, 'steps': 46259, 'loss/train': 1.031083106994629}
+02/26/2022 13:29:49 - INFO - codeparrot_training - Step 46260: {'lr': 7.452481106557746e-06, 'samples': 23685632, 'steps': 46260, 'loss/train': 1.3681429624557495}
+02/26/2022 13:29:53 - INFO - codeparrot_training - Step 46261: {'lr': 7.44851626127771e-06, 'samples': 23686144, 'steps': 46261, 'loss/train': 0.2631775140762329}
+02/26/2022 13:29:58 - INFO - codeparrot_training - Step 46262: {'lr': 7.444552455011216e-06, 'samples': 23686656, 'steps': 46262, 'loss/train': 1.740143060684204}
+02/26/2022 13:30:02 - INFO - codeparrot_training - Step 46263: {'lr': 7.440589687775251e-06, 'samples': 23687168, 'steps': 46263, 'loss/train': 1.8481998443603516}
+02/26/2022 13:30:08 - INFO - codeparrot_training - Step 46264: {'lr': 7.436627959586744e-06, 'samples': 23687680, 'steps': 46264, 'loss/train': 3.0240542888641357}
+02/26/2022 13:30:12 - INFO - codeparrot_training - Step 46265: {'lr': 7.432667270462712e-06, 'samples': 23688192, 'steps': 46265, 'loss/train': 1.746176838874817}
+02/26/2022 13:30:17 - INFO - codeparrot_training - Step 46266: {'lr': 7.428707620420138e-06, 'samples': 23688704, 'steps': 46266, 'loss/train': 0.054960038512945175}
+02/26/2022 13:30:21 - INFO - codeparrot_training - Step 46267: {'lr': 7.424749009475929e-06, 'samples': 23689216, 'steps': 46267, 'loss/train': 1.7184443473815918}
+02/26/2022 13:30:26 - INFO - codeparrot_training - Step 46268: {'lr': 7.4207914376470395e-06, 'samples': 23689728, 'steps': 46268, 'loss/train': 1.4960073232650757}
+02/26/2022 13:30:30 - INFO - codeparrot_training - Step 46269: {'lr': 7.4168349049504864e-06, 'samples': 23690240, 'steps': 46269, 'loss/train': 2.735445499420166}
+02/26/2022 13:30:35 - INFO - codeparrot_training - Step 46270: {'lr': 7.4128794114032e-06, 'samples': 23690752, 'steps': 46270, 'loss/train': 2.057077407836914}
+02/26/2022 13:30:39 - INFO - codeparrot_training - Step 46271: {'lr': 7.408924957022084e-06, 'samples': 23691264, 'steps': 46271, 'loss/train': 1.2970287799835205}
+02/26/2022 13:30:44 - INFO - codeparrot_training - Step 46272: {'lr': 7.4049715418240955e-06, 'samples': 23691776, 'steps': 46272, 'loss/train': 1.7952051162719727}
+02/26/2022 13:30:48 - INFO - codeparrot_training - Step 46273: {'lr': 7.401019165826195e-06, 'samples': 23692288, 'steps': 46273, 'loss/train': 1.2831530570983887}
+02/26/2022 13:30:54 - INFO - codeparrot_training - Step 46274: {'lr': 7.397067829045284e-06, 'samples': 23692800, 'steps': 46274, 'loss/train': 1.8026372194290161}
+02/26/2022 13:30:57 - INFO - codeparrot_training - Step 46275: {'lr': 7.393117531498322e-06, 'samples': 23693312, 'steps': 46275, 'loss/train': 2.117004632949829}
+02/26/2022 13:31:03 - INFO - codeparrot_training - Step 46276: {'lr': 7.389168273202157e-06, 'samples': 23693824, 'steps': 46276, 'loss/train': 1.5674726963043213}
+02/26/2022 13:31:06 - INFO - codeparrot_training - Step 46277: {'lr': 7.385220054173802e-06, 'samples': 23694336, 'steps': 46277, 'loss/train': 0.9597045183181763}
+02/26/2022 13:31:12 - INFO - codeparrot_training - Step 46278: {'lr': 7.381272874430078e-06, 'samples': 23694848, 'steps': 46278, 'loss/train': 1.9084405899047852}
+02/26/2022 13:31:15 - INFO - codeparrot_training - Step 46279: {'lr': 7.377326733988027e-06, 'samples': 23695360, 'steps': 46279, 'loss/train': 1.8546720743179321}
+02/26/2022 13:31:21 - INFO - codeparrot_training - Step 46280: {'lr': 7.373381632864384e-06, 'samples': 23695872, 'steps': 46280, 'loss/train': 1.6214810609817505}
+02/26/2022 13:31:24 - INFO - codeparrot_training - Step 46281: {'lr': 7.369437571076165e-06, 'samples': 23696384, 'steps': 46281, 'loss/train': 2.410975933074951}
+02/26/2022 13:31:30 - INFO - codeparrot_training - Step 46282: {'lr': 7.365494548640217e-06, 'samples': 23696896, 'steps': 46282, 'loss/train': 1.1699903011322021}
+02/26/2022 13:31:33 - INFO - codeparrot_training - Step 46283: {'lr': 7.361552565573499e-06, 'samples': 23697408, 'steps': 46283, 'loss/train': 1.9359527826309204}
+02/26/2022 13:31:39 - INFO - codeparrot_training - Step 46284: {'lr': 7.357611621892774e-06, 'samples': 23697920, 'steps': 46284, 'loss/train': 1.2825313806533813}
+02/26/2022 13:31:44 - INFO - codeparrot_training - Step 46285: {'lr': 7.353671717615029e-06, 'samples': 23698432, 'steps': 46285, 'loss/train': 0.47180798649787903}
+02/26/2022 13:31:48 - INFO - codeparrot_training - Step 46286: {'lr': 7.349732852757085e-06, 'samples': 23698944, 'steps': 46286, 'loss/train': 1.7396972179412842}
+02/26/2022 13:31:53 - INFO - codeparrot_training - Step 46287: {'lr': 7.345795027335844e-06, 'samples': 23699456, 'steps': 46287, 'loss/train': 2.2069430351257324}
+02/26/2022 13:31:57 - INFO - codeparrot_training - Step 46288: {'lr': 7.341858241368182e-06, 'samples': 23699968, 'steps': 46288, 'loss/train': 1.7909590005874634}
+02/26/2022 13:32:03 - INFO - codeparrot_training - Step 46289: {'lr': 7.3379224948709465e-06, 'samples': 23700480, 'steps': 46289, 'loss/train': 2.207418441772461}
+02/26/2022 13:32:06 - INFO - codeparrot_training - Step 46290: {'lr': 7.333987787860958e-06, 'samples': 23700992, 'steps': 46290, 'loss/train': 0.9212024211883545}
+02/26/2022 13:32:12 - INFO - codeparrot_training - Step 46291: {'lr': 7.330054120355146e-06, 'samples': 23701504, 'steps': 46291, 'loss/train': 2.1827926635742188}
+02/26/2022 13:32:15 - INFO - codeparrot_training - Step 46292: {'lr': 7.32612149237033e-06, 'samples': 23702016, 'steps': 46292, 'loss/train': 0.8848879337310791}
+02/26/2022 13:32:21 - INFO - codeparrot_training - Step 46293: {'lr': 7.322189903923332e-06, 'samples': 23702528, 'steps': 46293, 'loss/train': 2.041585922241211}
+02/26/2022 13:32:24 - INFO - codeparrot_training - Step 46294: {'lr': 7.318259355030999e-06, 'samples': 23703040, 'steps': 46294, 'loss/train': 2.701075792312622}
+02/26/2022 13:32:30 - INFO - codeparrot_training - Step 46295: {'lr': 7.314329845710205e-06, 'samples': 23703552, 'steps': 46295, 'loss/train': 1.5426957607269287}
+02/26/2022 13:32:33 - INFO - codeparrot_training - Step 46296: {'lr': 7.310401375977771e-06, 'samples': 23704064, 'steps': 46296, 'loss/train': 1.7253940105438232}
+02/26/2022 13:32:39 - INFO - codeparrot_training - Step 46297: {'lr': 7.306473945850517e-06, 'samples': 23704576, 'steps': 46297, 'loss/train': 0.8173562288284302}
+02/26/2022 13:32:42 - INFO - codeparrot_training - Step 46298: {'lr': 7.302547555345234e-06, 'samples': 23705088, 'steps': 46298, 'loss/train': 0.8185781240463257}
+02/26/2022 13:32:48 - INFO - codeparrot_training - Step 46299: {'lr': 7.2986222044788e-06, 'samples': 23705600, 'steps': 46299, 'loss/train': 2.2433414459228516}
+02/26/2022 13:32:52 - INFO - codeparrot_training - Step 46300: {'lr': 7.294697893267976e-06, 'samples': 23706112, 'steps': 46300, 'loss/train': 1.1419854164123535}
+02/26/2022 13:32:57 - INFO - codeparrot_training - Step 46301: {'lr': 7.29077462172964e-06, 'samples': 23706624, 'steps': 46301, 'loss/train': 4.863748073577881}
+02/26/2022 13:33:01 - INFO - codeparrot_training - Step 46302: {'lr': 7.2868523898805e-06, 'samples': 23707136, 'steps': 46302, 'loss/train': 1.9545962810516357}
+02/26/2022 13:33:06 - INFO - codeparrot_training - Step 46303: {'lr': 7.282931197737458e-06, 'samples': 23707648, 'steps': 46303, 'loss/train': 2.0479788780212402}
+02/26/2022 13:33:10 - INFO - codeparrot_training - Step 46304: {'lr': 7.279011045317252e-06, 'samples': 23708160, 'steps': 46304, 'loss/train': 0.6766265034675598}
+02/26/2022 13:33:15 - INFO - codeparrot_training - Step 46305: {'lr': 7.275091932636702e-06, 'samples': 23708672, 'steps': 46305, 'loss/train': 0.6496062874794006}
+02/26/2022 13:33:19 - INFO - codeparrot_training - Step 46306: {'lr': 7.271173859712571e-06, 'samples': 23709184, 'steps': 46306, 'loss/train': 0.18399237096309662}
+02/26/2022 13:33:25 - INFO - codeparrot_training - Step 46307: {'lr': 7.267256826561652e-06, 'samples': 23709696, 'steps': 46307, 'loss/train': 1.5856776237487793}
+02/26/2022 13:33:29 - INFO - codeparrot_training - Step 46308: {'lr': 7.263340833200738e-06, 'samples': 23710208, 'steps': 46308, 'loss/train': 2.046168327331543}
+02/26/2022 13:33:34 - INFO - codeparrot_training - Step 46309: {'lr': 7.25942587964662e-06, 'samples': 23710720, 'steps': 46309, 'loss/train': 1.9194923639297485}
+02/26/2022 13:33:38 - INFO - codeparrot_training - Step 46310: {'lr': 7.255511965916006e-06, 'samples': 23711232, 'steps': 46310, 'loss/train': 1.292546272277832}
+02/26/2022 13:33:43 - INFO - codeparrot_training - Step 46311: {'lr': 7.25159909202569e-06, 'samples': 23711744, 'steps': 46311, 'loss/train': 1.5153529644012451}
+02/26/2022 13:33:47 - INFO - codeparrot_training - Step 46312: {'lr': 7.247687257992463e-06, 'samples': 23712256, 'steps': 46312, 'loss/train': 1.8743751049041748}
+02/26/2022 13:33:52 - INFO - codeparrot_training - Step 46313: {'lr': 7.243776463833035e-06, 'samples': 23712768, 'steps': 46313, 'loss/train': 2.3574328422546387}
+02/26/2022 13:33:56 - INFO - codeparrot_training - Step 46314: {'lr': 7.239866709564252e-06, 'samples': 23713280, 'steps': 46314, 'loss/train': 2.1100828647613525}
+02/26/2022 13:34:01 - INFO - codeparrot_training - Step 46315: {'lr': 7.235957995202741e-06, 'samples': 23713792, 'steps': 46315, 'loss/train': 2.3595588207244873}
+02/26/2022 13:34:05 - INFO - codeparrot_training - Step 46316: {'lr': 7.232050320765321e-06, 'samples': 23714304, 'steps': 46316, 'loss/train': 1.548946738243103}
+02/26/2022 13:34:11 - INFO - codeparrot_training - Step 46317: {'lr': 7.2281436862686735e-06, 'samples': 23714816, 'steps': 46317, 'loss/train': 0.9009461998939514}
+02/26/2022 13:34:14 - INFO - codeparrot_training - Step 46318: {'lr': 7.224238091729618e-06, 'samples': 23715328, 'steps': 46318, 'loss/train': 2.0814337730407715}
+02/26/2022 13:34:21 - INFO - codeparrot_training - Step 46319: {'lr': 7.220333537164808e-06, 'samples': 23715840, 'steps': 46319, 'loss/train': 2.3470852375030518}
+02/26/2022 13:34:24 - INFO - codeparrot_training - Step 46320: {'lr': 7.216430022591009e-06, 'samples': 23716352, 'steps': 46320, 'loss/train': 1.624616026878357}
+02/26/2022 13:34:30 - INFO - codeparrot_training - Step 46321: {'lr': 7.2125275480249e-06, 'samples': 23716864, 'steps': 46321, 'loss/train': 1.901342749595642}
+02/26/2022 13:34:33 - INFO - codeparrot_training - Step 46322: {'lr': 7.208626113483274e-06, 'samples': 23717376, 'steps': 46322, 'loss/train': 0.9478985071182251}
+02/26/2022 13:34:39 - INFO - codeparrot_training - Step 46323: {'lr': 7.204725718982785e-06, 'samples': 23717888, 'steps': 46323, 'loss/train': 3.80117130279541}
+02/26/2022 13:34:42 - INFO - codeparrot_training - Step 46324: {'lr': 7.200826364540142e-06, 'samples': 23718400, 'steps': 46324, 'loss/train': 1.3132497072219849}
+02/26/2022 13:34:48 - INFO - codeparrot_training - Step 46325: {'lr': 7.196928050172052e-06, 'samples': 23718912, 'steps': 46325, 'loss/train': 2.0616514682769775}
+02/26/2022 13:34:51 - INFO - codeparrot_training - Step 46326: {'lr': 7.193030775895254e-06, 'samples': 23719424, 'steps': 46326, 'loss/train': 1.5557838678359985}
+02/26/2022 13:34:57 - INFO - codeparrot_training - Step 46327: {'lr': 7.1891345417263995e-06, 'samples': 23719936, 'steps': 46327, 'loss/train': 1.473690152168274}
+02/26/2022 13:35:00 - INFO - codeparrot_training - Step 46328: {'lr': 7.185239347682199e-06, 'samples': 23720448, 'steps': 46328, 'loss/train': 1.9749585390090942}
+02/26/2022 13:35:06 - INFO - codeparrot_training - Step 46329: {'lr': 7.181345193779304e-06, 'samples': 23720960, 'steps': 46329, 'loss/train': 0.7785266041755676}
+02/26/2022 13:35:09 - INFO - codeparrot_training - Step 46330: {'lr': 7.177452080034452e-06, 'samples': 23721472, 'steps': 46330, 'loss/train': 0.4698942005634308}
+02/26/2022 13:35:15 - INFO - codeparrot_training - Step 46331: {'lr': 7.173560006464269e-06, 'samples': 23721984, 'steps': 46331, 'loss/train': 2.16408109664917}
+02/26/2022 13:35:19 - INFO - codeparrot_training - Step 46332: {'lr': 7.169668973085463e-06, 'samples': 23722496, 'steps': 46332, 'loss/train': 2.0029828548431396}
+02/26/2022 13:35:24 - INFO - codeparrot_training - Step 46333: {'lr': 7.165778979914689e-06, 'samples': 23723008, 'steps': 46333, 'loss/train': 2.0213077068328857}
+02/26/2022 13:35:28 - INFO - codeparrot_training - Step 46334: {'lr': 7.161890026968598e-06, 'samples': 23723520, 'steps': 46334, 'loss/train': 2.0433108806610107}
+02/26/2022 13:35:33 - INFO - codeparrot_training - Step 46335: {'lr': 7.158002114263845e-06, 'samples': 23724032, 'steps': 46335, 'loss/train': 1.7285903692245483}
+02/26/2022 13:35:37 - INFO - codeparrot_training - Step 46336: {'lr': 7.154115241817166e-06, 'samples': 23724544, 'steps': 46336, 'loss/train': 1.5634167194366455}
+02/26/2022 13:35:42 - INFO - codeparrot_training - Step 46337: {'lr': 7.150229409645076e-06, 'samples': 23725056, 'steps': 46337, 'loss/train': 1.351056694984436}
+02/26/2022 13:35:46 - INFO - codeparrot_training - Step 46338: {'lr': 7.146344617764311e-06, 'samples': 23725568, 'steps': 46338, 'loss/train': 2.9512109756469727}
+02/26/2022 13:35:51 - INFO - codeparrot_training - Step 46339: {'lr': 7.142460866191497e-06, 'samples': 23726080, 'steps': 46339, 'loss/train': 2.253892660140991}
+02/26/2022 13:35:55 - INFO - codeparrot_training - Step 46340: {'lr': 7.138578154943287e-06, 'samples': 23726592, 'steps': 46340, 'loss/train': 2.0414609909057617}
+02/26/2022 13:36:00 - INFO - codeparrot_training - Step 46341: {'lr': 7.134696484036251e-06, 'samples': 23727104, 'steps': 46341, 'loss/train': 0.4035230875015259}
+02/26/2022 13:36:04 - INFO - codeparrot_training - Step 46342: {'lr': 7.130815853487071e-06, 'samples': 23727616, 'steps': 46342, 'loss/train': 0.42983752489089966}
+02/26/2022 13:36:10 - INFO - codeparrot_training - Step 46343: {'lr': 7.1269362633123725e-06, 'samples': 23728128, 'steps': 46343, 'loss/train': 2.3718860149383545}
+02/26/2022 13:36:13 - INFO - codeparrot_training - Step 46344: {'lr': 7.1230577135287525e-06, 'samples': 23728640, 'steps': 46344, 'loss/train': 0.7256945967674255}
+02/26/2022 13:36:20 - INFO - codeparrot_training - Step 46345: {'lr': 7.119180204152781e-06, 'samples': 23729152, 'steps': 46345, 'loss/train': 1.1530717611312866}
+02/26/2022 13:36:23 - INFO - codeparrot_training - Step 46346: {'lr': 7.115303735201168e-06, 'samples': 23729664, 'steps': 46346, 'loss/train': 2.5768351554870605}
+02/26/2022 13:36:29 - INFO - codeparrot_training - Step 46347: {'lr': 7.111428306690482e-06, 'samples': 23730176, 'steps': 46347, 'loss/train': 1.0652800798416138}
+02/26/2022 13:36:33 - INFO - codeparrot_training - Step 46348: {'lr': 7.107553918637266e-06, 'samples': 23730688, 'steps': 46348, 'loss/train': 1.7955538034439087}
+02/26/2022 13:36:38 - INFO - codeparrot_training - Step 46349: {'lr': 7.103680571058202e-06, 'samples': 23731200, 'steps': 46349, 'loss/train': 1.219046950340271}
+02/26/2022 13:36:41 - INFO - codeparrot_training - Step 46350: {'lr': 7.099808263969832e-06, 'samples': 23731712, 'steps': 46350, 'loss/train': 2.186893939971924}
+02/26/2022 13:36:47 - INFO - codeparrot_training - Step 46351: {'lr': 7.095936997388752e-06, 'samples': 23732224, 'steps': 46351, 'loss/train': 1.5777232646942139}
+02/26/2022 13:36:51 - INFO - codeparrot_training - Step 46352: {'lr': 7.092066771331507e-06, 'samples': 23732736, 'steps': 46352, 'loss/train': 0.8324767351150513}
+02/26/2022 13:36:56 - INFO - codeparrot_training - Step 46353: {'lr': 7.0881975858147764e-06, 'samples': 23733248, 'steps': 46353, 'loss/train': 0.571177065372467}
+02/26/2022 13:37:00 - INFO - codeparrot_training - Step 46354: {'lr': 7.084329440855048e-06, 'samples': 23733760, 'steps': 46354, 'loss/train': 2.2714080810546875}
+02/26/2022 13:37:06 - INFO - codeparrot_training - Step 46355: {'lr': 7.080462336468918e-06, 'samples': 23734272, 'steps': 46355, 'loss/train': 1.245651364326477}
+02/26/2022 13:37:10 - INFO - codeparrot_training - Step 46356: {'lr': 7.076596272672931e-06, 'samples': 23734784, 'steps': 46356, 'loss/train': 1.6411590576171875}
+02/26/2022 13:37:15 - INFO - codeparrot_training - Step 46357: {'lr': 7.072731249483738e-06, 'samples': 23735296, 'steps': 46357, 'loss/train': 1.6794277429580688}
+02/26/2022 13:37:19 - INFO - codeparrot_training - Step 46358: {'lr': 7.068867266917745e-06, 'samples': 23735808, 'steps': 46358, 'loss/train': 2.4114811420440674}
+02/26/2022 13:37:24 - INFO - codeparrot_training - Step 46359: {'lr': 7.065004324991631e-06, 'samples': 23736320, 'steps': 46359, 'loss/train': 0.2611892819404602}
+02/26/2022 13:37:28 - INFO - codeparrot_training - Step 46360: {'lr': 7.061142423721856e-06, 'samples': 23736832, 'steps': 46360, 'loss/train': 1.831305980682373}
+02/26/2022 13:37:33 - INFO - codeparrot_training - Step 46361: {'lr': 7.057281563125045e-06, 'samples': 23737344, 'steps': 46361, 'loss/train': 2.3734524250030518}
+02/26/2022 13:37:37 - INFO - codeparrot_training - Step 46362: {'lr': 7.053421743217714e-06, 'samples': 23737856, 'steps': 46362, 'loss/train': 2.57153582572937}
+02/26/2022 13:37:42 - INFO - codeparrot_training - Step 46363: {'lr': 7.049562964016349e-06, 'samples': 23738368, 'steps': 46363, 'loss/train': 1.5510889291763306}
+02/26/2022 13:37:46 - INFO - codeparrot_training - Step 46364: {'lr': 7.045705225537491e-06, 'samples': 23738880, 'steps': 46364, 'loss/train': 0.845072865486145}
+02/26/2022 13:37:52 - INFO - codeparrot_training - Step 46365: {'lr': 7.0418485277977115e-06, 'samples': 23739392, 'steps': 46365, 'loss/train': 1.5776699781417847}
+02/26/2022 13:37:56 - INFO - codeparrot_training - Step 46366: {'lr': 7.037992870813525e-06, 'samples': 23739904, 'steps': 46366, 'loss/train': 1.4175328016281128}
+02/26/2022 13:38:01 - INFO - codeparrot_training - Step 46367: {'lr': 7.034138254601391e-06, 'samples': 23740416, 'steps': 46367, 'loss/train': 1.720585823059082}
+02/26/2022 13:38:05 - INFO - codeparrot_training - Step 46368: {'lr': 7.030284679177851e-06, 'samples': 23740928, 'steps': 46368, 'loss/train': 2.760316848754883}
+02/26/2022 13:38:10 - INFO - codeparrot_training - Step 46369: {'lr': 7.0264321445594745e-06, 'samples': 23741440, 'steps': 46369, 'loss/train': 1.2613515853881836}
+02/26/2022 13:38:14 - INFO - codeparrot_training - Step 46370: {'lr': 7.022580650762667e-06, 'samples': 23741952, 'steps': 46370, 'loss/train': 1.168245553970337}
+02/26/2022 13:38:19 - INFO - codeparrot_training - Step 46371: {'lr': 7.0187301978039965e-06, 'samples': 23742464, 'steps': 46371, 'loss/train': 2.0983407497406006}
+02/26/2022 13:38:23 - INFO - codeparrot_training - Step 46372: {'lr': 7.014880785699895e-06, 'samples': 23742976, 'steps': 46372, 'loss/train': 1.5561175346374512}
+02/26/2022 13:38:28 - INFO - codeparrot_training - Step 46373: {'lr': 7.011032414466906e-06, 'samples': 23743488, 'steps': 46373, 'loss/train': 2.3228728771209717}
+02/26/2022 13:38:32 - INFO - codeparrot_training - Step 46374: {'lr': 7.007185084121487e-06, 'samples': 23744000, 'steps': 46374, 'loss/train': 1.7513600587844849}
+02/26/2022 13:38:37 - INFO - codeparrot_training - Step 46375: {'lr': 7.003338794680153e-06, 'samples': 23744512, 'steps': 46375, 'loss/train': 2.057473659515381}
+02/26/2022 13:38:41 - INFO - codeparrot_training - Step 46376: {'lr': 6.999493546159336e-06, 'samples': 23745024, 'steps': 46376, 'loss/train': 0.9103937745094299}
+02/26/2022 13:38:47 - INFO - codeparrot_training - Step 46377: {'lr': 6.995649338575521e-06, 'samples': 23745536, 'steps': 46377, 'loss/train': 1.9823440313339233}
+02/26/2022 13:38:51 - INFO - codeparrot_training - Step 46378: {'lr': 6.9918061719451975e-06, 'samples': 23746048, 'steps': 46378, 'loss/train': 1.6942082643508911}
+02/26/2022 13:38:56 - INFO - codeparrot_training - Step 46379: {'lr': 6.9879640462848225e-06, 'samples': 23746560, 'steps': 46379, 'loss/train': 1.7283331155776978}
+02/26/2022 13:38:59 - INFO - codeparrot_training - Step 46380: {'lr': 6.984122961610828e-06, 'samples': 23747072, 'steps': 46380, 'loss/train': 2.2948594093322754}
+02/26/2022 13:39:05 - INFO - codeparrot_training - Step 46381: {'lr': 6.9802829179396734e-06, 'samples': 23747584, 'steps': 46381, 'loss/train': 1.6163698434829712}
+02/26/2022 13:39:09 - INFO - codeparrot_training - Step 46382: {'lr': 6.976443915287817e-06, 'samples': 23748096, 'steps': 46382, 'loss/train': 3.2901971340179443}
+02/26/2022 13:39:14 - INFO - codeparrot_training - Step 46383: {'lr': 6.972605953671745e-06, 'samples': 23748608, 'steps': 46383, 'loss/train': 2.113461971282959}
+02/26/2022 13:39:20 - INFO - codeparrot_training - Step 46384: {'lr': 6.968769033107836e-06, 'samples': 23749120, 'steps': 46384, 'loss/train': 2.248030424118042}
+02/26/2022 13:39:23 - INFO - codeparrot_training - Step 46385: {'lr': 6.964933153612546e-06, 'samples': 23749632, 'steps': 46385, 'loss/train': 1.8545007705688477}
+02/26/2022 13:39:29 - INFO - codeparrot_training - Step 46386: {'lr': 6.961098315202308e-06, 'samples': 23750144, 'steps': 46386, 'loss/train': 1.8316305875778198}
+02/26/2022 13:39:33 - INFO - codeparrot_training - Step 46387: {'lr': 6.957264517893552e-06, 'samples': 23750656, 'steps': 46387, 'loss/train': 1.458087682723999}
+02/26/2022 13:39:38 - INFO - codeparrot_training - Step 46388: {'lr': 6.953431761702711e-06, 'samples': 23751168, 'steps': 46388, 'loss/train': 1.8730438947677612}
+02/26/2022 13:39:42 - INFO - codeparrot_training - Step 46389: {'lr': 6.949600046646187e-06, 'samples': 23751680, 'steps': 46389, 'loss/train': 1.6114581823349}
+02/26/2022 13:39:47 - INFO - codeparrot_training - Step 46390: {'lr': 6.945769372740413e-06, 'samples': 23752192, 'steps': 46390, 'loss/train': 1.8649907112121582}
+02/26/2022 13:39:51 - INFO - codeparrot_training - Step 46391: {'lr': 6.941939740001735e-06, 'samples': 23752704, 'steps': 46391, 'loss/train': 1.341841220855713}
+02/26/2022 13:39:56 - INFO - codeparrot_training - Step 46392: {'lr': 6.938111148446668e-06, 'samples': 23753216, 'steps': 46392, 'loss/train': 2.4057626724243164}
+02/26/2022 13:40:00 - INFO - codeparrot_training - Step 46393: {'lr': 6.9342835980915345e-06, 'samples': 23753728, 'steps': 46393, 'loss/train': 1.709973692893982}
+02/26/2022 13:40:05 - INFO - codeparrot_training - Step 46394: {'lr': 6.930457088952735e-06, 'samples': 23754240, 'steps': 46394, 'loss/train': 1.5739259719848633}
+02/26/2022 13:40:09 - INFO - codeparrot_training - Step 46395: {'lr': 6.926631621046647e-06, 'samples': 23754752, 'steps': 46395, 'loss/train': 1.7106016874313354}
+02/26/2022 13:40:14 - INFO - codeparrot_training - Step 46396: {'lr': 6.92280719438973e-06, 'samples': 23755264, 'steps': 46396, 'loss/train': 2.3995444774627686}
+02/26/2022 13:40:18 - INFO - codeparrot_training - Step 46397: {'lr': 6.918983808998331e-06, 'samples': 23755776, 'steps': 46397, 'loss/train': 0.45177268981933594}
+02/26/2022 13:40:23 - INFO - codeparrot_training - Step 46398: {'lr': 6.915161464888797e-06, 'samples': 23756288, 'steps': 46398, 'loss/train': 1.8366248607635498}
+02/26/2022 13:40:27 - INFO - codeparrot_training - Step 46399: {'lr': 6.911340162077506e-06, 'samples': 23756800, 'steps': 46399, 'loss/train': 1.68247652053833}
+02/26/2022 13:40:32 - INFO - codeparrot_training - Step 46400: {'lr': 6.907519900580861e-06, 'samples': 23757312, 'steps': 46400, 'loss/train': 0.4413408637046814}
+02/26/2022 13:40:36 - INFO - codeparrot_training - Step 46401: {'lr': 6.903700680415209e-06, 'samples': 23757824, 'steps': 46401, 'loss/train': 1.572052001953125}
+02/26/2022 13:40:42 - INFO - codeparrot_training - Step 46402: {'lr': 6.899882501596927e-06, 'samples': 23758336, 'steps': 46402, 'loss/train': 1.7348308563232422}
+02/26/2022 13:40:45 - INFO - codeparrot_training - Step 46403: {'lr': 6.896065364142307e-06, 'samples': 23758848, 'steps': 46403, 'loss/train': 0.9265848994255066}
+02/26/2022 13:40:51 - INFO - codeparrot_training - Step 46404: {'lr': 6.892249268067807e-06, 'samples': 23759360, 'steps': 46404, 'loss/train': 1.5321924686431885}
+02/26/2022 13:40:54 - INFO - codeparrot_training - Step 46405: {'lr': 6.888434213389694e-06, 'samples': 23759872, 'steps': 46405, 'loss/train': 1.8817692995071411}
+02/26/2022 13:41:00 - INFO - codeparrot_training - Step 46406: {'lr': 6.8846202001243145e-06, 'samples': 23760384, 'steps': 46406, 'loss/train': 1.6436309814453125}
+02/26/2022 13:41:03 - INFO - codeparrot_training - Step 46407: {'lr': 6.880807228288016e-06, 'samples': 23760896, 'steps': 46407, 'loss/train': 2.03767991065979}
+02/26/2022 13:41:09 - INFO - codeparrot_training - Step 46408: {'lr': 6.876995297897176e-06, 'samples': 23761408, 'steps': 46408, 'loss/train': 1.8102270364761353}
+02/26/2022 13:41:12 - INFO - codeparrot_training - Step 46409: {'lr': 6.873184408968059e-06, 'samples': 23761920, 'steps': 46409, 'loss/train': 1.194539189338684}
+02/26/2022 13:41:18 - INFO - codeparrot_training - Step 46410: {'lr': 6.869374561517066e-06, 'samples': 23762432, 'steps': 46410, 'loss/train': 0.08907842636108398}
+02/26/2022 13:41:21 - INFO - codeparrot_training - Step 46411: {'lr': 6.86556575556041e-06, 'samples': 23762944, 'steps': 46411, 'loss/train': 2.4222910404205322}
+02/26/2022 13:41:27 - INFO - codeparrot_training - Step 46412: {'lr': 6.86175799111452e-06, 'samples': 23763456, 'steps': 46412, 'loss/train': 2.30281925201416}
+02/26/2022 13:41:31 - INFO - codeparrot_training - Step 46413: {'lr': 6.857951268195605e-06, 'samples': 23763968, 'steps': 46413, 'loss/train': 1.2040692567825317}
+02/26/2022 13:41:36 - INFO - codeparrot_training - Step 46414: {'lr': 6.854145586820071e-06, 'samples': 23764480, 'steps': 46414, 'loss/train': 2.3244926929473877}
+02/26/2022 13:41:40 - INFO - codeparrot_training - Step 46415: {'lr': 6.850340947004124e-06, 'samples': 23764992, 'steps': 46415, 'loss/train': 1.1270174980163574}
+02/26/2022 13:41:45 - INFO - codeparrot_training - Step 46416: {'lr': 6.846537348764114e-06, 'samples': 23765504, 'steps': 46416, 'loss/train': 2.4419026374816895}
+02/26/2022 13:41:49 - INFO - codeparrot_training - Step 46417: {'lr': 6.842734792116334e-06, 'samples': 23766016, 'steps': 46417, 'loss/train': 1.7396389245986938}
+02/26/2022 13:41:54 - INFO - codeparrot_training - Step 46418: {'lr': 6.838933277077103e-06, 'samples': 23766528, 'steps': 46418, 'loss/train': 2.305845022201538}
+02/26/2022 13:41:58 - INFO - codeparrot_training - Step 46419: {'lr': 6.83513280366263e-06, 'samples': 23767040, 'steps': 46419, 'loss/train': 1.3567595481872559}
+02/26/2022 13:42:03 - INFO - codeparrot_training - Step 46420: {'lr': 6.8313333718892365e-06, 'samples': 23767552, 'steps': 46420, 'loss/train': 2.64897084236145}
+02/26/2022 13:42:07 - INFO - codeparrot_training - Step 46421: {'lr': 6.827534981773187e-06, 'samples': 23768064, 'steps': 46421, 'loss/train': 2.4512853622436523}
+02/26/2022 13:42:13 - INFO - codeparrot_training - Step 46422: {'lr': 6.823737633330801e-06, 'samples': 23768576, 'steps': 46422, 'loss/train': 0.6194876432418823}
+02/26/2022 13:42:16 - INFO - codeparrot_training - Step 46423: {'lr': 6.819941326578289e-06, 'samples': 23769088, 'steps': 46423, 'loss/train': 2.2124860286712646}
+02/26/2022 13:42:22 - INFO - codeparrot_training - Step 46424: {'lr': 6.816146061531914e-06, 'samples': 23769600, 'steps': 46424, 'loss/train': 2.097661256790161}
+02/26/2022 13:42:25 - INFO - codeparrot_training - Step 46425: {'lr': 6.812351838207942e-06, 'samples': 23770112, 'steps': 46425, 'loss/train': 0.35688576102256775}
+02/26/2022 13:42:31 - INFO - codeparrot_training - Step 46426: {'lr': 6.808558656622665e-06, 'samples': 23770624, 'steps': 46426, 'loss/train': 1.4441722631454468}
+02/26/2022 13:42:34 - INFO - codeparrot_training - Step 46427: {'lr': 6.804766516792293e-06, 'samples': 23771136, 'steps': 46427, 'loss/train': 1.9445942640304565}
+02/26/2022 13:42:40 - INFO - codeparrot_training - Step 46428: {'lr': 6.800975418733063e-06, 'samples': 23771648, 'steps': 46428, 'loss/train': 1.642012596130371}
+02/26/2022 13:42:43 - INFO - codeparrot_training - Step 46429: {'lr': 6.797185362461239e-06, 'samples': 23772160, 'steps': 46429, 'loss/train': 0.5557148456573486}
+02/26/2022 13:42:49 - INFO - codeparrot_training - Step 46430: {'lr': 6.7933963479930585e-06, 'samples': 23772672, 'steps': 46430, 'loss/train': 0.997287929058075}
+02/26/2022 13:42:52 - INFO - codeparrot_training - Step 46431: {'lr': 6.78960837534473e-06, 'samples': 23773184, 'steps': 46431, 'loss/train': 0.7129632234573364}
+02/26/2022 13:42:58 - INFO - codeparrot_training - Step 46432: {'lr': 6.785821444532492e-06, 'samples': 23773696, 'steps': 46432, 'loss/train': 1.0941447019577026}
+02/26/2022 13:43:02 - INFO - codeparrot_training - Step 46433: {'lr': 6.782035555572552e-06, 'samples': 23774208, 'steps': 46433, 'loss/train': 2.6145706176757812}
+02/26/2022 13:43:07 - INFO - codeparrot_training - Step 46434: {'lr': 6.778250708481148e-06, 'samples': 23774720, 'steps': 46434, 'loss/train': 1.8168832063674927}
+02/26/2022 13:43:11 - INFO - codeparrot_training - Step 46435: {'lr': 6.7744669032745175e-06, 'samples': 23775232, 'steps': 46435, 'loss/train': 2.020702600479126}
+02/26/2022 13:43:16 - INFO - codeparrot_training - Step 46436: {'lr': 6.770684139968814e-06, 'samples': 23775744, 'steps': 46436, 'loss/train': 1.5341795682907104}
+02/26/2022 13:43:20 - INFO - codeparrot_training - Step 46437: {'lr': 6.766902418580273e-06, 'samples': 23776256, 'steps': 46437, 'loss/train': 1.170082688331604}
+02/26/2022 13:43:25 - INFO - codeparrot_training - Step 46438: {'lr': 6.76312173912505e-06, 'samples': 23776768, 'steps': 46438, 'loss/train': 0.9335953593254089}
+02/26/2022 13:43:29 - INFO - codeparrot_training - Step 46439: {'lr': 6.759342101619409e-06, 'samples': 23777280, 'steps': 46439, 'loss/train': 2.61860990524292}
+02/26/2022 13:43:34 - INFO - codeparrot_training - Step 46440: {'lr': 6.755563506079504e-06, 'samples': 23777792, 'steps': 46440, 'loss/train': 1.5336909294128418}
+02/26/2022 13:43:38 - INFO - codeparrot_training - Step 46441: {'lr': 6.751785952521544e-06, 'samples': 23778304, 'steps': 46441, 'loss/train': 2.984496831893921}
+02/26/2022 13:43:43 - INFO - codeparrot_training - Step 46442: {'lr': 6.748009440961655e-06, 'samples': 23778816, 'steps': 46442, 'loss/train': 1.4492080211639404}
+02/26/2022 13:43:47 - INFO - codeparrot_training - Step 46443: {'lr': 6.744233971416075e-06, 'samples': 23779328, 'steps': 46443, 'loss/train': 1.7095447778701782}
+02/26/2022 13:43:52 - INFO - codeparrot_training - Step 46444: {'lr': 6.740459543900956e-06, 'samples': 23779840, 'steps': 46444, 'loss/train': 1.5660254955291748}
+02/26/2022 13:43:56 - INFO - codeparrot_training - Step 46445: {'lr': 6.736686158432481e-06, 'samples': 23780352, 'steps': 46445, 'loss/train': 1.6093417406082153}
+02/26/2022 13:44:02 - INFO - codeparrot_training - Step 46446: {'lr': 6.732913815026748e-06, 'samples': 23780864, 'steps': 46446, 'loss/train': 1.5200092792510986}
+02/26/2022 13:44:05 - INFO - codeparrot_training - Step 46447: {'lr': 6.72914251370002e-06, 'samples': 23781376, 'steps': 46447, 'loss/train': 3.6204516887664795}
+02/26/2022 13:44:11 - INFO - codeparrot_training - Step 46448: {'lr': 6.725372254468343e-06, 'samples': 23781888, 'steps': 46448, 'loss/train': 2.036477565765381}
+02/26/2022 13:44:17 - INFO - codeparrot_training - Step 46449: {'lr': 6.721603037348006e-06, 'samples': 23782400, 'steps': 46449, 'loss/train': 1.6362109184265137}
+02/26/2022 13:44:20 - INFO - codeparrot_training - Step 46450: {'lr': 6.717834862354999e-06, 'samples': 23782912, 'steps': 46450, 'loss/train': 1.953580617904663}
+02/26/2022 13:44:26 - INFO - codeparrot_training - Step 46451: {'lr': 6.7140677295055575e-06, 'samples': 23783424, 'steps': 46451, 'loss/train': 1.2113935947418213}
+02/26/2022 13:44:29 - INFO - codeparrot_training - Step 46452: {'lr': 6.71030163881578e-06, 'samples': 23783936, 'steps': 46452, 'loss/train': 1.748292326927185}
+02/26/2022 13:44:35 - INFO - codeparrot_training - Step 46453: {'lr': 6.706536590301876e-06, 'samples': 23784448, 'steps': 46453, 'loss/train': 1.4966521263122559}
+02/26/2022 13:44:38 - INFO - codeparrot_training - Step 46454: {'lr': 6.70277258397986e-06, 'samples': 23784960, 'steps': 46454, 'loss/train': 1.5227965116500854}
+02/26/2022 13:44:44 - INFO - codeparrot_training - Step 46455: {'lr': 6.699009619865914e-06, 'samples': 23785472, 'steps': 46455, 'loss/train': 1.4812519550323486}
+02/26/2022 13:44:47 - INFO - codeparrot_training - Step 46456: {'lr': 6.695247697976164e-06, 'samples': 23785984, 'steps': 46456, 'loss/train': 1.727266788482666}
+02/26/2022 13:44:51 - INFO - codeparrot_training - Step 46457: {'lr': 6.691486818326709e-06, 'samples': 23786496, 'steps': 46457, 'loss/train': 1.5667589902877808}
+02/26/2022 13:44:57 - INFO - codeparrot_training - Step 46458: {'lr': 6.687726980933673e-06, 'samples': 23787008, 'steps': 46458, 'loss/train': 1.1811867952346802}
+02/26/2022 13:45:00 - INFO - codeparrot_training - Step 46459: {'lr': 6.683968185813183e-06, 'samples': 23787520, 'steps': 46459, 'loss/train': 1.806470513343811}
+02/26/2022 13:45:06 - INFO - codeparrot_training - Step 46460: {'lr': 6.680210432981254e-06, 'samples': 23788032, 'steps': 46460, 'loss/train': 2.3082308769226074}
+02/26/2022 13:45:09 - INFO - codeparrot_training - Step 46461: {'lr': 6.676453722454068e-06, 'samples': 23788544, 'steps': 46461, 'loss/train': 2.1304683685302734}
+02/26/2022 13:45:15 - INFO - codeparrot_training - Step 46462: {'lr': 6.672698054247695e-06, 'samples': 23789056, 'steps': 46462, 'loss/train': 2.252291440963745}
+02/26/2022 13:45:20 - INFO - codeparrot_training - Step 46463: {'lr': 6.668943428378233e-06, 'samples': 23789568, 'steps': 46463, 'loss/train': 0.9569962024688721}
+02/26/2022 13:45:24 - INFO - codeparrot_training - Step 46464: {'lr': 6.665189844861724e-06, 'samples': 23790080, 'steps': 46464, 'loss/train': 1.9051259756088257}
+02/26/2022 13:45:29 - INFO - codeparrot_training - Step 46465: {'lr': 6.661437303714269e-06, 'samples': 23790592, 'steps': 46465, 'loss/train': 4.860052108764648}
+02/26/2022 13:45:33 - INFO - codeparrot_training - Step 46466: {'lr': 6.657685804951991e-06, 'samples': 23791104, 'steps': 46466, 'loss/train': 1.9033058881759644}
+02/26/2022 13:45:39 - INFO - codeparrot_training - Step 46467: {'lr': 6.653935348590878e-06, 'samples': 23791616, 'steps': 46467, 'loss/train': 2.4031310081481934}
+02/26/2022 13:45:43 - INFO - codeparrot_training - Step 46468: {'lr': 6.650185934647029e-06, 'samples': 23792128, 'steps': 46468, 'loss/train': 1.2548633813858032}
+02/26/2022 13:45:46 - INFO - codeparrot_training - Step 46469: {'lr': 6.646437563136543e-06, 'samples': 23792640, 'steps': 46469, 'loss/train': 1.8490608930587769}
+02/26/2022 13:45:52 - INFO - codeparrot_training - Step 46470: {'lr': 6.642690234075432e-06, 'samples': 23793152, 'steps': 46470, 'loss/train': 2.378690004348755}
+02/26/2022 13:45:55 - INFO - codeparrot_training - Step 46471: {'lr': 6.63894394747977e-06, 'samples': 23793664, 'steps': 46471, 'loss/train': 0.8607810735702515}
+02/26/2022 13:46:01 - INFO - codeparrot_training - Step 46472: {'lr': 6.635198703365569e-06, 'samples': 23794176, 'steps': 46472, 'loss/train': 2.0531651973724365}
+02/26/2022 13:46:04 - INFO - codeparrot_training - Step 46473: {'lr': 6.631454501748902e-06, 'samples': 23794688, 'steps': 46473, 'loss/train': 2.18971848487854}
+02/26/2022 13:46:10 - INFO - codeparrot_training - Step 46474: {'lr': 6.627711342645837e-06, 'samples': 23795200, 'steps': 46474, 'loss/train': 1.747352123260498}
+02/26/2022 13:46:13 - INFO - codeparrot_training - Step 46475: {'lr': 6.6239692260723624e-06, 'samples': 23795712, 'steps': 46475, 'loss/train': 1.4273356199264526}
+02/26/2022 13:46:19 - INFO - codeparrot_training - Step 46476: {'lr': 6.620228152044494e-06, 'samples': 23796224, 'steps': 46476, 'loss/train': 1.5860884189605713}
+02/26/2022 13:46:22 - INFO - codeparrot_training - Step 46477: {'lr': 6.616488120578329e-06, 'samples': 23796736, 'steps': 46477, 'loss/train': 1.7078667879104614}
+02/26/2022 13:46:28 - INFO - codeparrot_training - Step 46478: {'lr': 6.612749131689827e-06, 'samples': 23797248, 'steps': 46478, 'loss/train': 2.1136834621429443}
+02/26/2022 13:46:31 - INFO - codeparrot_training - Step 46479: {'lr': 6.609011185395031e-06, 'samples': 23797760, 'steps': 46479, 'loss/train': 3.049741506576538}
+02/26/2022 13:46:38 - INFO - codeparrot_training - Step 46480: {'lr': 6.605274281709927e-06, 'samples': 23798272, 'steps': 46480, 'loss/train': 1.4769575595855713}
+02/26/2022 13:46:41 - INFO - codeparrot_training - Step 46481: {'lr': 6.6015384206505605e-06, 'samples': 23798784, 'steps': 46481, 'loss/train': 1.2307194471359253}
+02/26/2022 13:46:47 - INFO - codeparrot_training - Step 46482: {'lr': 6.597803602232916e-06, 'samples': 23799296, 'steps': 46482, 'loss/train': 1.357944130897522}
+02/26/2022 13:46:50 - INFO - codeparrot_training - Step 46483: {'lr': 6.594069826472981e-06, 'samples': 23799808, 'steps': 46483, 'loss/train': 1.6806366443634033}
+02/26/2022 13:46:56 - INFO - codeparrot_training - Step 46484: {'lr': 6.590337093386772e-06, 'samples': 23800320, 'steps': 46484, 'loss/train': 1.6275931596755981}
+02/26/2022 13:46:59 - INFO - codeparrot_training - Step 46485: {'lr': 6.586605402990275e-06, 'samples': 23800832, 'steps': 46485, 'loss/train': 1.729817509651184}
+02/26/2022 13:47:05 - INFO - codeparrot_training - Step 46486: {'lr': 6.582874755299451e-06, 'samples': 23801344, 'steps': 46486, 'loss/train': 0.7806394100189209}
+02/26/2022 13:47:08 - INFO - codeparrot_training - Step 46487: {'lr': 6.579145150330284e-06, 'samples': 23801856, 'steps': 46487, 'loss/train': 0.62368243932724}
+02/26/2022 13:47:14 - INFO - codeparrot_training - Step 46488: {'lr': 6.57541658809882e-06, 'samples': 23802368, 'steps': 46488, 'loss/train': 1.6624560356140137}
+02/26/2022 13:47:17 - INFO - codeparrot_training - Step 46489: {'lr': 6.571689068620906e-06, 'samples': 23802880, 'steps': 46489, 'loss/train': 2.3215036392211914}
+02/26/2022 13:47:23 - INFO - codeparrot_training - Step 46490: {'lr': 6.567962591912613e-06, 'samples': 23803392, 'steps': 46490, 'loss/train': 2.1143174171447754}
+02/26/2022 13:47:26 - INFO - codeparrot_training - Step 46491: {'lr': 6.564237157989872e-06, 'samples': 23803904, 'steps': 46491, 'loss/train': 2.7534046173095703}
+02/26/2022 13:47:32 - INFO - codeparrot_training - Step 46492: {'lr': 6.5605127668686425e-06, 'samples': 23804416, 'steps': 46492, 'loss/train': 2.5225470066070557}
+02/26/2022 13:47:35 - INFO - codeparrot_training - Step 46493: {'lr': 6.556789418564857e-06, 'samples': 23804928, 'steps': 46493, 'loss/train': 1.5890918970108032}
+02/26/2022 13:47:41 - INFO - codeparrot_training - Step 46494: {'lr': 6.553067113094502e-06, 'samples': 23805440, 'steps': 46494, 'loss/train': 2.932908535003662}
+02/26/2022 13:47:45 - INFO - codeparrot_training - Step 46495: {'lr': 6.5493458504734815e-06, 'samples': 23805952, 'steps': 46495, 'loss/train': 1.5942814350128174}
+02/26/2022 13:47:51 - INFO - codeparrot_training - Step 46496: {'lr': 6.545625630717783e-06, 'samples': 23806464, 'steps': 46496, 'loss/train': 1.4911079406738281}
+02/26/2022 13:47:54 - INFO - codeparrot_training - Step 46497: {'lr': 6.541906453843311e-06, 'samples': 23806976, 'steps': 46497, 'loss/train': 0.730410099029541}
+02/26/2022 13:47:59 - INFO - codeparrot_training - Step 46498: {'lr': 6.5381883198660234e-06, 'samples': 23807488, 'steps': 46498, 'loss/train': 1.453140139579773}
+02/26/2022 13:48:03 - INFO - codeparrot_training - Step 46499: {'lr': 6.534471228801769e-06, 'samples': 23808000, 'steps': 46499, 'loss/train': 1.7993172407150269}
+02/26/2022 13:48:09 - INFO - codeparrot_training - Step 46500: {'lr': 6.530755180666592e-06, 'samples': 23808512, 'steps': 46500, 'loss/train': 0.24077600240707397}
+02/26/2022 13:48:12 - INFO - codeparrot_training - Step 46501: {'lr': 6.52704017547634e-06, 'samples': 23809024, 'steps': 46501, 'loss/train': 1.7888671159744263}
+02/26/2022 13:48:18 - INFO - codeparrot_training - Step 46502: {'lr': 6.523326213246916e-06, 'samples': 23809536, 'steps': 46502, 'loss/train': 0.8614785075187683}
+02/26/2022 13:48:21 - INFO - codeparrot_training - Step 46503: {'lr': 6.519613293994253e-06, 'samples': 23810048, 'steps': 46503, 'loss/train': 0.784758985042572}
+02/26/2022 13:48:27 - INFO - codeparrot_training - Step 46504: {'lr': 6.515901417734255e-06, 'samples': 23810560, 'steps': 46504, 'loss/train': 1.494071125984192}
+02/26/2022 13:48:31 - INFO - codeparrot_training - Step 46505: {'lr': 6.512190584482825e-06, 'samples': 23811072, 'steps': 46505, 'loss/train': 1.955809473991394}
+02/26/2022 13:48:36 - INFO - codeparrot_training - Step 46506: {'lr': 6.508480794255895e-06, 'samples': 23811584, 'steps': 46506, 'loss/train': 1.6849819421768188}
+02/26/2022 13:48:40 - INFO - codeparrot_training - Step 46507: {'lr': 6.5047720470692575e-06, 'samples': 23812096, 'steps': 46507, 'loss/train': 1.6729947328567505}
+02/26/2022 13:48:45 - INFO - codeparrot_training - Step 46508: {'lr': 6.5010643429388724e-06, 'samples': 23812608, 'steps': 46508, 'loss/train': 1.6640431880950928}
+02/26/2022 13:48:49 - INFO - codeparrot_training - Step 46509: {'lr': 6.497357681880589e-06, 'samples': 23813120, 'steps': 46509, 'loss/train': 1.933899998664856}
+02/26/2022 13:48:55 - INFO - codeparrot_training - Step 46510: {'lr': 6.493652063910366e-06, 'samples': 23813632, 'steps': 46510, 'loss/train': 2.2764484882354736}
+02/26/2022 13:48:58 - INFO - codeparrot_training - Step 46511: {'lr': 6.48994748904394e-06, 'samples': 23814144, 'steps': 46511, 'loss/train': 1.3437808752059937}
+02/26/2022 13:49:04 - INFO - codeparrot_training - Step 46512: {'lr': 6.4862439572973e-06, 'samples': 23814656, 'steps': 46512, 'loss/train': 2.2047863006591797}
+02/26/2022 13:49:07 - INFO - codeparrot_training - Step 46513: {'lr': 6.482541468686265e-06, 'samples': 23815168, 'steps': 46513, 'loss/train': 0.4005439281463623}
+02/26/2022 13:49:13 - INFO - codeparrot_training - Step 46514: {'lr': 6.478840023226712e-06, 'samples': 23815680, 'steps': 46514, 'loss/train': 1.5261503458023071}
+02/26/2022 13:49:16 - INFO - codeparrot_training - Step 46515: {'lr': 6.475139620934433e-06, 'samples': 23816192, 'steps': 46515, 'loss/train': 2.8783137798309326}
+02/26/2022 13:49:22 - INFO - codeparrot_training - Step 46516: {'lr': 6.471440261825362e-06, 'samples': 23816704, 'steps': 46516, 'loss/train': 1.5007215738296509}
+02/26/2022 13:49:26 - INFO - codeparrot_training - Step 46517: {'lr': 6.467741945915289e-06, 'samples': 23817216, 'steps': 46517, 'loss/train': 1.7192862033843994}
+02/26/2022 13:49:31 - INFO - codeparrot_training - Step 46518: {'lr': 6.464044673220121e-06, 'samples': 23817728, 'steps': 46518, 'loss/train': 1.9812179803848267}
+02/26/2022 13:49:35 - INFO - codeparrot_training - Step 46519: {'lr': 6.460348443755621e-06, 'samples': 23818240, 'steps': 46519, 'loss/train': 1.9676893949508667}
+02/26/2022 13:49:40 - INFO - codeparrot_training - Step 46520: {'lr': 6.456653257537665e-06, 'samples': 23818752, 'steps': 46520, 'loss/train': 1.057741641998291}
+02/26/2022 13:49:44 - INFO - codeparrot_training - Step 46521: {'lr': 6.452959114582102e-06, 'samples': 23819264, 'steps': 46521, 'loss/train': 1.8243581056594849}
+02/26/2022 13:49:49 - INFO - codeparrot_training - Step 46522: {'lr': 6.44926601490467e-06, 'samples': 23819776, 'steps': 46522, 'loss/train': 2.279841184616089}
+02/26/2022 13:49:53 - INFO - codeparrot_training - Step 46523: {'lr': 6.445573958521273e-06, 'samples': 23820288, 'steps': 46523, 'loss/train': 1.6697983741760254}
+02/26/2022 13:49:58 - INFO - codeparrot_training - Step 46524: {'lr': 6.441882945447702e-06, 'samples': 23820800, 'steps': 46524, 'loss/train': 2.1427297592163086}
+02/26/2022 13:50:02 - INFO - codeparrot_training - Step 46525: {'lr': 6.438192975699781e-06, 'samples': 23821312, 'steps': 46525, 'loss/train': 1.2807435989379883}
+02/26/2022 13:50:07 - INFO - codeparrot_training - Step 46526: {'lr': 6.434504049293271e-06, 'samples': 23821824, 'steps': 46526, 'loss/train': 1.1216572523117065}
+02/26/2022 13:50:11 - INFO - codeparrot_training - Step 46527: {'lr': 6.430816166244024e-06, 'samples': 23822336, 'steps': 46527, 'loss/train': 1.4811056852340698}
+02/26/2022 13:50:18 - INFO - codeparrot_training - Step 46528: {'lr': 6.427129326567804e-06, 'samples': 23822848, 'steps': 46528, 'loss/train': 1.458317756652832}
+02/26/2022 13:50:22 - INFO - codeparrot_training - Step 46529: {'lr': 6.42344353028046e-06, 'samples': 23823360, 'steps': 46529, 'loss/train': 0.5162584185600281}
+02/26/2022 13:50:27 - INFO - codeparrot_training - Step 46530: {'lr': 6.419758777397672e-06, 'samples': 23823872, 'steps': 46530, 'loss/train': 1.766250729560852}
+02/26/2022 13:50:31 - INFO - codeparrot_training - Step 46531: {'lr': 6.416075067935345e-06, 'samples': 23824384, 'steps': 46531, 'loss/train': 1.853021264076233}
+02/26/2022 13:50:36 - INFO - codeparrot_training - Step 46532: {'lr': 6.4123924019091896e-06, 'samples': 23824896, 'steps': 46532, 'loss/train': 2.168959856033325}
+02/26/2022 13:50:40 - INFO - codeparrot_training - Step 46533: {'lr': 6.408710779334998e-06, 'samples': 23825408, 'steps': 46533, 'loss/train': 1.7554508447647095}
+02/26/2022 13:50:45 - INFO - codeparrot_training - Step 46534: {'lr': 6.4050302002285345e-06, 'samples': 23825920, 'steps': 46534, 'loss/train': 1.4360580444335938}
+02/26/2022 13:50:49 - INFO - codeparrot_training - Step 46535: {'lr': 6.401350664605565e-06, 'samples': 23826432, 'steps': 46535, 'loss/train': 1.4463597536087036}
+02/26/2022 13:50:54 - INFO - codeparrot_training - Step 46536: {'lr': 6.397672172481883e-06, 'samples': 23826944, 'steps': 46536, 'loss/train': 3.1432900428771973}
+02/26/2022 13:50:58 - INFO - codeparrot_training - Step 46537: {'lr': 6.3939947238732255e-06, 'samples': 23827456, 'steps': 46537, 'loss/train': 2.0650248527526855}
+02/26/2022 13:51:05 - INFO - codeparrot_training - Step 46538: {'lr': 6.3903183187953015e-06, 'samples': 23827968, 'steps': 46538, 'loss/train': 1.4216917753219604}
+02/26/2022 13:51:08 - INFO - codeparrot_training - Step 46539: {'lr': 6.386642957263905e-06, 'samples': 23828480, 'steps': 46539, 'loss/train': 1.4810951948165894}
+02/26/2022 13:51:14 - INFO - codeparrot_training - Step 46540: {'lr': 6.3829686392948005e-06, 'samples': 23828992, 'steps': 46540, 'loss/train': 3.243936777114868}
+02/26/2022 13:51:18 - INFO - codeparrot_training - Step 46541: {'lr': 6.3792953649036975e-06, 'samples': 23829504, 'steps': 46541, 'loss/train': 1.8178826570510864}
+02/26/2022 13:51:23 - INFO - codeparrot_training - Step 46542: {'lr': 6.375623134106307e-06, 'samples': 23830016, 'steps': 46542, 'loss/train': 2.0278468132019043}
+02/26/2022 13:51:27 - INFO - codeparrot_training - Step 46543: {'lr': 6.371951946918419e-06, 'samples': 23830528, 'steps': 46543, 'loss/train': 1.3376773595809937}
+02/26/2022 13:51:32 - INFO - codeparrot_training - Step 46544: {'lr': 6.368281803355691e-06, 'samples': 23831040, 'steps': 46544, 'loss/train': 0.41307809948921204}
+02/26/2022 13:51:36 - INFO - codeparrot_training - Step 46545: {'lr': 6.364612703433942e-06, 'samples': 23831552, 'steps': 46545, 'loss/train': 1.3543895483016968}
+02/26/2022 13:51:41 - INFO - codeparrot_training - Step 46546: {'lr': 6.360944647168798e-06, 'samples': 23832064, 'steps': 46546, 'loss/train': 0.6874322891235352}
+02/26/2022 13:51:45 - INFO - codeparrot_training - Step 46547: {'lr': 6.357277634575998e-06, 'samples': 23832576, 'steps': 46547, 'loss/train': 0.3340718150138855}
+02/26/2022 13:51:52 - INFO - codeparrot_training - Step 46548: {'lr': 6.35361166567125e-06, 'samples': 23833088, 'steps': 46548, 'loss/train': 1.3821195363998413}
+02/26/2022 13:51:56 - INFO - codeparrot_training - Step 46549: {'lr': 6.349946740470292e-06, 'samples': 23833600, 'steps': 46549, 'loss/train': 2.128309965133667}
+02/26/2022 13:52:01 - INFO - codeparrot_training - Step 46550: {'lr': 6.34628285898875e-06, 'samples': 23834112, 'steps': 46550, 'loss/train': 1.1764681339263916}
+02/26/2022 13:52:05 - INFO - codeparrot_training - Step 46551: {'lr': 6.34262002124239e-06, 'samples': 23834624, 'steps': 46551, 'loss/train': 1.3445438146591187}
+02/26/2022 13:52:10 - INFO - codeparrot_training - Step 46552: {'lr': 6.338958227246866e-06, 'samples': 23835136, 'steps': 46552, 'loss/train': 1.153898000717163}
+02/26/2022 13:52:14 - INFO - codeparrot_training - Step 46553: {'lr': 6.335297477017887e-06, 'samples': 23835648, 'steps': 46553, 'loss/train': 1.9815213680267334}
+02/26/2022 13:52:19 - INFO - codeparrot_training - Step 46554: {'lr': 6.331637770571108e-06, 'samples': 23836160, 'steps': 46554, 'loss/train': 1.7304085493087769}
+02/26/2022 13:52:23 - INFO - codeparrot_training - Step 46555: {'lr': 6.327979107922238e-06, 'samples': 23836672, 'steps': 46555, 'loss/train': 2.0602192878723145}
+02/26/2022 13:52:28 - INFO - codeparrot_training - Step 46556: {'lr': 6.324321489086904e-06, 'samples': 23837184, 'steps': 46556, 'loss/train': 1.6734329462051392}
+02/26/2022 13:52:32 - INFO - codeparrot_training - Step 46557: {'lr': 6.3206649140808145e-06, 'samples': 23837696, 'steps': 46557, 'loss/train': 1.937697410583496}
+02/26/2022 13:52:39 - INFO - codeparrot_training - Step 46558: {'lr': 6.317009382919625e-06, 'samples': 23838208, 'steps': 46558, 'loss/train': 1.1814743280410767}
+02/26/2022 13:52:42 - INFO - codeparrot_training - Step 46559: {'lr': 6.313354895618989e-06, 'samples': 23838720, 'steps': 46559, 'loss/train': 1.5727505683898926}
+02/26/2022 13:52:48 - INFO - codeparrot_training - Step 46560: {'lr': 6.309701452194561e-06, 'samples': 23839232, 'steps': 46560, 'loss/train': 1.6556941270828247}
+02/26/2022 13:52:51 - INFO - codeparrot_training - Step 46561: {'lr': 6.306049052661994e-06, 'samples': 23839744, 'steps': 46561, 'loss/train': 1.7878711223602295}
+02/26/2022 13:52:57 - INFO - codeparrot_training - Step 46562: {'lr': 6.302397697036916e-06, 'samples': 23840256, 'steps': 46562, 'loss/train': 2.382051944732666}
+02/26/2022 13:53:00 - INFO - codeparrot_training - Step 46563: {'lr': 6.298747385335007e-06, 'samples': 23840768, 'steps': 46563, 'loss/train': 1.2224739789962769}
+02/26/2022 13:53:06 - INFO - codeparrot_training - Step 46564: {'lr': 6.29509811757184e-06, 'samples': 23841280, 'steps': 46564, 'loss/train': 1.1466196775436401}
+02/26/2022 13:53:11 - INFO - codeparrot_training - Step 46565: {'lr': 6.291449893763123e-06, 'samples': 23841792, 'steps': 46565, 'loss/train': 1.770437479019165}
+02/26/2022 13:53:15 - INFO - codeparrot_training - Step 46566: {'lr': 6.287802713924456e-06, 'samples': 23842304, 'steps': 46566, 'loss/train': 1.4644834995269775}
+02/26/2022 13:53:20 - INFO - codeparrot_training - Step 46567: {'lr': 6.284156578071437e-06, 'samples': 23842816, 'steps': 46567, 'loss/train': 1.7263942956924438}
+02/26/2022 13:53:24 - INFO - codeparrot_training - Step 46568: {'lr': 6.28051148621972e-06, 'samples': 23843328, 'steps': 46568, 'loss/train': 1.174325942993164}
+02/26/2022 13:53:29 - INFO - codeparrot_training - Step 46569: {'lr': 6.2768674383848755e-06, 'samples': 23843840, 'steps': 46569, 'loss/train': 1.6952998638153076}
+02/26/2022 13:53:33 - INFO - codeparrot_training - Step 46570: {'lr': 6.273224434582559e-06, 'samples': 23844352, 'steps': 46570, 'loss/train': 0.7334764003753662}
+02/26/2022 13:53:38 - INFO - codeparrot_training - Step 46571: {'lr': 6.26958247482834e-06, 'samples': 23844864, 'steps': 46571, 'loss/train': 1.1163926124572754}
+02/26/2022 13:53:42 - INFO - codeparrot_training - Step 46572: {'lr': 6.2659415591378445e-06, 'samples': 23845376, 'steps': 46572, 'loss/train': 2.25423264503479}
+02/26/2022 13:53:49 - INFO - codeparrot_training - Step 46573: {'lr': 6.2623016875266445e-06, 'samples': 23845888, 'steps': 46573, 'loss/train': 1.303412914276123}
+02/26/2022 13:53:53 - INFO - codeparrot_training - Step 46574: {'lr': 6.258662860010394e-06, 'samples': 23846400, 'steps': 46574, 'loss/train': 2.097568988800049}
+02/26/2022 13:53:58 - INFO - codeparrot_training - Step 46575: {'lr': 6.255025076604609e-06, 'samples': 23846912, 'steps': 46575, 'loss/train': 1.623725175857544}
+02/26/2022 13:54:02 - INFO - codeparrot_training - Step 46576: {'lr': 6.251388337324887e-06, 'samples': 23847424, 'steps': 46576, 'loss/train': 0.6567850112915039}
+02/26/2022 13:54:05 - INFO - codeparrot_training - Step 46577: {'lr': 6.247752642186827e-06, 'samples': 23847936, 'steps': 46577, 'loss/train': 2.17887806892395}
+02/26/2022 13:54:11 - INFO - codeparrot_training - Step 46578: {'lr': 6.244117991206e-06, 'samples': 23848448, 'steps': 46578, 'loss/train': 1.637640357017517}
+02/26/2022 13:54:14 - INFO - codeparrot_training - Step 46579: {'lr': 6.240484384397949e-06, 'samples': 23848960, 'steps': 46579, 'loss/train': 1.7565633058547974}
+02/26/2022 13:54:20 - INFO - codeparrot_training - Step 46580: {'lr': 6.2368518217783e-06, 'samples': 23849472, 'steps': 46580, 'loss/train': 2.4432973861694336}
+02/26/2022 13:54:23 - INFO - codeparrot_training - Step 46581: {'lr': 6.2332203033625415e-06, 'samples': 23849984, 'steps': 46581, 'loss/train': 0.6739732027053833}
+02/26/2022 13:54:29 - INFO - codeparrot_training - Step 46582: {'lr': 6.229589829166271e-06, 'samples': 23850496, 'steps': 46582, 'loss/train': 2.023789167404175}
+02/26/2022 13:54:32 - INFO - codeparrot_training - Step 46583: {'lr': 6.225960399205033e-06, 'samples': 23851008, 'steps': 46583, 'loss/train': 0.9450035691261292}
+02/26/2022 13:54:40 - INFO - codeparrot_training - Step 46584: {'lr': 6.2223320134943976e-06, 'samples': 23851520, 'steps': 46584, 'loss/train': 1.9493979215621948}
+02/26/2022 13:54:43 - INFO - codeparrot_training - Step 46585: {'lr': 6.2187046720498516e-06, 'samples': 23852032, 'steps': 46585, 'loss/train': 2.1133739948272705}
+02/26/2022 13:54:49 - INFO - codeparrot_training - Step 46586: {'lr': 6.215078374886968e-06, 'samples': 23852544, 'steps': 46586, 'loss/train': 2.5516552925109863}
+02/26/2022 13:54:52 - INFO - codeparrot_training - Step 46587: {'lr': 6.21145312202126e-06, 'samples': 23853056, 'steps': 46587, 'loss/train': 0.6385740041732788}
+02/26/2022 13:54:58 - INFO - codeparrot_training - Step 46588: {'lr': 6.207828913468355e-06, 'samples': 23853568, 'steps': 46588, 'loss/train': 1.7570998668670654}
+02/26/2022 13:55:01 - INFO - codeparrot_training - Step 46589: {'lr': 6.204205749243602e-06, 'samples': 23854080, 'steps': 46589, 'loss/train': 0.6544342637062073}
+02/26/2022 13:55:07 - INFO - codeparrot_training - Step 46590: {'lr': 6.200583629362683e-06, 'samples': 23854592, 'steps': 46590, 'loss/train': 2.156109571456909}
+02/26/2022 13:55:10 - INFO - codeparrot_training - Step 46591: {'lr': 6.196962553841001e-06, 'samples': 23855104, 'steps': 46591, 'loss/train': 0.5243734121322632}
+02/26/2022 13:55:16 - INFO - codeparrot_training - Step 46592: {'lr': 6.1933425226941566e-06, 'samples': 23855616, 'steps': 46592, 'loss/train': 1.1431992053985596}
+02/26/2022 13:55:20 - INFO - codeparrot_training - Step 46593: {'lr': 6.1897235359376084e-06, 'samples': 23856128, 'steps': 46593, 'loss/train': 0.17018328607082367}
+02/26/2022 13:55:27 - INFO - codeparrot_training - Step 46594: {'lr': 6.186105593586844e-06, 'samples': 23856640, 'steps': 46594, 'loss/train': 2.735783338546753}
+02/26/2022 13:55:31 - INFO - codeparrot_training - Step 46595: {'lr': 6.182488695657379e-06, 'samples': 23857152, 'steps': 46595, 'loss/train': 2.716249465942383}
+02/26/2022 13:55:36 - INFO - codeparrot_training - Step 46596: {'lr': 6.178872842164729e-06, 'samples': 23857664, 'steps': 46596, 'loss/train': 0.9551397562026978}
+02/26/2022 13:55:39 - INFO - codeparrot_training - Step 46597: {'lr': 6.175258033124381e-06, 'samples': 23858176, 'steps': 46597, 'loss/train': 1.4555304050445557}
+02/26/2022 13:55:45 - INFO - codeparrot_training - Step 46598: {'lr': 6.171644268551796e-06, 'samples': 23858688, 'steps': 46598, 'loss/train': 1.7013649940490723}
+02/26/2022 13:55:48 - INFO - codeparrot_training - Step 46599: {'lr': 6.168031548462433e-06, 'samples': 23859200, 'steps': 46599, 'loss/train': 1.5595775842666626}
+02/26/2022 13:55:54 - INFO - codeparrot_training - Step 46600: {'lr': 6.164419872871835e-06, 'samples': 23859712, 'steps': 46600, 'loss/train': 1.9842448234558105}
+02/26/2022 13:55:57 - INFO - codeparrot_training - Step 46601: {'lr': 6.1608092417954346e-06, 'samples': 23860224, 'steps': 46601, 'loss/train': 1.2747728824615479}
+02/26/2022 13:56:03 - INFO - codeparrot_training - Step 46602: {'lr': 6.157199655248691e-06, 'samples': 23860736, 'steps': 46602, 'loss/train': 1.8581600189208984}
+02/26/2022 13:56:06 - INFO - codeparrot_training - Step 46603: {'lr': 6.153591113247036e-06, 'samples': 23861248, 'steps': 46603, 'loss/train': 1.6494197845458984}
+02/26/2022 13:56:12 - INFO - codeparrot_training - Step 46604: {'lr': 6.149983615806015e-06, 'samples': 23861760, 'steps': 46604, 'loss/train': 1.5091625452041626}
+02/26/2022 13:56:15 - INFO - codeparrot_training - Step 46605: {'lr': 6.14637716294103e-06, 'samples': 23862272, 'steps': 46605, 'loss/train': 0.6269817352294922}
+02/26/2022 13:56:23 - INFO - codeparrot_training - Step 46606: {'lr': 6.142771754667514e-06, 'samples': 23862784, 'steps': 46606, 'loss/train': 1.720922827720642}
+02/26/2022 13:56:26 - INFO - codeparrot_training - Step 46607: {'lr': 6.139167391000927e-06, 'samples': 23863296, 'steps': 46607, 'loss/train': 1.9761912822723389}
+02/26/2022 13:56:32 - INFO - codeparrot_training - Step 46608: {'lr': 6.135564071956729e-06, 'samples': 23863808, 'steps': 46608, 'loss/train': 1.5667837858200073}
+02/26/2022 13:56:35 - INFO - codeparrot_training - Step 46609: {'lr': 6.1319617975503505e-06, 'samples': 23864320, 'steps': 46609, 'loss/train': 1.2838783264160156}
+02/26/2022 13:56:41 - INFO - codeparrot_training - Step 46610: {'lr': 6.128360567797198e-06, 'samples': 23864832, 'steps': 46610, 'loss/train': 1.78585684299469}
+02/26/2022 13:56:44 - INFO - codeparrot_training - Step 46611: {'lr': 6.124760382712674e-06, 'samples': 23865344, 'steps': 46611, 'loss/train': 1.6675465106964111}
+02/26/2022 13:56:50 - INFO - codeparrot_training - Step 46612: {'lr': 6.121161242312268e-06, 'samples': 23865856, 'steps': 46612, 'loss/train': 2.134647846221924}
+02/26/2022 13:56:53 - INFO - codeparrot_training - Step 46613: {'lr': 6.117563146611382e-06, 'samples': 23866368, 'steps': 46613, 'loss/train': 2.284396171569824}
+02/26/2022 13:56:59 - INFO - codeparrot_training - Step 46614: {'lr': 6.113966095625395e-06, 'samples': 23866880, 'steps': 46614, 'loss/train': 1.4648711681365967}
+02/26/2022 13:57:02 - INFO - codeparrot_training - Step 46615: {'lr': 6.110370089369766e-06, 'samples': 23867392, 'steps': 46615, 'loss/train': 2.062326431274414}
+02/26/2022 13:57:08 - INFO - codeparrot_training - Step 46616: {'lr': 6.106775127859815e-06, 'samples': 23867904, 'steps': 46616, 'loss/train': 2.564812421798706}
+02/26/2022 13:57:11 - INFO - codeparrot_training - Step 46617: {'lr': 6.103181211111031e-06, 'samples': 23868416, 'steps': 46617, 'loss/train': 1.0981943607330322}
+02/26/2022 13:57:16 - INFO - codeparrot_training - Step 46618: {'lr': 6.099588339138762e-06, 'samples': 23868928, 'steps': 46618, 'loss/train': 1.1418571472167969}
+02/26/2022 13:57:20 - INFO - codeparrot_training - Step 46619: {'lr': 6.09599651195844e-06, 'samples': 23869440, 'steps': 46619, 'loss/train': 1.1658334732055664}
+02/26/2022 13:57:27 - INFO - codeparrot_training - Step 46620: {'lr': 6.092405729585359e-06, 'samples': 23869952, 'steps': 46620, 'loss/train': 1.969346046447754}
+02/26/2022 13:57:30 - INFO - codeparrot_training - Step 46621: {'lr': 6.0888159920350075e-06, 'samples': 23870464, 'steps': 46621, 'loss/train': 2.5843257904052734}
+02/26/2022 13:57:36 - INFO - codeparrot_training - Step 46622: {'lr': 6.085227299322676e-06, 'samples': 23870976, 'steps': 46622, 'loss/train': 1.7043955326080322}
+02/26/2022 13:57:39 - INFO - codeparrot_training - Step 46623: {'lr': 6.081639651463855e-06, 'samples': 23871488, 'steps': 46623, 'loss/train': 1.9177359342575073}
+02/26/2022 13:57:45 - INFO - codeparrot_training - Step 46624: {'lr': 6.0780530484737805e-06, 'samples': 23872000, 'steps': 46624, 'loss/train': 3.2926406860351562}
+02/26/2022 13:57:48 - INFO - codeparrot_training - Step 46625: {'lr': 6.0744674903678855e-06, 'samples': 23872512, 'steps': 46625, 'loss/train': 1.9187465906143188}
+02/26/2022 13:57:54 - INFO - codeparrot_training - Step 46626: {'lr': 6.07088297716149e-06, 'samples': 23873024, 'steps': 46626, 'loss/train': 1.8744899034500122}
+02/26/2022 13:57:57 - INFO - codeparrot_training - Step 46627: {'lr': 6.06729950887e-06, 'samples': 23873536, 'steps': 46627, 'loss/train': 1.0220211744308472}
+02/26/2022 13:58:03 - INFO - codeparrot_training - Step 46628: {'lr': 6.063717085508763e-06, 'samples': 23874048, 'steps': 46628, 'loss/train': 1.101572036743164}
+02/26/2022 13:58:06 - INFO - codeparrot_training - Step 46629: {'lr': 6.060135707093073e-06, 'samples': 23874560, 'steps': 46629, 'loss/train': 1.3812395334243774}
+02/26/2022 13:58:13 - INFO - codeparrot_training - Step 46630: {'lr': 6.056555373638306e-06, 'samples': 23875072, 'steps': 46630, 'loss/train': 1.8080193996429443}
+02/26/2022 13:58:17 - INFO - codeparrot_training - Step 46631: {'lr': 6.052976085159783e-06, 'samples': 23875584, 'steps': 46631, 'loss/train': 0.9176732301712036}
+02/26/2022 13:58:22 - INFO - codeparrot_training - Step 46632: {'lr': 6.049397841672882e-06, 'samples': 23876096, 'steps': 46632, 'loss/train': 1.2067385911941528}
+02/26/2022 13:58:26 - INFO - codeparrot_training - Step 46633: {'lr': 6.045820643192895e-06, 'samples': 23876608, 'steps': 46633, 'loss/train': 1.5077482461929321}
+02/26/2022 13:58:32 - INFO - codeparrot_training - Step 46634: {'lr': 6.042244489735116e-06, 'samples': 23877120, 'steps': 46634, 'loss/train': 2.1659486293792725}
+02/26/2022 13:58:35 - INFO - codeparrot_training - Step 46635: {'lr': 6.038669381314921e-06, 'samples': 23877632, 'steps': 46635, 'loss/train': 1.721977710723877}
+02/26/2022 13:58:41 - INFO - codeparrot_training - Step 46636: {'lr': 6.035095317947603e-06, 'samples': 23878144, 'steps': 46636, 'loss/train': 1.7914661169052124}
+02/26/2022 13:58:44 - INFO - codeparrot_training - Step 46637: {'lr': 6.0315222996484565e-06, 'samples': 23878656, 'steps': 46637, 'loss/train': 1.6718448400497437}
+02/26/2022 13:58:50 - INFO - codeparrot_training - Step 46638: {'lr': 6.027950326432774e-06, 'samples': 23879168, 'steps': 46638, 'loss/train': 1.4683914184570312}
+02/26/2022 13:58:53 - INFO - codeparrot_training - Step 46639: {'lr': 6.024379398315933e-06, 'samples': 23879680, 'steps': 46639, 'loss/train': 0.6901495456695557}
+02/26/2022 13:59:01 - INFO - codeparrot_training - Step 46640: {'lr': 6.020809515313141e-06, 'samples': 23880192, 'steps': 46640, 'loss/train': 1.9465771913528442}
+02/26/2022 13:59:04 - INFO - codeparrot_training - Step 46641: {'lr': 6.0172406774397785e-06, 'samples': 23880704, 'steps': 46641, 'loss/train': 1.105635643005371}
+02/26/2022 13:59:10 - INFO - codeparrot_training - Step 46642: {'lr': 6.013672884711025e-06, 'samples': 23881216, 'steps': 46642, 'loss/train': 1.9301466941833496}
+02/26/2022 13:59:13 - INFO - codeparrot_training - Step 46643: {'lr': 6.0101061371422575e-06, 'samples': 23881728, 'steps': 46643, 'loss/train': 1.4662811756134033}
+02/26/2022 13:59:18 - INFO - codeparrot_training - Step 46644: {'lr': 6.0065404347486864e-06, 'samples': 23882240, 'steps': 46644, 'loss/train': 1.681962251663208}
+02/26/2022 13:59:22 - INFO - codeparrot_training - Step 46645: {'lr': 6.002975777545689e-06, 'samples': 23882752, 'steps': 46645, 'loss/train': 1.3840441703796387}
+02/26/2022 13:59:28 - INFO - codeparrot_training - Step 46646: {'lr': 5.99941216554839e-06, 'samples': 23883264, 'steps': 46646, 'loss/train': 2.5264551639556885}
+02/26/2022 13:59:31 - INFO - codeparrot_training - Step 46647: {'lr': 5.995849598772169e-06, 'samples': 23883776, 'steps': 46647, 'loss/train': 1.7247811555862427}
+02/26/2022 13:59:37 - INFO - codeparrot_training - Step 46648: {'lr': 5.992288077232261e-06, 'samples': 23884288, 'steps': 46648, 'loss/train': 1.6359807252883911}
+02/26/2022 13:59:40 - INFO - codeparrot_training - Step 46649: {'lr': 5.988727600943878e-06, 'samples': 23884800, 'steps': 46649, 'loss/train': 2.1761908531188965}
+02/26/2022 13:59:46 - INFO - codeparrot_training - Step 46650: {'lr': 5.985168169922311e-06, 'samples': 23885312, 'steps': 46650, 'loss/train': 2.2371673583984375}
+02/26/2022 13:59:49 - INFO - codeparrot_training - Step 46651: {'lr': 5.981609784182801e-06, 'samples': 23885824, 'steps': 46651, 'loss/train': 1.9049779176712036}
+02/26/2022 13:59:56 - INFO - codeparrot_training - Step 46652: {'lr': 5.978052443740584e-06, 'samples': 23886336, 'steps': 46652, 'loss/train': 1.652034878730774}
+02/26/2022 14:00:00 - INFO - codeparrot_training - Step 46653: {'lr': 5.974496148610897e-06, 'samples': 23886848, 'steps': 46653, 'loss/train': 2.0710084438323975}
+02/26/2022 14:00:05 - INFO - codeparrot_training - Step 46654: {'lr': 5.970940898809008e-06, 'samples': 23887360, 'steps': 46654, 'loss/train': 1.9836301803588867}
+02/26/2022 14:00:09 - INFO - codeparrot_training - Step 46655: {'lr': 5.9673866943501245e-06, 'samples': 23887872, 'steps': 46655, 'loss/train': 1.204217791557312}
+02/26/2022 14:00:14 - INFO - codeparrot_training - Step 46656: {'lr': 5.963833535249458e-06, 'samples': 23888384, 'steps': 46656, 'loss/train': 1.724589228630066}
+02/26/2022 14:00:18 - INFO - codeparrot_training - Step 46657: {'lr': 5.960281421522218e-06, 'samples': 23888896, 'steps': 46657, 'loss/train': 2.331415891647339}
+02/26/2022 14:00:23 - INFO - codeparrot_training - Step 46658: {'lr': 5.95673035318367e-06, 'samples': 23889408, 'steps': 46658, 'loss/train': 0.16043223440647125}
+02/26/2022 14:00:27 - INFO - codeparrot_training - Step 46659: {'lr': 5.95318033024897e-06, 'samples': 23889920, 'steps': 46659, 'loss/train': 3.100290060043335}
+02/26/2022 14:00:32 - INFO - codeparrot_training - Step 46660: {'lr': 5.9496313527333814e-06, 'samples': 23890432, 'steps': 46660, 'loss/train': 0.701377272605896}
+02/26/2022 14:00:36 - INFO - codeparrot_training - Step 46661: {'lr': 5.946083420652032e-06, 'samples': 23890944, 'steps': 46661, 'loss/train': 1.7472103834152222}
+02/26/2022 14:00:41 - INFO - codeparrot_training - Step 46662: {'lr': 5.942536534020215e-06, 'samples': 23891456, 'steps': 46662, 'loss/train': 1.168807864189148}
+02/26/2022 14:00:45 - INFO - codeparrot_training - Step 46663: {'lr': 5.938990692853058e-06, 'samples': 23891968, 'steps': 46663, 'loss/train': 3.1657886505126953}
+02/26/2022 14:00:50 - INFO - codeparrot_training - Step 46664: {'lr': 5.93544589716577e-06, 'samples': 23892480, 'steps': 46664, 'loss/train': 2.089587688446045}
+02/26/2022 14:00:54 - INFO - codeparrot_training - Step 46665: {'lr': 5.931902146973506e-06, 'samples': 23892992, 'steps': 46665, 'loss/train': 1.3023558855056763}
+02/26/2022 14:00:59 - INFO - codeparrot_training - Step 46666: {'lr': 5.928359442291503e-06, 'samples': 23893504, 'steps': 46666, 'loss/train': 1.4478703737258911}
+02/26/2022 14:01:03 - INFO - codeparrot_training - Step 46667: {'lr': 5.924817783134889e-06, 'samples': 23894016, 'steps': 46667, 'loss/train': 1.4270223379135132}
+02/26/2022 14:01:10 - INFO - codeparrot_training - Step 46668: {'lr': 5.921277169518874e-06, 'samples': 23894528, 'steps': 46668, 'loss/train': 3.0140647888183594}
+02/26/2022 14:01:13 - INFO - codeparrot_training - Step 46669: {'lr': 5.917737601458556e-06, 'samples': 23895040, 'steps': 46669, 'loss/train': 2.3817083835601807}
+02/26/2022 14:01:19 - INFO - codeparrot_training - Step 46670: {'lr': 5.914199078969202e-06, 'samples': 23895552, 'steps': 46670, 'loss/train': 0.9312611818313599}
+02/26/2022 14:01:22 - INFO - codeparrot_training - Step 46671: {'lr': 5.910661602065881e-06, 'samples': 23896064, 'steps': 46671, 'loss/train': 2.0018935203552246}
+02/26/2022 14:01:28 - INFO - codeparrot_training - Step 46672: {'lr': 5.907125170763805e-06, 'samples': 23896576, 'steps': 46672, 'loss/train': 1.690332055091858}
+02/26/2022 14:01:31 - INFO - codeparrot_training - Step 46673: {'lr': 5.903589785078073e-06, 'samples': 23897088, 'steps': 46673, 'loss/train': 1.9078359603881836}
+02/26/2022 14:01:37 - INFO - codeparrot_training - Step 46674: {'lr': 5.900055445023839e-06, 'samples': 23897600, 'steps': 46674, 'loss/train': 1.2770737409591675}
+02/26/2022 14:01:40 - INFO - codeparrot_training - Step 46675: {'lr': 5.8965221506162845e-06, 'samples': 23898112, 'steps': 46675, 'loss/train': 2.472773790359497}
+02/26/2022 14:01:46 - INFO - codeparrot_training - Step 46676: {'lr': 5.892989901870538e-06, 'samples': 23898624, 'steps': 46676, 'loss/train': 2.0403401851654053}
+02/26/2022 14:01:49 - INFO - codeparrot_training - Step 46677: {'lr': 5.889458698801642e-06, 'samples': 23899136, 'steps': 46677, 'loss/train': 2.0397324562072754}
+02/26/2022 14:01:57 - INFO - codeparrot_training - Step 46678: {'lr': 5.885928541424862e-06, 'samples': 23899648, 'steps': 46678, 'loss/train': 2.402076244354248}
+02/26/2022 14:02:00 - INFO - codeparrot_training - Step 46679: {'lr': 5.882399429755187e-06, 'samples': 23900160, 'steps': 46679, 'loss/train': 1.7445820569992065}
+02/26/2022 14:02:06 - INFO - codeparrot_training - Step 46680: {'lr': 5.878871363807853e-06, 'samples': 23900672, 'steps': 46680, 'loss/train': 1.826703429222107}
+02/26/2022 14:02:09 - INFO - codeparrot_training - Step 46681: {'lr': 5.875344343597877e-06, 'samples': 23901184, 'steps': 46681, 'loss/train': 2.3417375087738037}
+02/26/2022 14:02:15 - INFO - codeparrot_training - Step 46682: {'lr': 5.871818369140442e-06, 'samples': 23901696, 'steps': 46682, 'loss/train': 1.435309886932373}
+02/26/2022 14:02:18 - INFO - codeparrot_training - Step 46683: {'lr': 5.8682934404505615e-06, 'samples': 23902208, 'steps': 46683, 'loss/train': 2.4600021839141846}
+02/26/2022 14:02:24 - INFO - codeparrot_training - Step 46684: {'lr': 5.8647695575434756e-06, 'samples': 23902720, 'steps': 46684, 'loss/train': 2.341658353805542}
+02/26/2022 14:02:27 - INFO - codeparrot_training - Step 46685: {'lr': 5.861246720434115e-06, 'samples': 23903232, 'steps': 46685, 'loss/train': 1.5958586931228638}
+02/26/2022 14:02:32 - INFO - codeparrot_training - Step 46686: {'lr': 5.857724929137692e-06, 'samples': 23903744, 'steps': 46686, 'loss/train': 1.2421454191207886}
+02/26/2022 14:02:36 - INFO - codeparrot_training - Step 46687: {'lr': 5.854204183669221e-06, 'samples': 23904256, 'steps': 46687, 'loss/train': 2.580960750579834}
+02/26/2022 14:02:43 - INFO - codeparrot_training - Step 46688: {'lr': 5.850684484043856e-06, 'samples': 23904768, 'steps': 46688, 'loss/train': 1.64923095703125}
+02/26/2022 14:02:47 - INFO - codeparrot_training - Step 46689: {'lr': 5.847165830276613e-06, 'samples': 23905280, 'steps': 46689, 'loss/train': 1.640005111694336}
+02/26/2022 14:02:52 - INFO - codeparrot_training - Step 46690: {'lr': 5.843648222382591e-06, 'samples': 23905792, 'steps': 46690, 'loss/train': 0.9614103436470032}
+02/26/2022 14:02:56 - INFO - codeparrot_training - Step 46691: {'lr': 5.840131660376835e-06, 'samples': 23906304, 'steps': 46691, 'loss/train': 2.1645987033843994}
+02/26/2022 14:03:02 - INFO - codeparrot_training - Step 46692: {'lr': 5.836616144274442e-06, 'samples': 23906816, 'steps': 46692, 'loss/train': 0.7547193765640259}
+02/26/2022 14:03:05 - INFO - codeparrot_training - Step 46693: {'lr': 5.833101674090457e-06, 'samples': 23907328, 'steps': 46693, 'loss/train': 1.377374291419983}
+02/26/2022 14:03:10 - INFO - codeparrot_training - Step 46694: {'lr': 5.829588249839923e-06, 'samples': 23907840, 'steps': 46694, 'loss/train': 0.6536068320274353}
+02/26/2022 14:03:14 - INFO - codeparrot_training - Step 46695: {'lr': 5.826075871537911e-06, 'samples': 23908352, 'steps': 46695, 'loss/train': 0.4406853914260864}
+02/26/2022 14:03:20 - INFO - codeparrot_training - Step 46696: {'lr': 5.8225645391994095e-06, 'samples': 23908864, 'steps': 46696, 'loss/train': 2.3112378120422363}
+02/26/2022 14:03:23 - INFO - codeparrot_training - Step 46697: {'lr': 5.819054252839573e-06, 'samples': 23909376, 'steps': 46697, 'loss/train': 4.267125606536865}
+02/26/2022 14:03:30 - INFO - codeparrot_training - Step 46698: {'lr': 5.815545012473334e-06, 'samples': 23909888, 'steps': 46698, 'loss/train': 1.966652274131775}
+02/26/2022 14:03:34 - INFO - codeparrot_training - Step 46699: {'lr': 5.81203681811579e-06, 'samples': 23910400, 'steps': 46699, 'loss/train': 0.8440201878547668}
+02/26/2022 14:03:39 - INFO - codeparrot_training - Step 46700: {'lr': 5.808529669781903e-06, 'samples': 23910912, 'steps': 46700, 'loss/train': 1.2726991176605225}
+02/26/2022 14:03:43 - INFO - codeparrot_training - Step 46701: {'lr': 5.805023567486745e-06, 'samples': 23911424, 'steps': 46701, 'loss/train': 1.7687829732894897}
+02/26/2022 14:03:49 - INFO - codeparrot_training - Step 46702: {'lr': 5.801518511245357e-06, 'samples': 23911936, 'steps': 46702, 'loss/train': 1.2757089138031006}
+02/26/2022 14:03:52 - INFO - codeparrot_training - Step 46703: {'lr': 5.798014501072701e-06, 'samples': 23912448, 'steps': 46703, 'loss/train': 2.372622489929199}
+02/26/2022 14:03:58 - INFO - codeparrot_training - Step 46704: {'lr': 5.794511536983793e-06, 'samples': 23912960, 'steps': 46704, 'loss/train': 1.5417133569717407}
+02/26/2022 14:04:01 - INFO - codeparrot_training - Step 46705: {'lr': 5.791009618993676e-06, 'samples': 23913472, 'steps': 46705, 'loss/train': 2.2302517890930176}
+02/26/2022 14:04:07 - INFO - codeparrot_training - Step 46706: {'lr': 5.7875087471173095e-06, 'samples': 23913984, 'steps': 46706, 'loss/train': 2.0182416439056396}
+02/26/2022 14:04:10 - INFO - codeparrot_training - Step 46707: {'lr': 5.784008921369738e-06, 'samples': 23914496, 'steps': 46707, 'loss/train': 1.498546838760376}
+02/26/2022 14:04:16 - INFO - codeparrot_training - Step 46708: {'lr': 5.780510141765893e-06, 'samples': 23915008, 'steps': 46708, 'loss/train': 1.0078778266906738}
+02/26/2022 14:04:19 - INFO - codeparrot_training - Step 46709: {'lr': 5.777012408320792e-06, 'samples': 23915520, 'steps': 46709, 'loss/train': 1.6712477207183838}
+02/26/2022 14:04:25 - INFO - codeparrot_training - Step 46710: {'lr': 5.7735157210494214e-06, 'samples': 23916032, 'steps': 46710, 'loss/train': 0.14980806410312653}
+02/26/2022 14:04:28 - INFO - codeparrot_training - Step 46711: {'lr': 5.770020079966826e-06, 'samples': 23916544, 'steps': 46711, 'loss/train': 1.2432053089141846}
+02/26/2022 14:04:33 - INFO - codeparrot_training - Step 46712: {'lr': 5.766525485087826e-06, 'samples': 23917056, 'steps': 46712, 'loss/train': 1.7230957746505737}
+02/26/2022 14:04:37 - INFO - codeparrot_training - Step 46713: {'lr': 5.763031936427521e-06, 'samples': 23917568, 'steps': 46713, 'loss/train': 2.444340229034424}
+02/26/2022 14:04:44 - INFO - codeparrot_training - Step 46714: {'lr': 5.759539434000816e-06, 'samples': 23918080, 'steps': 46714, 'loss/train': 2.7467739582061768}
+02/26/2022 14:04:48 - INFO - codeparrot_training - Step 46715: {'lr': 5.756047977822726e-06, 'samples': 23918592, 'steps': 46715, 'loss/train': 2.4739832878112793}
+02/26/2022 14:04:53 - INFO - codeparrot_training - Step 46716: {'lr': 5.752557567908129e-06, 'samples': 23919104, 'steps': 46716, 'loss/train': 2.059391975402832}
+02/26/2022 14:04:57 - INFO - codeparrot_training - Step 46717: {'lr': 5.74906820427204e-06, 'samples': 23919616, 'steps': 46717, 'loss/train': 1.3544613122940063}
+02/26/2022 14:05:02 - INFO - codeparrot_training - Step 46718: {'lr': 5.745579886929336e-06, 'samples': 23920128, 'steps': 46718, 'loss/train': 1.9531452655792236}
+02/26/2022 14:05:06 - INFO - codeparrot_training - Step 46719: {'lr': 5.742092615895089e-06, 'samples': 23920640, 'steps': 46719, 'loss/train': 1.3006643056869507}
+02/26/2022 14:05:11 - INFO - codeparrot_training - Step 46720: {'lr': 5.738606391184093e-06, 'samples': 23921152, 'steps': 46720, 'loss/train': 1.5238136053085327}
+02/26/2022 14:05:14 - INFO - codeparrot_training - Step 46721: {'lr': 5.73512121281139e-06, 'samples': 23921664, 'steps': 46721, 'loss/train': 2.314146041870117}
+02/26/2022 14:05:20 - INFO - codeparrot_training - Step 46722: {'lr': 5.7316370807918015e-06, 'samples': 23922176, 'steps': 46722, 'loss/train': 1.0148773193359375}
+02/26/2022 14:05:23 - INFO - codeparrot_training - Step 46723: {'lr': 5.7281539951403726e-06, 'samples': 23922688, 'steps': 46723, 'loss/train': 1.2166365385055542}
+02/26/2022 14:05:31 - INFO - codeparrot_training - Step 46724: {'lr': 5.724671955871951e-06, 'samples': 23923200, 'steps': 46724, 'loss/train': 1.886130928993225}
+02/26/2022 14:05:34 - INFO - codeparrot_training - Step 46725: {'lr': 5.721190963001499e-06, 'samples': 23923712, 'steps': 46725, 'loss/train': 1.9318009614944458}
+02/26/2022 14:05:40 - INFO - codeparrot_training - Step 46726: {'lr': 5.717711016543836e-06, 'samples': 23924224, 'steps': 46726, 'loss/train': 1.7105162143707275}
+02/26/2022 14:05:43 - INFO - codeparrot_training - Step 46727: {'lr': 5.714232116513979e-06, 'samples': 23924736, 'steps': 46727, 'loss/train': 1.3184913396835327}
+02/26/2022 14:05:49 - INFO - codeparrot_training - Step 46728: {'lr': 5.710754262926776e-06, 'samples': 23925248, 'steps': 46728, 'loss/train': 1.3114069700241089}
+02/26/2022 14:05:52 - INFO - codeparrot_training - Step 46729: {'lr': 5.707277455797105e-06, 'samples': 23925760, 'steps': 46729, 'loss/train': 1.8186323642730713}
+02/26/2022 14:05:58 - INFO - codeparrot_training - Step 46730: {'lr': 5.703801695139899e-06, 'samples': 23926272, 'steps': 46730, 'loss/train': 2.53918719291687}
+02/26/2022 14:06:01 - INFO - codeparrot_training - Step 46731: {'lr': 5.7003269809700335e-06, 'samples': 23926784, 'steps': 46731, 'loss/train': 2.002119779586792}
+02/26/2022 14:06:07 - INFO - codeparrot_training - Step 46732: {'lr': 5.696853313302414e-06, 'samples': 23927296, 'steps': 46732, 'loss/train': 1.4633177518844604}
+02/26/2022 14:06:10 - INFO - codeparrot_training - Step 46733: {'lr': 5.693380692151889e-06, 'samples': 23927808, 'steps': 46733, 'loss/train': 2.9672770500183105}
+02/26/2022 14:06:18 - INFO - codeparrot_training - Step 46734: {'lr': 5.689909117533309e-06, 'samples': 23928320, 'steps': 46734, 'loss/train': 1.9776338338851929}
+02/26/2022 14:06:21 - INFO - codeparrot_training - Step 46735: {'lr': 5.686438589461606e-06, 'samples': 23928832, 'steps': 46735, 'loss/train': 1.1969789266586304}
+02/26/2022 14:06:26 - INFO - codeparrot_training - Step 46736: {'lr': 5.68296910795163e-06, 'samples': 23929344, 'steps': 46736, 'loss/train': 1.754347562789917}
+02/26/2022 14:06:30 - INFO - codeparrot_training - Step 46737: {'lr': 5.679500673018257e-06, 'samples': 23929856, 'steps': 46737, 'loss/train': 1.4279024600982666}
+02/26/2022 14:06:36 - INFO - codeparrot_training - Step 46738: {'lr': 5.67603328467628e-06, 'samples': 23930368, 'steps': 46738, 'loss/train': 1.0800254344940186}
+02/26/2022 14:06:39 - INFO - codeparrot_training - Step 46739: {'lr': 5.672566942940605e-06, 'samples': 23930880, 'steps': 46739, 'loss/train': 2.2174394130706787}
+02/26/2022 14:06:45 - INFO - codeparrot_training - Step 46740: {'lr': 5.66910164782608e-06, 'samples': 23931392, 'steps': 46740, 'loss/train': 0.46291375160217285}
+02/26/2022 14:06:48 - INFO - codeparrot_training - Step 46741: {'lr': 5.6656373993475565e-06, 'samples': 23931904, 'steps': 46741, 'loss/train': 1.7776920795440674}
+02/26/2022 14:06:54 - INFO - codeparrot_training - Step 46742: {'lr': 5.662174197519826e-06, 'samples': 23932416, 'steps': 46742, 'loss/train': 1.7433170080184937}
+02/26/2022 14:06:57 - INFO - codeparrot_training - Step 46743: {'lr': 5.658712042357766e-06, 'samples': 23932928, 'steps': 46743, 'loss/train': 1.5523154735565186}
+02/26/2022 14:07:04 - INFO - codeparrot_training - Step 46744: {'lr': 5.655250933876199e-06, 'samples': 23933440, 'steps': 46744, 'loss/train': 2.678130626678467}
+02/26/2022 14:07:08 - INFO - codeparrot_training - Step 46745: {'lr': 5.651790872089946e-06, 'samples': 23933952, 'steps': 46745, 'loss/train': 1.7714459896087646}
+02/26/2022 14:07:13 - INFO - codeparrot_training - Step 46746: {'lr': 5.648331857013855e-06, 'samples': 23934464, 'steps': 46746, 'loss/train': 2.5082831382751465}
+02/26/2022 14:07:17 - INFO - codeparrot_training - Step 46747: {'lr': 5.644873888662666e-06, 'samples': 23934976, 'steps': 46747, 'loss/train': 2.53240966796875}
+02/26/2022 14:07:22 - INFO - codeparrot_training - Step 46748: {'lr': 5.641416967051283e-06, 'samples': 23935488, 'steps': 46748, 'loss/train': 1.7202811241149902}
+02/26/2022 14:07:26 - INFO - codeparrot_training - Step 46749: {'lr': 5.637961092194471e-06, 'samples': 23936000, 'steps': 46749, 'loss/train': 1.3155291080474854}
+02/26/2022 14:07:31 - INFO - codeparrot_training - Step 46750: {'lr': 5.634506264107053e-06, 'samples': 23936512, 'steps': 46750, 'loss/train': 2.2414345741271973}
+02/26/2022 14:07:35 - INFO - codeparrot_training - Step 46751: {'lr': 5.631052482803795e-06, 'samples': 23937024, 'steps': 46751, 'loss/train': 2.1189589500427246}
+02/26/2022 14:07:40 - INFO - codeparrot_training - Step 46752: {'lr': 5.627599748299516e-06, 'samples': 23937536, 'steps': 46752, 'loss/train': 2.3657515048980713}
+02/26/2022 14:07:44 - INFO - codeparrot_training - Step 46753: {'lr': 5.624148060608986e-06, 'samples': 23938048, 'steps': 46753, 'loss/train': 1.374780535697937}
+02/26/2022 14:07:49 - INFO - codeparrot_training - Step 46754: {'lr': 5.620697419747078e-06, 'samples': 23938560, 'steps': 46754, 'loss/train': 0.5056179165840149}
+02/26/2022 14:07:53 - INFO - codeparrot_training - Step 46755: {'lr': 5.6172478257284224e-06, 'samples': 23939072, 'steps': 46755, 'loss/train': 0.8102360367774963}
+02/26/2022 14:07:58 - INFO - codeparrot_training - Step 46756: {'lr': 5.613799278567922e-06, 'samples': 23939584, 'steps': 46756, 'loss/train': 2.0188512802124023}
+02/26/2022 14:08:02 - INFO - codeparrot_training - Step 46757: {'lr': 5.610351778280259e-06, 'samples': 23940096, 'steps': 46757, 'loss/train': 1.7047706842422485}
+02/26/2022 14:08:07 - INFO - codeparrot_training - Step 46758: {'lr': 5.606905324880313e-06, 'samples': 23940608, 'steps': 46758, 'loss/train': 1.7402641773223877}
+02/26/2022 14:08:11 - INFO - codeparrot_training - Step 46759: {'lr': 5.603459918382736e-06, 'samples': 23941120, 'steps': 46759, 'loss/train': 2.0662107467651367}
+02/26/2022 14:08:18 - INFO - codeparrot_training - Step 46760: {'lr': 5.600015558802352e-06, 'samples': 23941632, 'steps': 46760, 'loss/train': 2.454516887664795}
+02/26/2022 14:08:21 - INFO - codeparrot_training - Step 46761: {'lr': 5.596572246153869e-06, 'samples': 23942144, 'steps': 46761, 'loss/train': 2.2911412715911865}
+02/26/2022 14:08:27 - INFO - codeparrot_training - Step 46762: {'lr': 5.593129980452083e-06, 'samples': 23942656, 'steps': 46762, 'loss/train': 1.1962132453918457}
+02/26/2022 14:08:30 - INFO - codeparrot_training - Step 46763: {'lr': 5.589688761711703e-06, 'samples': 23943168, 'steps': 46763, 'loss/train': 2.023801565170288}
+02/26/2022 14:08:36 - INFO - codeparrot_training - Step 46764: {'lr': 5.586248589947523e-06, 'samples': 23943680, 'steps': 46764, 'loss/train': 2.108915328979492}
+02/26/2022 14:08:39 - INFO - codeparrot_training - Step 46765: {'lr': 5.582809465174199e-06, 'samples': 23944192, 'steps': 46765, 'loss/train': 2.0662779808044434}
+02/26/2022 14:08:45 - INFO - codeparrot_training - Step 46766: {'lr': 5.579371387406551e-06, 'samples': 23944704, 'steps': 46766, 'loss/train': 1.2213648557662964}
+02/26/2022 14:08:50 - INFO - codeparrot_training - Step 46767: {'lr': 5.575934356659235e-06, 'samples': 23945216, 'steps': 46767, 'loss/train': 0.6193291544914246}
+02/26/2022 14:08:54 - INFO - codeparrot_training - Step 46768: {'lr': 5.572498372947016e-06, 'samples': 23945728, 'steps': 46768, 'loss/train': 1.8453292846679688}
+02/26/2022 14:09:01 - INFO - codeparrot_training - Step 46769: {'lr': 5.569063436284577e-06, 'samples': 23946240, 'steps': 46769, 'loss/train': 1.2371264696121216}
+02/26/2022 14:09:04 - INFO - codeparrot_training - Step 46770: {'lr': 5.5656295466866845e-06, 'samples': 23946752, 'steps': 46770, 'loss/train': 2.5571675300598145}
+02/26/2022 14:09:08 - INFO - codeparrot_training - Step 46771: {'lr': 5.5621967041679935e-06, 'samples': 23947264, 'steps': 46771, 'loss/train': 1.4545338153839111}
+02/26/2022 14:09:13 - INFO - codeparrot_training - Step 46772: {'lr': 5.558764908743269e-06, 'samples': 23947776, 'steps': 46772, 'loss/train': 2.4007105827331543}
+02/26/2022 14:09:17 - INFO - codeparrot_training - Step 46773: {'lr': 5.555334160427139e-06, 'samples': 23948288, 'steps': 46773, 'loss/train': 2.6095728874206543}
+02/26/2022 14:09:22 - INFO - codeparrot_training - Step 46774: {'lr': 5.551904459234369e-06, 'samples': 23948800, 'steps': 46774, 'loss/train': 1.1040265560150146}
+02/26/2022 14:09:26 - INFO - codeparrot_training - Step 46775: {'lr': 5.548475805179587e-06, 'samples': 23949312, 'steps': 46775, 'loss/train': 1.3597338199615479}
+02/26/2022 14:09:31 - INFO - codeparrot_training - Step 46776: {'lr': 5.545048198277558e-06, 'samples': 23949824, 'steps': 46776, 'loss/train': 1.2872953414916992}
+02/26/2022 14:09:35 - INFO - codeparrot_training - Step 46777: {'lr': 5.5416216385428816e-06, 'samples': 23950336, 'steps': 46777, 'loss/train': 2.621816635131836}
+02/26/2022 14:09:40 - INFO - codeparrot_training - Step 46778: {'lr': 5.538196125990269e-06, 'samples': 23950848, 'steps': 46778, 'loss/train': 1.016654133796692}
+02/26/2022 14:09:44 - INFO - codeparrot_training - Step 46779: {'lr': 5.53477166063443e-06, 'samples': 23951360, 'steps': 46779, 'loss/train': 1.1899681091308594}
+02/26/2022 14:09:51 - INFO - codeparrot_training - Step 46780: {'lr': 5.531348242489992e-06, 'samples': 23951872, 'steps': 46780, 'loss/train': 2.40132212638855}
+02/26/2022 14:09:54 - INFO - codeparrot_training - Step 46781: {'lr': 5.5279258715716096e-06, 'samples': 23952384, 'steps': 46781, 'loss/train': 1.9497724771499634}
+02/26/2022 14:10:00 - INFO - codeparrot_training - Step 46782: {'lr': 5.524504547893994e-06, 'samples': 23952896, 'steps': 46782, 'loss/train': 1.4064050912857056}
+02/26/2022 14:10:03 - INFO - codeparrot_training - Step 46783: {'lr': 5.521084271471744e-06, 'samples': 23953408, 'steps': 46783, 'loss/train': 1.335425853729248}
+02/26/2022 14:10:09 - INFO - codeparrot_training - Step 46784: {'lr': 5.517665042319542e-06, 'samples': 23953920, 'steps': 46784, 'loss/train': 1.67551589012146}
+02/26/2022 14:10:15 - INFO - codeparrot_training - Step 46785: {'lr': 5.5142468604520435e-06, 'samples': 23954432, 'steps': 46785, 'loss/train': 1.8045179843902588}
+02/26/2022 14:10:18 - INFO - codeparrot_training - Step 46786: {'lr': 5.5108297258838755e-06, 'samples': 23954944, 'steps': 46786, 'loss/train': 1.6436898708343506}
+02/26/2022 14:10:23 - INFO - codeparrot_training - Step 46787: {'lr': 5.507413638629693e-06, 'samples': 23955456, 'steps': 46787, 'loss/train': 1.7725476026535034}
+02/26/2022 14:10:27 - INFO - codeparrot_training - Step 46788: {'lr': 5.5039985987040686e-06, 'samples': 23955968, 'steps': 46788, 'loss/train': 1.9082043170928955}
+02/26/2022 14:10:34 - INFO - codeparrot_training - Step 46789: {'lr': 5.500584606121739e-06, 'samples': 23956480, 'steps': 46789, 'loss/train': 1.1810981035232544}
+02/26/2022 14:10:38 - INFO - codeparrot_training - Step 46790: {'lr': 5.497171660897221e-06, 'samples': 23956992, 'steps': 46790, 'loss/train': 3.068063735961914}
+02/26/2022 14:10:43 - INFO - codeparrot_training - Step 46791: {'lr': 5.493759763045225e-06, 'samples': 23957504, 'steps': 46791, 'loss/train': 1.1218230724334717}
+02/26/2022 14:10:47 - INFO - codeparrot_training - Step 46792: {'lr': 5.490348912580268e-06, 'samples': 23958016, 'steps': 46792, 'loss/train': 1.0819718837738037}
+02/26/2022 14:10:52 - INFO - codeparrot_training - Step 46793: {'lr': 5.486939109517086e-06, 'samples': 23958528, 'steps': 46793, 'loss/train': 1.4107733964920044}
+02/26/2022 14:10:56 - INFO - codeparrot_training - Step 46794: {'lr': 5.483530353870142e-06, 'samples': 23959040, 'steps': 46794, 'loss/train': 0.8035626411437988}
+02/26/2022 14:10:59 - INFO - codeparrot_training - Step 46795: {'lr': 5.480122645654145e-06, 'samples': 23959552, 'steps': 46795, 'loss/train': 2.2564501762390137}
+02/26/2022 14:11:05 - INFO - codeparrot_training - Step 46796: {'lr': 5.47671598488364e-06, 'samples': 23960064, 'steps': 46796, 'loss/train': 0.7502437233924866}
+02/26/2022 14:11:08 - INFO - codeparrot_training - Step 46797: {'lr': 5.4733103715732535e-06, 'samples': 23960576, 'steps': 46797, 'loss/train': 1.5987889766693115}
+02/26/2022 14:11:14 - INFO - codeparrot_training - Step 46798: {'lr': 5.469905805737585e-06, 'samples': 23961088, 'steps': 46798, 'loss/train': 1.4029862880706787}
+02/26/2022 14:11:17 - INFO - codeparrot_training - Step 46799: {'lr': 5.466502287391179e-06, 'samples': 23961600, 'steps': 46799, 'loss/train': 2.0328757762908936}
+02/26/2022 14:11:23 - INFO - codeparrot_training - Step 46800: {'lr': 5.463099816548578e-06, 'samples': 23962112, 'steps': 46800, 'loss/train': 2.150216817855835}
+02/26/2022 14:11:26 - INFO - codeparrot_training - Step 46801: {'lr': 5.459698393224466e-06, 'samples': 23962624, 'steps': 46801, 'loss/train': 2.3617630004882812}
+02/26/2022 14:11:32 - INFO - codeparrot_training - Step 46802: {'lr': 5.456298017433331e-06, 'samples': 23963136, 'steps': 46802, 'loss/train': 1.958127737045288}
+02/26/2022 14:11:35 - INFO - codeparrot_training - Step 46803: {'lr': 5.4528986891898e-06, 'samples': 23963648, 'steps': 46803, 'loss/train': 0.9350897669792175}
+02/26/2022 14:11:41 - INFO - codeparrot_training - Step 46804: {'lr': 5.449500408508334e-06, 'samples': 23964160, 'steps': 46804, 'loss/train': 1.5976330041885376}
+02/26/2022 14:11:44 - INFO - codeparrot_training - Step 46805: {'lr': 5.446103175403616e-06, 'samples': 23964672, 'steps': 46805, 'loss/train': 1.387372612953186}
+02/26/2022 14:11:52 - INFO - codeparrot_training - Step 46806: {'lr': 5.442706989890106e-06, 'samples': 23965184, 'steps': 46806, 'loss/train': 0.8588217496871948}
+02/26/2022 14:11:55 - INFO - codeparrot_training - Step 46807: {'lr': 5.439311851982404e-06, 'samples': 23965696, 'steps': 46807, 'loss/train': 2.1299164295196533}
+02/26/2022 14:12:00 - INFO - codeparrot_training - Step 46808: {'lr': 5.435917761694998e-06, 'samples': 23966208, 'steps': 46808, 'loss/train': 1.5720378160476685}
+02/26/2022 14:12:04 - INFO - codeparrot_training - Step 46809: {'lr': 5.432524719042487e-06, 'samples': 23966720, 'steps': 46809, 'loss/train': 1.6974263191223145}
+02/26/2022 14:12:09 - INFO - codeparrot_training - Step 46810: {'lr': 5.42913272403936e-06, 'samples': 23967232, 'steps': 46810, 'loss/train': 3.2405030727386475}
+02/26/2022 14:12:13 - INFO - codeparrot_training - Step 46811: {'lr': 5.425741776700216e-06, 'samples': 23967744, 'steps': 46811, 'loss/train': 0.40171346068382263}
+02/26/2022 14:12:18 - INFO - codeparrot_training - Step 46812: {'lr': 5.4223518770394884e-06, 'samples': 23968256, 'steps': 46812, 'loss/train': 1.8758174180984497}
+02/26/2022 14:12:22 - INFO - codeparrot_training - Step 46813: {'lr': 5.418963025071777e-06, 'samples': 23968768, 'steps': 46813, 'loss/train': 1.6103911399841309}
+02/26/2022 14:12:27 - INFO - codeparrot_training - Step 46814: {'lr': 5.415575220811514e-06, 'samples': 23969280, 'steps': 46814, 'loss/train': 1.8378190994262695}
+02/26/2022 14:12:31 - INFO - codeparrot_training - Step 46815: {'lr': 5.412188464273327e-06, 'samples': 23969792, 'steps': 46815, 'loss/train': 0.42845645546913147}
+02/26/2022 14:12:39 - INFO - codeparrot_training - Step 46816: {'lr': 5.40880275547162e-06, 'samples': 23970304, 'steps': 46816, 'loss/train': 0.4071495234966278}
+02/26/2022 14:12:42 - INFO - codeparrot_training - Step 46817: {'lr': 5.405418094420939e-06, 'samples': 23970816, 'steps': 46817, 'loss/train': 1.4022789001464844}
+02/26/2022 14:12:48 - INFO - codeparrot_training - Step 46818: {'lr': 5.40203448113577e-06, 'samples': 23971328, 'steps': 46818, 'loss/train': 2.10654616355896}
+02/26/2022 14:12:51 - INFO - codeparrot_training - Step 46819: {'lr': 5.398651915630631e-06, 'samples': 23971840, 'steps': 46819, 'loss/train': 1.2048009634017944}
+02/26/2022 14:12:57 - INFO - codeparrot_training - Step 46820: {'lr': 5.39527039792001e-06, 'samples': 23972352, 'steps': 46820, 'loss/train': 0.7934987545013428}
+02/26/2022 14:13:00 - INFO - codeparrot_training - Step 46821: {'lr': 5.391889928018368e-06, 'samples': 23972864, 'steps': 46821, 'loss/train': 1.607393503189087}
+02/26/2022 14:13:06 - INFO - codeparrot_training - Step 46822: {'lr': 5.388510505940164e-06, 'samples': 23973376, 'steps': 46822, 'loss/train': 3.01770281791687}
+02/26/2022 14:13:09 - INFO - codeparrot_training - Step 46823: {'lr': 5.385132131699943e-06, 'samples': 23973888, 'steps': 46823, 'loss/train': 1.8295824527740479}
+02/26/2022 14:13:15 - INFO - codeparrot_training - Step 46824: {'lr': 5.381754805312139e-06, 'samples': 23974400, 'steps': 46824, 'loss/train': 1.118936538696289}
+02/26/2022 14:13:19 - INFO - codeparrot_training - Step 46825: {'lr': 5.378378526791239e-06, 'samples': 23974912, 'steps': 46825, 'loss/train': 1.9161953926086426}
+02/26/2022 14:13:26 - INFO - codeparrot_training - Step 46826: {'lr': 5.375003296151676e-06, 'samples': 23975424, 'steps': 46826, 'loss/train': 1.5617352724075317}
+02/26/2022 14:13:29 - INFO - codeparrot_training - Step 46827: {'lr': 5.371629113407911e-06, 'samples': 23975936, 'steps': 46827, 'loss/train': 0.651698112487793}
+02/26/2022 14:13:35 - INFO - codeparrot_training - Step 46828: {'lr': 5.3682559785744045e-06, 'samples': 23976448, 'steps': 46828, 'loss/train': 1.2936040163040161}
+02/26/2022 14:13:38 - INFO - codeparrot_training - Step 46829: {'lr': 5.364883891665645e-06, 'samples': 23976960, 'steps': 46829, 'loss/train': 1.6932175159454346}
+02/26/2022 14:13:44 - INFO - codeparrot_training - Step 46830: {'lr': 5.36151285269601e-06, 'samples': 23977472, 'steps': 46830, 'loss/train': 0.7406253218650818}
+02/26/2022 14:13:47 - INFO - codeparrot_training - Step 46831: {'lr': 5.35814286167996e-06, 'samples': 23977984, 'steps': 46831, 'loss/train': 1.9916950464248657}
+02/26/2022 14:13:53 - INFO - codeparrot_training - Step 46832: {'lr': 5.3547739186319836e-06, 'samples': 23978496, 'steps': 46832, 'loss/train': 0.9536425471305847}
+02/26/2022 14:13:56 - INFO - codeparrot_training - Step 46833: {'lr': 5.351406023566457e-06, 'samples': 23979008, 'steps': 46833, 'loss/train': 1.3169853687286377}
+02/26/2022 14:14:02 - INFO - codeparrot_training - Step 46834: {'lr': 5.348039176497816e-06, 'samples': 23979520, 'steps': 46834, 'loss/train': 2.296553373336792}
+02/26/2022 14:14:06 - INFO - codeparrot_training - Step 46835: {'lr': 5.3446733774404895e-06, 'samples': 23980032, 'steps': 46835, 'loss/train': 2.886805534362793}
+02/26/2022 14:14:13 - INFO - codeparrot_training - Step 46836: {'lr': 5.341308626408886e-06, 'samples': 23980544, 'steps': 46836, 'loss/train': 1.98550546169281}
+02/26/2022 14:14:16 - INFO - codeparrot_training - Step 46837: {'lr': 5.337944923417465e-06, 'samples': 23981056, 'steps': 46837, 'loss/train': 2.3220250606536865}
+02/26/2022 14:14:22 - INFO - codeparrot_training - Step 46838: {'lr': 5.334582268480575e-06, 'samples': 23981568, 'steps': 46838, 'loss/train': 2.136589288711548}
+02/26/2022 14:14:25 - INFO - codeparrot_training - Step 46839: {'lr': 5.3312206616126235e-06, 'samples': 23982080, 'steps': 46839, 'loss/train': 2.316633462905884}
+02/26/2022 14:14:31 - INFO - codeparrot_training - Step 46840: {'lr': 5.327860102828041e-06, 'samples': 23982592, 'steps': 46840, 'loss/train': 2.326310396194458}
+02/26/2022 14:14:34 - INFO - codeparrot_training - Step 46841: {'lr': 5.324500592141235e-06, 'samples': 23983104, 'steps': 46841, 'loss/train': 1.8999918699264526}
+02/26/2022 14:14:40 - INFO - codeparrot_training - Step 46842: {'lr': 5.321142129566581e-06, 'samples': 23983616, 'steps': 46842, 'loss/train': 0.9576194286346436}
+02/26/2022 14:14:43 - INFO - codeparrot_training - Step 46843: {'lr': 5.317784715118429e-06, 'samples': 23984128, 'steps': 46843, 'loss/train': 1.371591329574585}
+02/26/2022 14:14:49 - INFO - codeparrot_training - Step 46844: {'lr': 5.314428348811212e-06, 'samples': 23984640, 'steps': 46844, 'loss/train': 2.247483968734741}
+02/26/2022 14:14:52 - INFO - codeparrot_training - Step 46845: {'lr': 5.311073030659252e-06, 'samples': 23985152, 'steps': 46845, 'loss/train': 2.291668176651001}
+02/26/2022 14:14:58 - INFO - codeparrot_training - Step 46846: {'lr': 5.307718760677011e-06, 'samples': 23985664, 'steps': 46846, 'loss/train': 1.7852822542190552}
+02/26/2022 14:15:01 - INFO - codeparrot_training - Step 46847: {'lr': 5.30436553887878e-06, 'samples': 23986176, 'steps': 46847, 'loss/train': 1.6136611700057983}
+02/26/2022 14:15:07 - INFO - codeparrot_training - Step 46848: {'lr': 5.30101336527894e-06, 'samples': 23986688, 'steps': 46848, 'loss/train': 1.0306272506713867}
+02/26/2022 14:15:10 - INFO - codeparrot_training - Step 46849: {'lr': 5.297662239891865e-06, 'samples': 23987200, 'steps': 46849, 'loss/train': 0.9531620740890503}
+02/26/2022 14:15:16 - INFO - codeparrot_training - Step 46850: {'lr': 5.294312162731935e-06, 'samples': 23987712, 'steps': 46850, 'loss/train': 1.41732656955719}
+02/26/2022 14:15:19 - INFO - codeparrot_training - Step 46851: {'lr': 5.290963133813415e-06, 'samples': 23988224, 'steps': 46851, 'loss/train': 1.2813444137573242}
+02/26/2022 14:15:27 - INFO - codeparrot_training - Step 46852: {'lr': 5.287615153150737e-06, 'samples': 23988736, 'steps': 46852, 'loss/train': 2.7821009159088135}
+02/26/2022 14:15:30 - INFO - codeparrot_training - Step 46853: {'lr': 5.284268220758171e-06, 'samples': 23989248, 'steps': 46853, 'loss/train': 2.277313709259033}
+02/26/2022 14:15:36 - INFO - codeparrot_training - Step 46854: {'lr': 5.2809223366501734e-06, 'samples': 23989760, 'steps': 46854, 'loss/train': 2.1209797859191895}
+02/26/2022 14:15:39 - INFO - codeparrot_training - Step 46855: {'lr': 5.27757750084093e-06, 'samples': 23990272, 'steps': 46855, 'loss/train': 2.236997127532959}
+02/26/2022 14:15:44 - INFO - codeparrot_training - Step 46856: {'lr': 5.274233713344845e-06, 'samples': 23990784, 'steps': 46856, 'loss/train': 1.7585680484771729}
+02/26/2022 14:15:48 - INFO - codeparrot_training - Step 46857: {'lr': 5.27089097417624e-06, 'samples': 23991296, 'steps': 46857, 'loss/train': 0.3344886600971222}
+02/26/2022 14:15:54 - INFO - codeparrot_training - Step 46858: {'lr': 5.267549283349438e-06, 'samples': 23991808, 'steps': 46858, 'loss/train': 1.0785777568817139}
+02/26/2022 14:15:57 - INFO - codeparrot_training - Step 46859: {'lr': 5.264208640878732e-06, 'samples': 23992320, 'steps': 46859, 'loss/train': 2.6864705085754395}
+02/26/2022 14:16:03 - INFO - codeparrot_training - Step 46860: {'lr': 5.260869046778471e-06, 'samples': 23992832, 'steps': 46860, 'loss/train': 1.2688500881195068}
+02/26/2022 14:16:06 - INFO - codeparrot_training - Step 46861: {'lr': 5.257530501062896e-06, 'samples': 23993344, 'steps': 46861, 'loss/train': 1.5397019386291504}
+02/26/2022 14:16:12 - INFO - codeparrot_training - Step 46862: {'lr': 5.254193003746355e-06, 'samples': 23993856, 'steps': 46862, 'loss/train': 2.134564161300659}
+02/26/2022 14:16:15 - INFO - codeparrot_training - Step 46863: {'lr': 5.250856554843142e-06, 'samples': 23994368, 'steps': 46863, 'loss/train': 1.7671452760696411}
+02/26/2022 14:16:22 - INFO - codeparrot_training - Step 46864: {'lr': 5.247521154367552e-06, 'samples': 23994880, 'steps': 46864, 'loss/train': 1.6623950004577637}
+02/26/2022 14:16:26 - INFO - codeparrot_training - Step 46865: {'lr': 5.24418680233385e-06, 'samples': 23995392, 'steps': 46865, 'loss/train': 1.0587592124938965}
+02/26/2022 14:16:31 - INFO - codeparrot_training - Step 46866: {'lr': 5.2408534987563316e-06, 'samples': 23995904, 'steps': 46866, 'loss/train': 1.7487235069274902}
+02/26/2022 14:16:35 - INFO - codeparrot_training - Step 46867: {'lr': 5.23752124364929e-06, 'samples': 23996416, 'steps': 46867, 'loss/train': 1.8403130769729614}
+02/26/2022 14:16:40 - INFO - codeparrot_training - Step 46868: {'lr': 5.234190037026992e-06, 'samples': 23996928, 'steps': 46868, 'loss/train': 1.5571342706680298}
+02/26/2022 14:16:44 - INFO - codeparrot_training - Step 46869: {'lr': 5.230859878903676e-06, 'samples': 23997440, 'steps': 46869, 'loss/train': 1.5209988355636597}
+02/26/2022 14:16:49 - INFO - codeparrot_training - Step 46870: {'lr': 5.227530769293637e-06, 'samples': 23997952, 'steps': 46870, 'loss/train': 1.699662208557129}
+02/26/2022 14:16:53 - INFO - codeparrot_training - Step 46871: {'lr': 5.224202708211168e-06, 'samples': 23998464, 'steps': 46871, 'loss/train': 2.245939016342163}
+02/26/2022 14:16:58 - INFO - codeparrot_training - Step 46872: {'lr': 5.220875695670452e-06, 'samples': 23998976, 'steps': 46872, 'loss/train': 2.2990434169769287}
+02/26/2022 14:17:02 - INFO - codeparrot_training - Step 46873: {'lr': 5.217549731685811e-06, 'samples': 23999488, 'steps': 46873, 'loss/train': 0.8813782334327698}
+02/26/2022 14:17:09 - INFO - codeparrot_training - Step 46874: {'lr': 5.214224816271429e-06, 'samples': 24000000, 'steps': 46874, 'loss/train': 0.6537315249443054}
+02/26/2022 14:17:13 - INFO - codeparrot_training - Step 46875: {'lr': 5.210900949441571e-06, 'samples': 24000512, 'steps': 46875, 'loss/train': 2.4819328784942627}
+02/26/2022 14:17:18 - INFO - codeparrot_training - Step 46876: {'lr': 5.207578131210533e-06, 'samples': 24001024, 'steps': 46876, 'loss/train': 1.7477836608886719}
+02/26/2022 14:17:22 - INFO - codeparrot_training - Step 46877: {'lr': 5.204256361592469e-06, 'samples': 24001536, 'steps': 46877, 'loss/train': 1.5514863729476929}
+02/26/2022 14:17:27 - INFO - codeparrot_training - Step 46878: {'lr': 5.200935640601617e-06, 'samples': 24002048, 'steps': 46878, 'loss/train': 2.7574987411499023}
+02/26/2022 14:17:31 - INFO - codeparrot_training - Step 46879: {'lr': 5.197615968252245e-06, 'samples': 24002560, 'steps': 46879, 'loss/train': 2.185171127319336}
+02/26/2022 14:17:36 - INFO - codeparrot_training - Step 46880: {'lr': 5.194297344558535e-06, 'samples': 24003072, 'steps': 46880, 'loss/train': 0.9565368294715881}
+02/26/2022 14:17:40 - INFO - codeparrot_training - Step 46881: {'lr': 5.190979769534782e-06, 'samples': 24003584, 'steps': 46881, 'loss/train': 1.2464548349380493}
+02/26/2022 14:17:45 - INFO - codeparrot_training - Step 46882: {'lr': 5.1876632431950845e-06, 'samples': 24004096, 'steps': 46882, 'loss/train': 1.5799823999404907}
+02/26/2022 14:17:49 - INFO - codeparrot_training - Step 46883: {'lr': 5.18434776555371e-06, 'samples': 24004608, 'steps': 46883, 'loss/train': 0.8450912833213806}
+02/26/2022 14:17:56 - INFO - codeparrot_training - Step 46884: {'lr': 5.181033336624841e-06, 'samples': 24005120, 'steps': 46884, 'loss/train': 1.6226387023925781}
+02/26/2022 14:17:59 - INFO - codeparrot_training - Step 46885: {'lr': 5.177719956422716e-06, 'samples': 24005632, 'steps': 46885, 'loss/train': 1.790230631828308}
+02/26/2022 14:18:05 - INFO - codeparrot_training - Step 46886: {'lr': 5.174407624961463e-06, 'samples': 24006144, 'steps': 46886, 'loss/train': 1.0855509042739868}
+02/26/2022 14:18:09 - INFO - codeparrot_training - Step 46887: {'lr': 5.171096342255349e-06, 'samples': 24006656, 'steps': 46887, 'loss/train': 2.2399508953094482}
+02/26/2022 14:18:14 - INFO - codeparrot_training - Step 46888: {'lr': 5.167786108318473e-06, 'samples': 24007168, 'steps': 46888, 'loss/train': 1.3225986957550049}
+02/26/2022 14:18:18 - INFO - codeparrot_training - Step 46889: {'lr': 5.164476923165101e-06, 'samples': 24007680, 'steps': 46889, 'loss/train': 1.6188924312591553}
+02/26/2022 14:18:23 - INFO - codeparrot_training - Step 46890: {'lr': 5.161168786809334e-06, 'samples': 24008192, 'steps': 46890, 'loss/train': 1.9687080383300781}
+02/26/2022 14:18:27 - INFO - codeparrot_training - Step 46891: {'lr': 5.157861699265382e-06, 'samples': 24008704, 'steps': 46891, 'loss/train': 1.893865942955017}
+02/26/2022 14:18:32 - INFO - codeparrot_training - Step 46892: {'lr': 5.1545556605473724e-06, 'samples': 24009216, 'steps': 46892, 'loss/train': 1.6712392568588257}
+02/26/2022 14:18:36 - INFO - codeparrot_training - Step 46893: {'lr': 5.151250670669544e-06, 'samples': 24009728, 'steps': 46893, 'loss/train': 2.116797924041748}
+02/26/2022 14:18:41 - INFO - codeparrot_training - Step 46894: {'lr': 5.147946729645997e-06, 'samples': 24010240, 'steps': 46894, 'loss/train': 0.9294131398200989}
+02/26/2022 14:18:45 - INFO - codeparrot_training - Step 46895: {'lr': 5.144643837490914e-06, 'samples': 24010752, 'steps': 46895, 'loss/train': 1.5480866432189941}
+02/26/2022 14:18:50 - INFO - codeparrot_training - Step 46896: {'lr': 5.1413419942183955e-06, 'samples': 24011264, 'steps': 46896, 'loss/train': 1.5727174282073975}
+02/26/2022 14:18:54 - INFO - codeparrot_training - Step 46897: {'lr': 5.138041199842653e-06, 'samples': 24011776, 'steps': 46897, 'loss/train': 2.2603986263275146}
+02/26/2022 14:19:02 - INFO - codeparrot_training - Step 46898: {'lr': 5.134741454377756e-06, 'samples': 24012288, 'steps': 46898, 'loss/train': 1.9924979209899902}
+02/26/2022 14:19:05 - INFO - codeparrot_training - Step 46899: {'lr': 5.13144275783789e-06, 'samples': 24012800, 'steps': 46899, 'loss/train': 1.704669713973999}
+02/26/2022 14:19:11 - INFO - codeparrot_training - Step 46900: {'lr': 5.128145110237154e-06, 'samples': 24013312, 'steps': 46900, 'loss/train': 1.674017310142517}
+02/26/2022 14:19:14 - INFO - codeparrot_training - Step 46901: {'lr': 5.124848511589703e-06, 'samples': 24013824, 'steps': 46901, 'loss/train': 1.8421716690063477}
+02/26/2022 14:19:20 - INFO - codeparrot_training - Step 46902: {'lr': 5.1215529619096365e-06, 'samples': 24014336, 'steps': 46902, 'loss/train': 1.8025825023651123}
+02/26/2022 14:19:23 - INFO - codeparrot_training - Step 46903: {'lr': 5.118258461211084e-06, 'samples': 24014848, 'steps': 46903, 'loss/train': 1.0603227615356445}
+02/26/2022 14:19:29 - INFO - codeparrot_training - Step 46904: {'lr': 5.114965009508143e-06, 'samples': 24015360, 'steps': 46904, 'loss/train': 1.1741602420806885}
+02/26/2022 14:19:32 - INFO - codeparrot_training - Step 46905: {'lr': 5.111672606814944e-06, 'samples': 24015872, 'steps': 46905, 'loss/train': 0.34088361263275146}
+02/26/2022 14:19:38 - INFO - codeparrot_training - Step 46906: {'lr': 5.1083812531455835e-06, 'samples': 24016384, 'steps': 46906, 'loss/train': 2.0534825325012207}
+02/26/2022 14:19:41 - INFO - codeparrot_training - Step 46907: {'lr': 5.105090948514163e-06, 'samples': 24016896, 'steps': 46907, 'loss/train': 2.3434557914733887}
+02/26/2022 14:19:48 - INFO - codeparrot_training - Step 46908: {'lr': 5.101801692934727e-06, 'samples': 24017408, 'steps': 46908, 'loss/train': 0.6817491054534912}
+02/26/2022 14:19:52 - INFO - codeparrot_training - Step 46909: {'lr': 5.098513486421458e-06, 'samples': 24017920, 'steps': 46909, 'loss/train': 2.259007453918457}
+02/26/2022 14:19:57 - INFO - codeparrot_training - Step 46910: {'lr': 5.095226328988373e-06, 'samples': 24018432, 'steps': 46910, 'loss/train': 1.4829856157302856}
+02/26/2022 14:20:01 - INFO - codeparrot_training - Step 46911: {'lr': 5.091940220649571e-06, 'samples': 24018944, 'steps': 46911, 'loss/train': 1.1623817682266235}
+02/26/2022 14:20:06 - INFO - codeparrot_training - Step 46912: {'lr': 5.088655161419126e-06, 'samples': 24019456, 'steps': 46912, 'loss/train': 1.6048622131347656}
+02/26/2022 14:20:10 - INFO - codeparrot_training - Step 46913: {'lr': 5.085371151311108e-06, 'samples': 24019968, 'steps': 46913, 'loss/train': 1.5661735534667969}
+02/26/2022 14:20:15 - INFO - codeparrot_training - Step 46914: {'lr': 5.082088190339618e-06, 'samples': 24020480, 'steps': 46914, 'loss/train': 2.6421194076538086}
+02/26/2022 14:20:19 - INFO - codeparrot_training - Step 46915: {'lr': 5.078806278518672e-06, 'samples': 24020992, 'steps': 46915, 'loss/train': 1.5811841487884521}
+02/26/2022 14:20:24 - INFO - codeparrot_training - Step 46916: {'lr': 5.075525415862342e-06, 'samples': 24021504, 'steps': 46916, 'loss/train': 2.1828420162200928}
+02/26/2022 14:20:28 - INFO - codeparrot_training - Step 46917: {'lr': 5.072245602384701e-06, 'samples': 24022016, 'steps': 46917, 'loss/train': 2.3859052658081055}
+02/26/2022 14:20:33 - INFO - codeparrot_training - Step 46918: {'lr': 5.068966838099793e-06, 'samples': 24022528, 'steps': 46918, 'loss/train': 1.5189543962478638}
+02/26/2022 14:20:37 - INFO - codeparrot_training - Step 46919: {'lr': 5.0656891230216325e-06, 'samples': 24023040, 'steps': 46919, 'loss/train': 1.4126026630401611}
+02/26/2022 14:20:45 - INFO - codeparrot_training - Step 46920: {'lr': 5.0624124571643225e-06, 'samples': 24023552, 'steps': 46920, 'loss/train': 1.1939884424209595}
+02/26/2022 14:20:48 - INFO - codeparrot_training - Step 46921: {'lr': 5.059136840541823e-06, 'samples': 24024064, 'steps': 46921, 'loss/train': 1.0683212280273438}
+02/26/2022 14:20:53 - INFO - codeparrot_training - Step 46922: {'lr': 5.055862273168233e-06, 'samples': 24024576, 'steps': 46922, 'loss/train': 1.4167451858520508}
+02/26/2022 14:20:57 - INFO - codeparrot_training - Step 46923: {'lr': 5.052588755057513e-06, 'samples': 24025088, 'steps': 46923, 'loss/train': 2.0334348678588867}
+02/26/2022 14:21:03 - INFO - codeparrot_training - Step 46924: {'lr': 5.049316286223765e-06, 'samples': 24025600, 'steps': 46924, 'loss/train': 2.108639717102051}
+02/26/2022 14:21:06 - INFO - codeparrot_training - Step 46925: {'lr': 5.046044866680949e-06, 'samples': 24026112, 'steps': 46925, 'loss/train': 1.158650517463684}
+02/26/2022 14:21:12 - INFO - codeparrot_training - Step 46926: {'lr': 5.042774496443081e-06, 'samples': 24026624, 'steps': 46926, 'loss/train': 1.2206578254699707}
+02/26/2022 14:21:15 - INFO - codeparrot_training - Step 46927: {'lr': 5.039505175524178e-06, 'samples': 24027136, 'steps': 46927, 'loss/train': 0.8434476852416992}
+02/26/2022 14:21:21 - INFO - codeparrot_training - Step 46928: {'lr': 5.0362369039382845e-06, 'samples': 24027648, 'steps': 46928, 'loss/train': 2.3522393703460693}
+02/26/2022 14:21:24 - INFO - codeparrot_training - Step 46929: {'lr': 5.032969681699362e-06, 'samples': 24028160, 'steps': 46929, 'loss/train': 2.226112127304077}
+02/26/2022 14:21:31 - INFO - codeparrot_training - Step 46930: {'lr': 5.029703508821398e-06, 'samples': 24028672, 'steps': 46930, 'loss/train': 1.7182066440582275}
+02/26/2022 14:21:35 - INFO - codeparrot_training - Step 46931: {'lr': 5.026438385318382e-06, 'samples': 24029184, 'steps': 46931, 'loss/train': 0.6680001020431519}
+02/26/2022 14:21:40 - INFO - codeparrot_training - Step 46932: {'lr': 5.0231743112043595e-06, 'samples': 24029696, 'steps': 46932, 'loss/train': 1.7184449434280396}
+02/26/2022 14:21:44 - INFO - codeparrot_training - Step 46933: {'lr': 5.019911286493234e-06, 'samples': 24030208, 'steps': 46933, 'loss/train': 1.6746971607208252}
+02/26/2022 14:21:49 - INFO - codeparrot_training - Step 46934: {'lr': 5.01664931119905e-06, 'samples': 24030720, 'steps': 46934, 'loss/train': 1.3019837141036987}
+02/26/2022 14:21:53 - INFO - codeparrot_training - Step 46935: {'lr': 5.013388385335715e-06, 'samples': 24031232, 'steps': 46935, 'loss/train': 1.9929417371749878}
+02/26/2022 14:21:58 - INFO - codeparrot_training - Step 46936: {'lr': 5.0101285089172445e-06, 'samples': 24031744, 'steps': 46936, 'loss/train': 2.174039125442505}
+02/26/2022 14:22:02 - INFO - codeparrot_training - Step 46937: {'lr': 5.006869681957599e-06, 'samples': 24032256, 'steps': 46937, 'loss/train': 2.351221799850464}
+02/26/2022 14:22:07 - INFO - codeparrot_training - Step 46938: {'lr': 5.003611904470739e-06, 'samples': 24032768, 'steps': 46938, 'loss/train': 1.9174375534057617}
+02/26/2022 14:22:11 - INFO - codeparrot_training - Step 46939: {'lr': 5.000355176470572e-06, 'samples': 24033280, 'steps': 46939, 'loss/train': 0.2696290910243988}
+02/26/2022 14:22:17 - INFO - codeparrot_training - Step 46940: {'lr': 4.997099497971114e-06, 'samples': 24033792, 'steps': 46940, 'loss/train': 2.0405056476593018}
+02/26/2022 14:22:20 - INFO - codeparrot_training - Step 46941: {'lr': 4.993844868986269e-06, 'samples': 24034304, 'steps': 46941, 'loss/train': 1.7250773906707764}
+02/26/2022 14:22:26 - INFO - codeparrot_training - Step 46942: {'lr': 4.990591289530028e-06, 'samples': 24034816, 'steps': 46942, 'loss/train': 2.0224058628082275}
+02/26/2022 14:22:29 - INFO - codeparrot_training - Step 46943: {'lr': 4.987338759616267e-06, 'samples': 24035328, 'steps': 46943, 'loss/train': 1.268270492553711}
+02/26/2022 14:22:36 - INFO - codeparrot_training - Step 46944: {'lr': 4.984087279258948e-06, 'samples': 24035840, 'steps': 46944, 'loss/train': 1.3498919010162354}
+02/26/2022 14:22:40 - INFO - codeparrot_training - Step 46945: {'lr': 4.980836848471976e-06, 'samples': 24036352, 'steps': 46945, 'loss/train': 0.8643433451652527}
+02/26/2022 14:22:45 - INFO - codeparrot_training - Step 46946: {'lr': 4.9775874672693674e-06, 'samples': 24036864, 'steps': 46946, 'loss/train': 1.3721646070480347}
+02/26/2022 14:22:49 - INFO - codeparrot_training - Step 46947: {'lr': 4.9743391356648905e-06, 'samples': 24037376, 'steps': 46947, 'loss/train': 2.0076985359191895}
+02/26/2022 14:22:54 - INFO - codeparrot_training - Step 46948: {'lr': 4.971091853672588e-06, 'samples': 24037888, 'steps': 46948, 'loss/train': 1.3897405862808228}
+02/26/2022 14:22:58 - INFO - codeparrot_training - Step 46949: {'lr': 4.967845621306283e-06, 'samples': 24038400, 'steps': 46949, 'loss/train': 2.2570042610168457}
+02/26/2022 14:23:03 - INFO - codeparrot_training - Step 46950: {'lr': 4.964600438579991e-06, 'samples': 24038912, 'steps': 46950, 'loss/train': 1.2331100702285767}
+02/26/2022 14:23:07 - INFO - codeparrot_training - Step 46951: {'lr': 4.961356305507481e-06, 'samples': 24039424, 'steps': 46951, 'loss/train': 2.7906792163848877}
+02/26/2022 14:23:12 - INFO - codeparrot_training - Step 46952: {'lr': 4.958113222102739e-06, 'samples': 24039936, 'steps': 46952, 'loss/train': 1.2655057907104492}
+02/26/2022 14:23:16 - INFO - codeparrot_training - Step 46953: {'lr': 4.954871188379617e-06, 'samples': 24040448, 'steps': 46953, 'loss/train': 1.2923851013183594}
+02/26/2022 14:23:23 - INFO - codeparrot_training - Step 46954: {'lr': 4.95163020435202e-06, 'samples': 24040960, 'steps': 46954, 'loss/train': 3.0559093952178955}
+02/26/2022 14:23:26 - INFO - codeparrot_training - Step 46955: {'lr': 4.948390270033826e-06, 'samples': 24041472, 'steps': 46955, 'loss/train': 1.5363434553146362}
+02/26/2022 14:23:32 - INFO - codeparrot_training - Step 46956: {'lr': 4.945151385438912e-06, 'samples': 24041984, 'steps': 46956, 'loss/train': 1.6995373964309692}
+02/26/2022 14:23:35 - INFO - codeparrot_training - Step 46957: {'lr': 4.941913550581184e-06, 'samples': 24042496, 'steps': 46957, 'loss/train': 0.6171913743019104}
+02/26/2022 14:23:41 - INFO - codeparrot_training - Step 46958: {'lr': 4.938676765474437e-06, 'samples': 24043008, 'steps': 46958, 'loss/train': 1.4346436262130737}
+02/26/2022 14:23:44 - INFO - codeparrot_training - Step 46959: {'lr': 4.935441030132631e-06, 'samples': 24043520, 'steps': 46959, 'loss/train': 1.4168815612792969}
+02/26/2022 14:23:50 - INFO - codeparrot_training - Step 46960: {'lr': 4.932206344569562e-06, 'samples': 24044032, 'steps': 46960, 'loss/train': 1.081868052482605}
+02/26/2022 14:23:53 - INFO - codeparrot_training - Step 46961: {'lr': 4.9289727087990784e-06, 'samples': 24044544, 'steps': 46961, 'loss/train': 1.7110190391540527}
+02/26/2022 14:23:59 - INFO - codeparrot_training - Step 46962: {'lr': 4.92574012283506e-06, 'samples': 24045056, 'steps': 46962, 'loss/train': 0.7463676333427429}
+02/26/2022 14:24:02 - INFO - codeparrot_training - Step 46963: {'lr': 4.9225085866913825e-06, 'samples': 24045568, 'steps': 46963, 'loss/train': 0.28492891788482666}
+02/26/2022 14:24:08 - INFO - codeparrot_training - Step 46964: {'lr': 4.919278100381841e-06, 'samples': 24046080, 'steps': 46964, 'loss/train': 0.6663783192634583}
+02/26/2022 14:24:11 - INFO - codeparrot_training - Step 46965: {'lr': 4.916048663920314e-06, 'samples': 24046592, 'steps': 46965, 'loss/train': 2.271629810333252}
+02/26/2022 14:24:19 - INFO - codeparrot_training - Step 46966: {'lr': 4.912820277320568e-06, 'samples': 24047104, 'steps': 46966, 'loss/train': 1.2768833637237549}
+02/26/2022 14:24:23 - INFO - codeparrot_training - Step 46967: {'lr': 4.9095929405965075e-06, 'samples': 24047616, 'steps': 46967, 'loss/train': 2.0954296588897705}
+02/26/2022 14:24:28 - INFO - codeparrot_training - Step 46968: {'lr': 4.906366653761901e-06, 'samples': 24048128, 'steps': 46968, 'loss/train': 2.1196422576904297}
+02/26/2022 14:24:31 - INFO - codeparrot_training - Step 46969: {'lr': 4.903141416830625e-06, 'samples': 24048640, 'steps': 46969, 'loss/train': 0.7138397693634033}
+02/26/2022 14:24:37 - INFO - codeparrot_training - Step 46970: {'lr': 4.8999172298164195e-06, 'samples': 24049152, 'steps': 46970, 'loss/train': 2.2004547119140625}
+02/26/2022 14:24:40 - INFO - codeparrot_training - Step 46971: {'lr': 4.896694092733189e-06, 'samples': 24049664, 'steps': 46971, 'loss/train': 2.0217840671539307}
+02/26/2022 14:24:46 - INFO - codeparrot_training - Step 46972: {'lr': 4.893472005594674e-06, 'samples': 24050176, 'steps': 46972, 'loss/train': 2.4144418239593506}
+02/26/2022 14:24:50 - INFO - codeparrot_training - Step 46973: {'lr': 4.890250968414694e-06, 'samples': 24050688, 'steps': 46973, 'loss/train': 1.228333830833435}
+02/26/2022 14:24:55 - INFO - codeparrot_training - Step 46974: {'lr': 4.887030981207019e-06, 'samples': 24051200, 'steps': 46974, 'loss/train': 1.450463891029358}
+02/26/2022 14:24:58 - INFO - codeparrot_training - Step 46975: {'lr': 4.883812043985497e-06, 'samples': 24051712, 'steps': 46975, 'loss/train': 1.1599994897842407}
+02/26/2022 14:25:06 - INFO - codeparrot_training - Step 46976: {'lr': 4.880594156763896e-06, 'samples': 24052224, 'steps': 46976, 'loss/train': 1.5338329076766968}
+02/26/2022 14:25:09 - INFO - codeparrot_training - Step 46977: {'lr': 4.877377319556009e-06, 'samples': 24052736, 'steps': 46977, 'loss/train': 2.1217143535614014}
+02/26/2022 14:25:15 - INFO - codeparrot_training - Step 46978: {'lr': 4.874161532375576e-06, 'samples': 24053248, 'steps': 46978, 'loss/train': 1.2750530242919922}
+02/26/2022 14:25:18 - INFO - codeparrot_training - Step 46979: {'lr': 4.8709467952364194e-06, 'samples': 24053760, 'steps': 46979, 'loss/train': 2.248365879058838}
+02/26/2022 14:25:24 - INFO - codeparrot_training - Step 46980: {'lr': 4.86773310815225e-06, 'samples': 24054272, 'steps': 46980, 'loss/train': 2.3472487926483154}
+02/26/2022 14:25:27 - INFO - codeparrot_training - Step 46981: {'lr': 4.864520471136919e-06, 'samples': 24054784, 'steps': 46981, 'loss/train': 1.949555516242981}
+02/26/2022 14:25:33 - INFO - codeparrot_training - Step 46982: {'lr': 4.861308884204108e-06, 'samples': 24055296, 'steps': 46982, 'loss/train': 2.0906949043273926}
+02/26/2022 14:25:36 - INFO - codeparrot_training - Step 46983: {'lr': 4.8580983473676396e-06, 'samples': 24055808, 'steps': 46983, 'loss/train': 2.1700401306152344}
+02/26/2022 14:25:42 - INFO - codeparrot_training - Step 46984: {'lr': 4.8548888606411986e-06, 'samples': 24056320, 'steps': 46984, 'loss/train': 1.7726908922195435}
+02/26/2022 14:25:45 - INFO - codeparrot_training - Step 46985: {'lr': 4.851680424038635e-06, 'samples': 24056832, 'steps': 46985, 'loss/train': 1.71482515335083}
+02/26/2022 14:25:51 - INFO - codeparrot_training - Step 46986: {'lr': 4.848473037573575e-06, 'samples': 24057344, 'steps': 46986, 'loss/train': 1.1402982473373413}
+02/26/2022 14:25:54 - INFO - codeparrot_training - Step 46987: {'lr': 4.845266701259815e-06, 'samples': 24057856, 'steps': 46987, 'loss/train': 2.1147561073303223}
+02/26/2022 14:26:00 - INFO - codeparrot_training - Step 46988: {'lr': 4.842061415111093e-06, 'samples': 24058368, 'steps': 46988, 'loss/train': 1.3244916200637817}
+02/26/2022 14:26:03 - INFO - codeparrot_training - Step 46989: {'lr': 4.838857179141148e-06, 'samples': 24058880, 'steps': 46989, 'loss/train': 1.4193236827850342}
+02/26/2022 14:26:09 - INFO - codeparrot_training - Step 46990: {'lr': 4.835653993363693e-06, 'samples': 24059392, 'steps': 46990, 'loss/train': 2.154270648956299}
+02/26/2022 14:26:12 - INFO - codeparrot_training - Step 46991: {'lr': 4.832451857792436e-06, 'samples': 24059904, 'steps': 46991, 'loss/train': 1.374183177947998}
+02/26/2022 14:26:19 - INFO - codeparrot_training - Step 46992: {'lr': 4.829250772441091e-06, 'samples': 24060416, 'steps': 46992, 'loss/train': 0.976502001285553}
+02/26/2022 14:26:23 - INFO - codeparrot_training - Step 46993: {'lr': 4.826050737323423e-06, 'samples': 24060928, 'steps': 46993, 'loss/train': 1.0698624849319458}
+02/26/2022 14:26:28 - INFO - codeparrot_training - Step 46994: {'lr': 4.822851752453089e-06, 'samples': 24061440, 'steps': 46994, 'loss/train': 1.904506802558899}
+02/26/2022 14:26:32 - INFO - codeparrot_training - Step 46995: {'lr': 4.8196538178438e-06, 'samples': 24061952, 'steps': 46995, 'loss/train': 2.0647518634796143}
+02/26/2022 14:26:37 - INFO - codeparrot_training - Step 46996: {'lr': 4.816456933509239e-06, 'samples': 24062464, 'steps': 46996, 'loss/train': 1.026369571685791}
+02/26/2022 14:26:41 - INFO - codeparrot_training - Step 46997: {'lr': 4.8132610994631455e-06, 'samples': 24062976, 'steps': 46997, 'loss/train': 2.110910415649414}
+02/26/2022 14:26:46 - INFO - codeparrot_training - Step 46998: {'lr': 4.810066315719175e-06, 'samples': 24063488, 'steps': 46998, 'loss/train': 1.6281605958938599}
+02/26/2022 14:26:52 - INFO - codeparrot_training - Step 46999: {'lr': 4.806872582291038e-06, 'samples': 24064000, 'steps': 46999, 'loss/train': 1.2316577434539795}
+02/26/2022 14:26:52 - INFO - codeparrot_training - Evaluating and saving model checkpoint