diff --git "a/log/debug_0.log" "b/log/debug_0.log"
--- "a/log/debug_0.log"
+++ "b/log/debug_0.log"
@@ -49596,3 +49596,1009 @@ Use FP16 precision: False
 02/26/2022 17:01:54 - INFO - codeparrot_training - Step 48998: {'lr': 5.374116399974871e-07, 'samples': 25087488, 'steps': 48998, 'loss/train': 2.223346471786499}
 02/26/2022 17:01:57 - INFO - codeparrot_training - Step 48999: {'lr': 5.363398808617414e-07, 'samples': 25088000, 'steps': 48999, 'loss/train': 1.8959709405899048}
 02/26/2022 17:01:57 - INFO - codeparrot_training - Evaluating and saving model checkpoint
+02/26/2022 17:02:15 - WARNING - huggingface_hub.repository - Several commits (49) will be pushed upstream.
+02/26/2022 17:02:15 - WARNING - huggingface_hub.repository - The progress bars may be unreliable.
+02/26/2022 17:02:50 - WARNING - huggingface_hub.repository - To https://huggingface.co/ncoop57/multi-code-clippy
+   8e26ae5..508ef45  floral-grass-11 -> floral-grass-11
+
+02/26/2022 17:02:58 - INFO - codeparrot_training - Step 49000: {'lr': 5.352691903491303e-07, 'samples': 25088512, 'steps': 49000, 'loss/train': 1.3508716821670532}
+02/26/2022 17:03:01 - INFO - codeparrot_training - Step 49001: {'lr': 5.341995684642053e-07, 'samples': 25089024, 'steps': 49001, 'loss/train': 0.7777153253555298}
+02/26/2022 17:03:07 - INFO - codeparrot_training - Step 49002: {'lr': 5.331310152115465e-07, 'samples': 25089536, 'steps': 49002, 'loss/train': 1.7808958292007446}
+02/26/2022 17:03:10 - INFO - codeparrot_training - Step 49003: {'lr': 5.320635305957333e-07, 'samples': 25090048, 'steps': 49003, 'loss/train': 2.8032636642456055}
+02/26/2022 17:03:16 - INFO - codeparrot_training - Step 49004: {'lr': 5.309971146213732e-07, 'samples': 25090560, 'steps': 49004, 'loss/train': 1.5926603078842163}
+02/26/2022 17:03:19 - INFO - codeparrot_training - Step 49005: {'lr': 5.299317672929904e-07, 'samples': 25091072, 'steps': 49005, 'loss/train': 1.5938676595687866}
+02/26/2022 17:03:25 - INFO - codeparrot_training - Step 49006: {'lr': 5.288674886151645e-07, 'samples': 25091584, 'steps': 49006, 'loss/train': 1.1151988506317139}
+02/26/2022 17:03:28 - INFO - codeparrot_training - Step 49007: {'lr': 5.278042785924475e-07, 'samples': 25092096, 'steps': 49007, 'loss/train': 1.8216667175292969}
+02/26/2022 17:03:34 - INFO - codeparrot_training - Step 49008: {'lr': 5.26742137229419e-07, 'samples': 25092608, 'steps': 49008, 'loss/train': 2.0989344120025635}
+02/26/2022 17:03:37 - INFO - codeparrot_training - Step 49009: {'lr': 5.256810645306032e-07, 'samples': 25093120, 'steps': 49009, 'loss/train': 0.9914582967758179}
+02/26/2022 17:03:43 - INFO - codeparrot_training - Step 49010: {'lr': 5.246210605005519e-07, 'samples': 25093632, 'steps': 49010, 'loss/train': 1.4833296537399292}
+02/26/2022 17:03:47 - INFO - codeparrot_training - Step 49011: {'lr': 5.235621251437894e-07, 'samples': 25094144, 'steps': 49011, 'loss/train': 0.8157903552055359}
+02/26/2022 17:03:52 - INFO - codeparrot_training - Step 49012: {'lr': 5.225042584648954e-07, 'samples': 25094656, 'steps': 49012, 'loss/train': 2.2832565307617188}
+02/26/2022 17:03:56 - INFO - codeparrot_training - Step 49013: {'lr': 5.214474604683384e-07, 'samples': 25095168, 'steps': 49013, 'loss/train': 2.02289080619812}
+02/26/2022 17:04:02 - INFO - codeparrot_training - Step 49014: {'lr': 5.203917311587258e-07, 'samples': 25095680, 'steps': 49014, 'loss/train': 1.2984391450881958}
+02/26/2022 17:04:05 - INFO - codeparrot_training - Step 49015: {'lr': 5.193370705404988e-07, 'samples': 25096192, 'steps': 49015, 'loss/train': 1.4447349309921265}
+02/26/2022 17:04:11 - INFO - codeparrot_training - Step 49016: {'lr': 5.182834786182366e-07, 'samples': 25096704, 'steps': 49016, 'loss/train': 2.0010018348693848}
+02/26/2022 17:04:14 - INFO - codeparrot_training - Step 49017: {'lr': 5.17230955396436e-07, 'samples': 25097216, 'steps': 49017, 'loss/train': 0.18538343906402588}
+02/26/2022 17:04:20 - INFO - codeparrot_training - Step 49018: {'lr': 5.161795008795656e-07, 'samples': 25097728, 'steps': 49018, 'loss/train': 1.284847378730774}
+02/26/2022 17:04:23 - INFO - codeparrot_training - Step 49019: {'lr': 5.151291150721771e-07, 'samples': 25098240, 'steps': 49019, 'loss/train': 1.2908810377120972}
+02/26/2022 17:04:30 - INFO - codeparrot_training - Step 49020: {'lr': 5.140797979787671e-07, 'samples': 25098752, 'steps': 49020, 'loss/train': 8.608312606811523}
+02/26/2022 17:04:33 - INFO - codeparrot_training - Step 49021: {'lr': 5.13031549603804e-07, 'samples': 25099264, 'steps': 49021, 'loss/train': 1.6041030883789062}
+02/26/2022 17:04:39 - INFO - codeparrot_training - Step 49022: {'lr': 5.119843699518123e-07, 'samples': 25099776, 'steps': 49022, 'loss/train': 1.7628525495529175}
+02/26/2022 17:04:42 - INFO - codeparrot_training - Step 49023: {'lr': 5.109382590272327e-07, 'samples': 25100288, 'steps': 49023, 'loss/train': 1.6956602334976196}
+02/26/2022 17:04:48 - INFO - codeparrot_training - Step 49024: {'lr': 5.098932168345893e-07, 'samples': 25100800, 'steps': 49024, 'loss/train': 3.1024675369262695}
+02/26/2022 17:04:51 - INFO - codeparrot_training - Step 49025: {'lr': 5.088492433783509e-07, 'samples': 25101312, 'steps': 49025, 'loss/train': 1.6666789054870605}
+02/26/2022 17:04:57 - INFO - codeparrot_training - Step 49026: {'lr': 5.078063386629583e-07, 'samples': 25101824, 'steps': 49026, 'loss/train': 1.6158421039581299}
+02/26/2022 17:05:00 - INFO - codeparrot_training - Step 49027: {'lr': 5.067645026929079e-07, 'samples': 25102336, 'steps': 49027, 'loss/train': 1.9067306518554688}
+02/26/2022 17:05:06 - INFO - codeparrot_training - Step 49028: {'lr': 5.057237354726685e-07, 'samples': 25102848, 'steps': 49028, 'loss/train': 1.9076614379882812}
+02/26/2022 17:05:09 - INFO - codeparrot_training - Step 49029: {'lr': 5.04684037006653e-07, 'samples': 25103360, 'steps': 49029, 'loss/train': 1.5937572717666626}
+02/26/2022 17:05:15 - INFO - codeparrot_training - Step 49030: {'lr': 5.036454072993857e-07, 'samples': 25103872, 'steps': 49030, 'loss/train': 1.298364281654358}
+02/26/2022 17:05:18 - INFO - codeparrot_training - Step 49031: {'lr': 5.026078463552519e-07, 'samples': 25104384, 'steps': 49031, 'loss/train': 1.7518583536148071}
+02/26/2022 17:05:24 - INFO - codeparrot_training - Step 49032: {'lr': 5.015713541787204e-07, 'samples': 25104896, 'steps': 49032, 'loss/train': 2.3967034816741943}
+02/26/2022 17:05:27 - INFO - codeparrot_training - Step 49033: {'lr': 5.005359307742319e-07, 'samples': 25105408, 'steps': 49033, 'loss/train': 1.2391053438186646}
+02/26/2022 17:05:33 - INFO - codeparrot_training - Step 49034: {'lr': 4.995015761462551e-07, 'samples': 25105920, 'steps': 49034, 'loss/train': 1.8676375150680542}
+02/26/2022 17:05:37 - INFO - codeparrot_training - Step 49035: {'lr': 4.984682902991477e-07, 'samples': 25106432, 'steps': 49035, 'loss/train': 0.9794620871543884}
+02/26/2022 17:05:42 - INFO - codeparrot_training - Step 49036: {'lr': 4.974360732374062e-07, 'samples': 25106944, 'steps': 49036, 'loss/train': 0.07919516414403915}
+02/26/2022 17:05:46 - INFO - codeparrot_training - Step 49037: {'lr': 4.964049249654156e-07, 'samples': 25107456, 'steps': 49037, 'loss/train': 3.1983420848846436}
+02/26/2022 17:05:51 - INFO - codeparrot_training - Step 49038: {'lr': 4.953748454875895e-07, 'samples': 25107968, 'steps': 49038, 'loss/train': 1.0089688301086426}
+02/26/2022 17:05:55 - INFO - codeparrot_training - Step 49039: {'lr': 4.943458348083407e-07, 'samples': 25108480, 'steps': 49039, 'loss/train': 1.2593165636062622}
+02/26/2022 17:06:00 - INFO - codeparrot_training - Step 49040: {'lr': 4.933178929321103e-07, 'samples': 25108992, 'steps': 49040, 'loss/train': 0.506611704826355}
+02/26/2022 17:06:04 - INFO - codeparrot_training - Step 49041: {'lr': 4.922910198632557e-07, 'samples': 25109504, 'steps': 49041, 'loss/train': 2.087669610977173}
+02/26/2022 17:06:09 - INFO - codeparrot_training - Step 49042: {'lr': 4.912652156062181e-07, 'samples': 25110016, 'steps': 49042, 'loss/train': 1.889102578163147}
+02/26/2022 17:06:13 - INFO - codeparrot_training - Step 49043: {'lr': 4.902404801653826e-07, 'samples': 25110528, 'steps': 49043, 'loss/train': 2.149672746658325}
+02/26/2022 17:06:18 - INFO - codeparrot_training - Step 49044: {'lr': 4.892168135451069e-07, 'samples': 25111040, 'steps': 49044, 'loss/train': 0.6079110503196716}
+02/26/2022 17:06:22 - INFO - codeparrot_training - Step 49045: {'lr': 4.881942157498043e-07, 'samples': 25111552, 'steps': 49045, 'loss/train': 0.18009762465953827}
+02/26/2022 17:06:28 - INFO - codeparrot_training - Step 49046: {'lr': 4.8717268678386e-07, 'samples': 25112064, 'steps': 49046, 'loss/train': 1.4845880270004272}
+02/26/2022 17:06:31 - INFO - codeparrot_training - Step 49047: {'lr': 4.861522266516316e-07, 'samples': 25112576, 'steps': 49047, 'loss/train': 0.9136321544647217}
+02/26/2022 17:06:37 - INFO - codeparrot_training - Step 49048: {'lr': 4.851328353574768e-07, 'samples': 25113088, 'steps': 49048, 'loss/train': 2.304622173309326}
+02/26/2022 17:06:40 - INFO - codeparrot_training - Step 49049: {'lr': 4.841145129058088e-07, 'samples': 25113600, 'steps': 49049, 'loss/train': 2.046469211578369}
+02/26/2022 17:06:46 - INFO - codeparrot_training - Step 49050: {'lr': 4.830972593009575e-07, 'samples': 25114112, 'steps': 49050, 'loss/train': 2.2296531200408936}
+02/26/2022 17:06:49 - INFO - codeparrot_training - Step 49051: {'lr': 4.820810745472803e-07, 'samples': 25114624, 'steps': 49051, 'loss/train': 1.6871041059494019}
+02/26/2022 17:06:55 - INFO - codeparrot_training - Step 49052: {'lr': 4.81065958649135e-07, 'samples': 25115136, 'steps': 49052, 'loss/train': 1.5846202373504639}
+02/26/2022 17:06:58 - INFO - codeparrot_training - Step 49053: {'lr': 4.800519116108792e-07, 'samples': 25115648, 'steps': 49053, 'loss/train': 1.346282958984375}
+02/26/2022 17:07:04 - INFO - codeparrot_training - Step 49054: {'lr': 4.790389334368428e-07, 'samples': 25116160, 'steps': 49054, 'loss/train': 1.1217204332351685}
+02/26/2022 17:07:07 - INFO - codeparrot_training - Step 49055: {'lr': 4.780270241313833e-07, 'samples': 25116672, 'steps': 49055, 'loss/train': 1.8381842374801636}
+02/26/2022 17:07:13 - INFO - codeparrot_training - Step 49056: {'lr': 4.770161836988308e-07, 'samples': 25117184, 'steps': 49056, 'loss/train': 1.8852794170379639}
+02/26/2022 17:07:17 - INFO - codeparrot_training - Step 49057: {'lr': 4.7600641214348706e-07, 'samples': 25117696, 'steps': 49057, 'loss/train': 1.2660834789276123}
+02/26/2022 17:07:22 - INFO - codeparrot_training - Step 49058: {'lr': 4.7499770946971e-07, 'samples': 25118208, 'steps': 49058, 'loss/train': 1.4219340085983276}
+02/26/2022 17:07:26 - INFO - codeparrot_training - Step 49059: {'lr': 4.7399007568180163e-07, 'samples': 25118720, 'steps': 49059, 'loss/train': 1.187503695487976}
+02/26/2022 17:07:31 - INFO - codeparrot_training - Step 49060: {'lr': 4.7298351078409185e-07, 'samples': 25119232, 'steps': 49060, 'loss/train': 1.719525694847107}
+02/26/2022 17:07:35 - INFO - codeparrot_training - Step 49061: {'lr': 4.7197801478088276e-07, 'samples': 25119744, 'steps': 49061, 'loss/train': 1.1388719081878662}
+02/26/2022 17:07:41 - INFO - codeparrot_training - Step 49062: {'lr': 4.7097358767650423e-07, 'samples': 25120256, 'steps': 49062, 'loss/train': 1.2376247644424438}
+02/26/2022 17:07:44 - INFO - codeparrot_training - Step 49063: {'lr': 4.699702294752306e-07, 'samples': 25120768, 'steps': 49063, 'loss/train': 2.1665024757385254}
+02/26/2022 17:07:50 - INFO - codeparrot_training - Step 49064: {'lr': 4.6896794018136404e-07, 'samples': 25121280, 'steps': 49064, 'loss/train': 2.526989221572876}
+02/26/2022 17:07:53 - INFO - codeparrot_training - Step 49065: {'lr': 4.679667197992066e-07, 'samples': 25121792, 'steps': 49065, 'loss/train': 2.167402982711792}
+02/26/2022 17:07:59 - INFO - codeparrot_training - Step 49066: {'lr': 4.6696656833303265e-07, 'samples': 25122304, 'steps': 49066, 'loss/train': 0.0786314532160759}
+02/26/2022 17:08:03 - INFO - codeparrot_training - Step 49067: {'lr': 4.659674857871443e-07, 'samples': 25122816, 'steps': 49067, 'loss/train': 1.325038194656372}
+02/26/2022 17:08:08 - INFO - codeparrot_training - Step 49068: {'lr': 4.6496947216581597e-07, 'samples': 25123328, 'steps': 49068, 'loss/train': 1.980177402496338}
+02/26/2022 17:08:12 - INFO - codeparrot_training - Step 49069: {'lr': 4.6397252747332196e-07, 'samples': 25123840, 'steps': 49069, 'loss/train': 1.6689904928207397}
+02/26/2022 17:08:17 - INFO - codeparrot_training - Step 49070: {'lr': 4.6297665171393666e-07, 'samples': 25124352, 'steps': 49070, 'loss/train': 1.5461705923080444}
+02/26/2022 17:08:21 - INFO - codeparrot_training - Step 49071: {'lr': 4.619818448919344e-07, 'samples': 25124864, 'steps': 49071, 'loss/train': 3.0336387157440186}
+02/26/2022 17:08:26 - INFO - codeparrot_training - Step 49072: {'lr': 4.609881070115618e-07, 'samples': 25125376, 'steps': 49072, 'loss/train': 1.640491008758545}
+02/26/2022 17:08:30 - INFO - codeparrot_training - Step 49073: {'lr': 4.5999543807706545e-07, 'samples': 25125888, 'steps': 49073, 'loss/train': 1.6270902156829834}
+02/26/2022 17:08:35 - INFO - codeparrot_training - Step 49074: {'lr': 4.5900383809271974e-07, 'samples': 25126400, 'steps': 49074, 'loss/train': 1.3789494037628174}
+02/26/2022 17:08:39 - INFO - codeparrot_training - Step 49075: {'lr': 4.580133070627435e-07, 'samples': 25126912, 'steps': 49075, 'loss/train': 1.1285934448242188}
+02/26/2022 17:08:44 - INFO - codeparrot_training - Step 49076: {'lr': 4.5702384499141106e-07, 'samples': 25127424, 'steps': 49076, 'loss/train': 0.973630964756012}
+02/26/2022 17:08:48 - INFO - codeparrot_training - Step 49077: {'lr': 4.56035451882969e-07, 'samples': 25127936, 'steps': 49077, 'loss/train': 2.1530659198760986}
+02/26/2022 17:08:53 - INFO - codeparrot_training - Step 49078: {'lr': 4.5504812774160855e-07, 'samples': 25128448, 'steps': 49078, 'loss/train': 1.0737115144729614}
+02/26/2022 17:08:57 - INFO - codeparrot_training - Step 49079: {'lr': 4.540618725716039e-07, 'samples': 25128960, 'steps': 49079, 'loss/train': 2.057705879211426}
+02/26/2022 17:09:02 - INFO - codeparrot_training - Step 49080: {'lr': 4.530766863771185e-07, 'samples': 25129472, 'steps': 49080, 'loss/train': 1.524463176727295}
+02/26/2022 17:09:06 - INFO - codeparrot_training - Step 49081: {'lr': 4.5209256916245443e-07, 'samples': 25129984, 'steps': 49081, 'loss/train': 1.3854734897613525}
+02/26/2022 17:09:12 - INFO - codeparrot_training - Step 49082: {'lr': 4.51109520931775e-07, 'samples': 25130496, 'steps': 49082, 'loss/train': 1.759461522102356}
+02/26/2022 17:09:16 - INFO - codeparrot_training - Step 49083: {'lr': 4.5012754168929916e-07, 'samples': 25131008, 'steps': 49083, 'loss/train': 1.9355318546295166}
+02/26/2022 17:09:21 - INFO - codeparrot_training - Step 49084: {'lr': 4.4914663143921785e-07, 'samples': 25131520, 'steps': 49084, 'loss/train': 1.1921107769012451}
+02/26/2022 17:09:25 - INFO - codeparrot_training - Step 49085: {'lr': 4.4816679018577775e-07, 'samples': 25132032, 'steps': 49085, 'loss/train': 2.0256187915802}
+02/26/2022 17:09:30 - INFO - codeparrot_training - Step 49086: {'lr': 4.4718801793314224e-07, 'samples': 25132544, 'steps': 49086, 'loss/train': 0.46422213315963745}
+02/26/2022 17:09:34 - INFO - codeparrot_training - Step 49087: {'lr': 4.4621031468550233e-07, 'samples': 25133056, 'steps': 49087, 'loss/train': 1.410831093788147}
+02/26/2022 17:09:39 - INFO - codeparrot_training - Step 49088: {'lr': 4.4523368044704915e-07, 'samples': 25133568, 'steps': 49088, 'loss/train': 1.8755733966827393}
+02/26/2022 17:09:43 - INFO - codeparrot_training - Step 49089: {'lr': 4.4425811522200154e-07, 'samples': 25134080, 'steps': 49089, 'loss/train': 0.9011662602424622}
+02/26/2022 17:09:48 - INFO - codeparrot_training - Step 49090: {'lr': 4.432836190144951e-07, 'samples': 25134592, 'steps': 49090, 'loss/train': 0.1797219067811966}
+02/26/2022 17:09:52 - INFO - codeparrot_training - Step 49091: {'lr': 4.4231019182869315e-07, 'samples': 25135104, 'steps': 49091, 'loss/train': 1.9175300598144531}
+02/26/2022 17:09:58 - INFO - codeparrot_training - Step 49092: {'lr': 4.4133783366881455e-07, 'samples': 25135616, 'steps': 49092, 'loss/train': 0.7467092275619507}
+02/26/2022 17:10:01 - INFO - codeparrot_training - Step 49093: {'lr': 4.4036654453899483e-07, 'samples': 25136128, 'steps': 49093, 'loss/train': 1.8505994081497192}
+02/26/2022 17:10:07 - INFO - codeparrot_training - Step 49094: {'lr': 4.393963244434251e-07, 'samples': 25136640, 'steps': 49094, 'loss/train': 1.1297352313995361}
+02/26/2022 17:10:10 - INFO - codeparrot_training - Step 49095: {'lr': 4.384271733862133e-07, 'samples': 25137152, 'steps': 49095, 'loss/train': 0.15467067062854767}
+02/26/2022 17:10:16 - INFO - codeparrot_training - Step 49096: {'lr': 4.374590913715226e-07, 'samples': 25137664, 'steps': 49096, 'loss/train': 0.8506312370300293}
+02/26/2022 17:10:19 - INFO - codeparrot_training - Step 49097: {'lr': 4.364920784035442e-07, 'samples': 25138176, 'steps': 49097, 'loss/train': 1.478582501411438}
+02/26/2022 17:10:25 - INFO - codeparrot_training - Step 49098: {'lr': 4.3552613448635815e-07, 'samples': 25138688, 'steps': 49098, 'loss/train': 1.641661286354065}
+02/26/2022 17:10:28 - INFO - codeparrot_training - Step 49099: {'lr': 4.345612596241555e-07, 'samples': 25139200, 'steps': 49099, 'loss/train': 2.45971941947937}
+02/26/2022 17:10:34 - INFO - codeparrot_training - Step 49100: {'lr': 4.335974538210441e-07, 'samples': 25139712, 'steps': 49100, 'loss/train': 1.9928321838378906}
+02/26/2022 17:10:37 - INFO - codeparrot_training - Step 49101: {'lr': 4.3263471708113177e-07, 'samples': 25140224, 'steps': 49101, 'loss/train': 1.9004757404327393}
+02/26/2022 17:10:43 - INFO - codeparrot_training - Step 49102: {'lr': 4.3167304940858185e-07, 'samples': 25140736, 'steps': 49102, 'loss/train': 1.4035385847091675}
+02/26/2022 17:10:49 - INFO - codeparrot_training - Step 49103: {'lr': 4.307124508075022e-07, 'samples': 25141248, 'steps': 49103, 'loss/train': 1.7648118734359741}
+02/26/2022 17:10:52 - INFO - codeparrot_training - Step 49104: {'lr': 4.297529212820006e-07, 'samples': 25141760, 'steps': 49104, 'loss/train': 1.4929388761520386}
+02/26/2022 17:10:57 - INFO - codeparrot_training - Step 49105: {'lr': 4.287944608361849e-07, 'samples': 25142272, 'steps': 49105, 'loss/train': 2.0259857177734375}
+02/26/2022 17:11:01 - INFO - codeparrot_training - Step 49106: {'lr': 4.278370694741629e-07, 'samples': 25142784, 'steps': 49106, 'loss/train': 2.1846115589141846}
+02/26/2022 17:11:07 - INFO - codeparrot_training - Step 49107: {'lr': 4.268807472000424e-07, 'samples': 25143296, 'steps': 49107, 'loss/train': 1.0400158166885376}
+02/26/2022 17:11:10 - INFO - codeparrot_training - Step 49108: {'lr': 4.2592549401793133e-07, 'samples': 25143808, 'steps': 49108, 'loss/train': 1.0118441581726074}
+02/26/2022 17:11:16 - INFO - codeparrot_training - Step 49109: {'lr': 4.2497130993188194e-07, 'samples': 25144320, 'steps': 49109, 'loss/train': 1.5299532413482666}
+02/26/2022 17:11:19 - INFO - codeparrot_training - Step 49110: {'lr': 4.240181949460298e-07, 'samples': 25144832, 'steps': 49110, 'loss/train': 2.021423816680908}
+02/26/2022 17:11:25 - INFO - codeparrot_training - Step 49111: {'lr': 4.230661490644272e-07, 'samples': 25145344, 'steps': 49111, 'loss/train': 1.249069094657898}
+02/26/2022 17:11:28 - INFO - codeparrot_training - Step 49112: {'lr': 4.2211517229115427e-07, 'samples': 25145856, 'steps': 49112, 'loss/train': 3.2623772621154785}
+02/26/2022 17:11:36 - INFO - codeparrot_training - Step 49113: {'lr': 4.21165264630291e-07, 'samples': 25146368, 'steps': 49113, 'loss/train': 1.301159143447876}
+02/26/2022 17:11:39 - INFO - codeparrot_training - Step 49114: {'lr': 4.202164260858898e-07, 'samples': 25146880, 'steps': 49114, 'loss/train': 1.7586778402328491}
+02/26/2022 17:11:45 - INFO - codeparrot_training - Step 49115: {'lr': 4.192686566620585e-07, 'samples': 25147392, 'steps': 49115, 'loss/train': 2.4094693660736084}
+02/26/2022 17:11:48 - INFO - codeparrot_training - Step 49116: {'lr': 4.1832195636282154e-07, 'samples': 25147904, 'steps': 49116, 'loss/train': 0.9420658946037292}
+02/26/2022 17:11:54 - INFO - codeparrot_training - Step 49117: {'lr': 4.1737632519223133e-07, 'samples': 25148416, 'steps': 49117, 'loss/train': 1.0134227275848389}
+02/26/2022 17:11:57 - INFO - codeparrot_training - Step 49118: {'lr': 4.164317631543679e-07, 'samples': 25148928, 'steps': 49118, 'loss/train': 2.1292717456817627}
+02/26/2022 17:12:03 - INFO - codeparrot_training - Step 49119: {'lr': 4.154882702532281e-07, 'samples': 25149440, 'steps': 49119, 'loss/train': 1.5120207071304321}
+02/26/2022 17:12:06 - INFO - codeparrot_training - Step 49120: {'lr': 4.145458464929197e-07, 'samples': 25149952, 'steps': 49120, 'loss/train': 1.184522271156311}
+02/26/2022 17:12:12 - INFO - codeparrot_training - Step 49121: {'lr': 4.1360449187741177e-07, 'samples': 25150464, 'steps': 49121, 'loss/train': 1.8079391717910767}
+02/26/2022 17:12:15 - INFO - codeparrot_training - Step 49122: {'lr': 4.1266420641081215e-07, 'samples': 25150976, 'steps': 49122, 'loss/train': 1.9366484880447388}
+02/26/2022 17:12:21 - INFO - codeparrot_training - Step 49123: {'lr': 4.117249900970621e-07, 'samples': 25151488, 'steps': 49123, 'loss/train': 0.13338087499141693}
+02/26/2022 17:12:24 - INFO - codeparrot_training - Step 49124: {'lr': 4.107868429402417e-07, 'samples': 25152000, 'steps': 49124, 'loss/train': 2.923515796661377}
+02/26/2022 17:12:32 - INFO - codeparrot_training - Step 49125: {'lr': 4.0984976494437553e-07, 'samples': 25152512, 'steps': 49125, 'loss/train': 1.2058892250061035}
+02/26/2022 17:12:35 - INFO - codeparrot_training - Step 49126: {'lr': 4.089137561134604e-07, 'samples': 25153024, 'steps': 49126, 'loss/train': 0.6896197199821472}
+02/26/2022 17:12:41 - INFO - codeparrot_training - Step 49127: {'lr': 4.079788164514653e-07, 'samples': 25153536, 'steps': 49127, 'loss/train': 0.6483401656150818}
+02/26/2022 17:12:44 - INFO - codeparrot_training - Step 49128: {'lr': 4.070449459624703e-07, 'samples': 25154048, 'steps': 49128, 'loss/train': 1.4917714595794678}
+02/26/2022 17:12:50 - INFO - codeparrot_training - Step 49129: {'lr': 4.0611214465044456e-07, 'samples': 25154560, 'steps': 49129, 'loss/train': 0.13677482306957245}
+02/26/2022 17:12:53 - INFO - codeparrot_training - Step 49130: {'lr': 4.05180412519357e-07, 'samples': 25155072, 'steps': 49130, 'loss/train': 2.2253036499023438}
+02/26/2022 17:12:59 - INFO - codeparrot_training - Step 49131: {'lr': 4.042497495732322e-07, 'samples': 25155584, 'steps': 49131, 'loss/train': 2.3909530639648438}
+02/26/2022 17:13:02 - INFO - codeparrot_training - Step 49132: {'lr': 4.033201558160393e-07, 'samples': 25156096, 'steps': 49132, 'loss/train': 2.2927629947662354}
+02/26/2022 17:13:08 - INFO - codeparrot_training - Step 49133: {'lr': 4.02391631251775e-07, 'samples': 25156608, 'steps': 49133, 'loss/train': 1.9030135869979858}
+02/26/2022 17:13:11 - INFO - codeparrot_training - Step 49134: {'lr': 4.0146417588440844e-07, 'samples': 25157120, 'steps': 49134, 'loss/train': 0.7534550428390503}
+02/26/2022 17:13:19 - INFO - codeparrot_training - Step 49135: {'lr': 4.0053778971790853e-07, 'samples': 25157632, 'steps': 49135, 'loss/train': 2.359708070755005}
+02/26/2022 17:13:22 - INFO - codeparrot_training - Step 49136: {'lr': 3.9961247275624445e-07, 'samples': 25158144, 'steps': 49136, 'loss/train': 2.3942036628723145}
+02/26/2022 17:13:28 - INFO - codeparrot_training - Step 49137: {'lr': 3.986882250033852e-07, 'samples': 25158656, 'steps': 49137, 'loss/train': 1.8561989068984985}
+02/26/2022 17:13:31 - INFO - codeparrot_training - Step 49138: {'lr': 3.977650464632998e-07, 'samples': 25159168, 'steps': 49138, 'loss/train': 0.273765504360199}
+02/26/2022 17:13:36 - INFO - codeparrot_training - Step 49139: {'lr': 3.968429371399296e-07, 'samples': 25159680, 'steps': 49139, 'loss/train': 2.343970775604248}
+02/26/2022 17:13:40 - INFO - codeparrot_training - Step 49140: {'lr': 3.9592189703718804e-07, 'samples': 25160192, 'steps': 49140, 'loss/train': 1.297652244567871}
+02/26/2022 17:13:46 - INFO - codeparrot_training - Step 49141: {'lr': 3.950019261590998e-07, 'samples': 25160704, 'steps': 49141, 'loss/train': 1.8379788398742676}
+02/26/2022 17:13:49 - INFO - codeparrot_training - Step 49142: {'lr': 3.9408302450955057e-07, 'samples': 25161216, 'steps': 49142, 'loss/train': 2.3740673065185547}
+02/26/2022 17:13:55 - INFO - codeparrot_training - Step 49143: {'lr': 3.9316519209250944e-07, 'samples': 25161728, 'steps': 49143, 'loss/train': 1.849450945854187}
+02/26/2022 17:13:58 - INFO - codeparrot_training - Step 49144: {'lr': 3.922484289118622e-07, 'samples': 25162240, 'steps': 49144, 'loss/train': 1.4418665170669556}
+02/26/2022 17:14:05 - INFO - codeparrot_training - Step 49145: {'lr': 3.9133273497157783e-07, 'samples': 25162752, 'steps': 49145, 'loss/train': 1.8954391479492188}
+02/26/2022 17:14:09 - INFO - codeparrot_training - Step 49146: {'lr': 3.9041811027556994e-07, 'samples': 25163264, 'steps': 49146, 'loss/train': 2.1025638580322266}
+02/26/2022 17:14:14 - INFO - codeparrot_training - Step 49147: {'lr': 3.8950455482775203e-07, 'samples': 25163776, 'steps': 49147, 'loss/train': 1.0569565296173096}
+02/26/2022 17:14:18 - INFO - codeparrot_training - Step 49148: {'lr': 3.885920686320099e-07, 'samples': 25164288, 'steps': 49148, 'loss/train': 1.853453278541565}
+02/26/2022 17:14:23 - INFO - codeparrot_training - Step 49149: {'lr': 3.876806516923126e-07, 'samples': 25164800, 'steps': 49149, 'loss/train': 0.3836728632450104}
+02/26/2022 17:14:27 - INFO - codeparrot_training - Step 49150: {'lr': 3.8677030401249035e-07, 'samples': 25165312, 'steps': 49150, 'loss/train': 2.1091792583465576}
+02/26/2022 17:14:32 - INFO - codeparrot_training - Step 49151: {'lr': 3.8586102559651226e-07, 'samples': 25165824, 'steps': 49151, 'loss/train': 0.9523171186447144}
+02/26/2022 17:14:36 - INFO - codeparrot_training - Step 49152: {'lr': 3.849528164482363e-07, 'samples': 25166336, 'steps': 49152, 'loss/train': 2.497467517852783}
+02/26/2022 17:14:41 - INFO - codeparrot_training - Step 49153: {'lr': 3.8404567657157607e-07, 'samples': 25166848, 'steps': 49153, 'loss/train': 1.6794003248214722}
+02/26/2022 17:14:45 - INFO - codeparrot_training - Step 49154: {'lr': 3.8313960597038956e-07, 'samples': 25167360, 'steps': 49154, 'loss/train': 1.7243382930755615}
+02/26/2022 17:14:52 - INFO - codeparrot_training - Step 49155: {'lr': 3.822346046485625e-07, 'samples': 25167872, 'steps': 49155, 'loss/train': 1.6368248462677002}
+02/26/2022 17:14:55 - INFO - codeparrot_training - Step 49156: {'lr': 3.8133067260998076e-07, 'samples': 25168384, 'steps': 49156, 'loss/train': 1.7094802856445312}
+02/26/2022 17:15:01 - INFO - codeparrot_training - Step 49157: {'lr': 3.804278098585301e-07, 'samples': 25168896, 'steps': 49157, 'loss/train': 6.278919696807861}
+02/26/2022 17:15:04 - INFO - codeparrot_training - Step 49158: {'lr': 3.795260163980685e-07, 'samples': 25169408, 'steps': 49158, 'loss/train': 2.52380633354187}
+02/26/2022 17:15:10 - INFO - codeparrot_training - Step 49159: {'lr': 3.7862529223242623e-07, 'samples': 25169920, 'steps': 49159, 'loss/train': 2.5490052700042725}
+02/26/2022 17:15:13 - INFO - codeparrot_training - Step 49160: {'lr': 3.777256373655169e-07, 'samples': 25170432, 'steps': 49160, 'loss/train': 2.0267715454101562}
+02/26/2022 17:15:19 - INFO - codeparrot_training - Step 49161: {'lr': 3.7682705180117074e-07, 'samples': 25170944, 'steps': 49161, 'loss/train': 1.6364741325378418}
+02/26/2022 17:15:22 - INFO - codeparrot_training - Step 49162: {'lr': 3.759295355431902e-07, 'samples': 25171456, 'steps': 49162, 'loss/train': 1.6684178113937378}
+02/26/2022 17:15:28 - INFO - codeparrot_training - Step 49163: {'lr': 3.7503308859551665e-07, 'samples': 25171968, 'steps': 49163, 'loss/train': 1.5207043886184692}
+02/26/2022 17:15:31 - INFO - codeparrot_training - Step 49164: {'lr': 3.7413771096189707e-07, 'samples': 25172480, 'steps': 49164, 'loss/train': 2.12703537940979}
+02/26/2022 17:15:37 - INFO - codeparrot_training - Step 49165: {'lr': 3.7324340264621724e-07, 'samples': 25172992, 'steps': 49165, 'loss/train': 1.8947240114212036}
+02/26/2022 17:15:40 - INFO - codeparrot_training - Step 49166: {'lr': 3.7235016365230746e-07, 'samples': 25173504, 'steps': 49166, 'loss/train': 1.1667351722717285}
+02/26/2022 17:15:46 - INFO - codeparrot_training - Step 49167: {'lr': 3.7145799398397016e-07, 'samples': 25174016, 'steps': 49167, 'loss/train': 2.478134870529175}
+02/26/2022 17:15:50 - INFO - codeparrot_training - Step 49168: {'lr': 3.7056689364503573e-07, 'samples': 25174528, 'steps': 49168, 'loss/train': 1.4744770526885986}
+02/26/2022 17:15:57 - INFO - codeparrot_training - Step 49169: {'lr': 3.696768626393343e-07, 'samples': 25175040, 'steps': 49169, 'loss/train': 0.8081668615341187}
+02/26/2022 17:16:01 - INFO - codeparrot_training - Step 49170: {'lr': 3.687879009706963e-07, 'samples': 25175552, 'steps': 49170, 'loss/train': 2.199540615081787}
+02/26/2022 17:16:06 - INFO - codeparrot_training - Step 49171: {'lr': 3.6790000864286855e-07, 'samples': 25176064, 'steps': 49171, 'loss/train': 1.9282736778259277}
+02/26/2022 17:16:10 - INFO - codeparrot_training - Step 49172: {'lr': 3.670131856597092e-07, 'samples': 25176576, 'steps': 49172, 'loss/train': 1.5354024171829224}
+02/26/2022 17:16:15 - INFO - codeparrot_training - Step 49173: {'lr': 3.66127432024993e-07, 'samples': 25177088, 'steps': 49173, 'loss/train': 2.1589131355285645}
+02/26/2022 17:16:19 - INFO - codeparrot_training - Step 49174: {'lr': 3.6524274774252244e-07, 'samples': 25177600, 'steps': 49174, 'loss/train': 2.42380952835083}
+02/26/2022 17:16:24 - INFO - codeparrot_training - Step 49175: {'lr': 3.6435913281607224e-07, 'samples': 25178112, 'steps': 49175, 'loss/train': 2.2026307582855225}
+02/26/2022 17:16:28 - INFO - codeparrot_training - Step 49176: {'lr': 3.63476587249445e-07, 'samples': 25178624, 'steps': 49176, 'loss/train': 1.486001968383789}
+02/26/2022 17:16:34 - INFO - codeparrot_training - Step 49177: {'lr': 3.625951110464154e-07, 'samples': 25179136, 'steps': 49177, 'loss/train': 1.705509901046753}
+02/26/2022 17:16:37 - INFO - codeparrot_training - Step 49178: {'lr': 3.6171470421075825e-07, 'samples': 25179648, 'steps': 49178, 'loss/train': 1.1683207750320435}
+02/26/2022 17:16:43 - INFO - codeparrot_training - Step 49179: {'lr': 3.608353667462483e-07, 'samples': 25180160, 'steps': 49179, 'loss/train': 0.4662502706050873}
+02/26/2022 17:16:46 - INFO - codeparrot_training - Step 49180: {'lr': 3.599570986566603e-07, 'samples': 25180672, 'steps': 49180, 'loss/train': 0.8370378613471985}
+02/26/2022 17:16:54 - INFO - codeparrot_training - Step 49181: {'lr': 3.590798999457412e-07, 'samples': 25181184, 'steps': 49181, 'loss/train': 1.2941521406173706}
+02/26/2022 17:16:57 - INFO - codeparrot_training - Step 49182: {'lr': 3.582037706172381e-07, 'samples': 25181696, 'steps': 49182, 'loss/train': 1.1602590084075928}
+02/26/2022 17:17:03 - INFO - codeparrot_training - Step 49183: {'lr': 3.573287106749534e-07, 'samples': 25182208, 'steps': 49183, 'loss/train': 1.65255868434906}
+02/26/2022 17:17:06 - INFO - codeparrot_training - Step 49184: {'lr': 3.5645472012257876e-07, 'samples': 25182720, 'steps': 49184, 'loss/train': 2.339512586593628}
+02/26/2022 17:17:10 - INFO - codeparrot_training - Step 49185: {'lr': 3.55581798963861e-07, 'samples': 25183232, 'steps': 49185, 'loss/train': 1.23650062084198}
+02/26/2022 17:17:15 - INFO - codeparrot_training - Step 49186: {'lr': 3.5470994720260276e-07, 'samples': 25183744, 'steps': 49186, 'loss/train': 1.4002877473831177}
+02/26/2022 17:17:19 - INFO - codeparrot_training - Step 49187: {'lr': 3.538391648424677e-07, 'samples': 25184256, 'steps': 49187, 'loss/train': 0.95508873462677}
+02/26/2022 17:17:24 - INFO - codeparrot_training - Step 49188: {'lr': 3.5296945188723063e-07, 'samples': 25184768, 'steps': 49188, 'loss/train': 2.977891683578491}
+02/26/2022 17:17:30 - INFO - codeparrot_training - Step 49189: {'lr': 3.52100808340583e-07, 'samples': 25185280, 'steps': 49189, 'loss/train': 2.0337424278259277}
+02/26/2022 17:17:33 - INFO - codeparrot_training - Step 49190: {'lr': 3.5123323420627183e-07, 'samples': 25185792, 'steps': 49190, 'loss/train': 1.6148189306259155}
+02/26/2022 17:17:41 - INFO - codeparrot_training - Step 49191: {'lr': 3.503667294879886e-07, 'samples': 25186304, 'steps': 49191, 'loss/train': 2.1883771419525146}
+02/26/2022 17:17:44 - INFO - codeparrot_training - Step 49192: {'lr': 3.495012941894804e-07, 'samples': 25186816, 'steps': 49192, 'loss/train': 1.573521375656128}
+02/26/2022 17:17:50 - INFO - codeparrot_training - Step 49193: {'lr': 3.486369283144386e-07, 'samples': 25187328, 'steps': 49193, 'loss/train': 2.792530059814453}
+02/26/2022 17:17:53 - INFO - codeparrot_training - Step 49194: {'lr': 3.4777363186655477e-07, 'samples': 25187840, 'steps': 49194, 'loss/train': 0.9519041180610657}
+02/26/2022 17:17:59 - INFO - codeparrot_training - Step 49195: {'lr': 3.469114048495481e-07, 'samples': 25188352, 'steps': 49195, 'loss/train': 1.179364800453186}
+02/26/2022 17:18:02 - INFO - codeparrot_training - Step 49196: {'lr': 3.4605024726708235e-07, 'samples': 25188864, 'steps': 49196, 'loss/train': 2.2198383808135986}
+02/26/2022 17:18:07 - INFO - codeparrot_training - Step 49197: {'lr': 3.4519015912287677e-07, 'samples': 25189376, 'steps': 49197, 'loss/train': 1.5995867252349854}
+02/26/2022 17:18:11 - INFO - codeparrot_training - Step 49198: {'lr': 3.443311404205951e-07, 'samples': 25189888, 'steps': 49198, 'loss/train': 1.0336480140686035}
+02/26/2022 17:18:16 - INFO - codeparrot_training - Step 49199: {'lr': 3.4347319116390106e-07, 'samples': 25190400, 'steps': 49199, 'loss/train': 0.5650877952575684}
+02/26/2022 17:18:20 - INFO - codeparrot_training - Step 49200: {'lr': 3.426163113565417e-07, 'samples': 25190912, 'steps': 49200, 'loss/train': 1.613423228263855}
+02/26/2022 17:18:26 - INFO - codeparrot_training - Step 49201: {'lr': 3.4176050100209744e-07, 'samples': 25191424, 'steps': 49201, 'loss/train': 1.7087675333023071}
+02/26/2022 17:18:29 - INFO - codeparrot_training - Step 49202: {'lr': 3.409057601042875e-07, 'samples': 25191936, 'steps': 49202, 'loss/train': 2.2779128551483154}
+02/26/2022 17:18:35 - INFO - codeparrot_training - Step 49203: {'lr': 3.4005208866677574e-07, 'samples': 25192448, 'steps': 49203, 'loss/train': 2.2067148685455322}
+02/26/2022 17:18:38 - INFO - codeparrot_training - Step 49204: {'lr': 3.39199486693198e-07, 'samples': 25192960, 'steps': 49204, 'loss/train': 1.5900170803070068}
+02/26/2022 17:18:45 - INFO - codeparrot_training - Step 49205: {'lr': 3.383479541872181e-07, 'samples': 25193472, 'steps': 49205, 'loss/train': 0.923997163772583}
+02/26/2022 17:18:49 - INFO - codeparrot_training - Step 49206: {'lr': 3.3749749115247197e-07, 'samples': 25193984, 'steps': 49206, 'loss/train': 1.6662477254867554}
+02/26/2022 17:18:54 - INFO - codeparrot_training - Step 49207: {'lr': 3.3664809759259565e-07, 'samples': 25194496, 'steps': 49207, 'loss/train': 0.8580371141433716}
+02/26/2022 17:18:58 - INFO - codeparrot_training - Step 49208: {'lr': 3.357997735112528e-07, 'samples': 25195008, 'steps': 49208, 'loss/train': 1.279490351676941}
+02/26/2022 17:19:03 - INFO - codeparrot_training - Step 49209: {'lr': 3.3495251891207946e-07, 'samples': 25195520, 'steps': 49209, 'loss/train': 1.272603988647461}
+02/26/2022 17:19:07 - INFO - codeparrot_training - Step 49210: {'lr': 3.3410633379868383e-07, 'samples': 25196032, 'steps': 49210, 'loss/train': 2.558316946029663}
+02/26/2022 17:19:12 - INFO - codeparrot_training - Step 49211: {'lr': 3.332612181747019e-07, 'samples': 25196544, 'steps': 49211, 'loss/train': 1.1882816553115845}
+02/26/2022 17:19:16 - INFO - codeparrot_training - Step 49212: {'lr': 3.324171720437419e-07, 'samples': 25197056, 'steps': 49212, 'loss/train': 1.201464056968689}
+02/26/2022 17:19:21 - INFO - codeparrot_training - Step 49213: {'lr': 3.31574195409412e-07, 'samples': 25197568, 'steps': 49213, 'loss/train': 1.187307596206665}
+02/26/2022 17:19:25 - INFO - codeparrot_training - Step 49214: {'lr': 3.3073228827537606e-07, 'samples': 25198080, 'steps': 49214, 'loss/train': 2.3933403491973877}
+02/26/2022 17:19:30 - INFO - codeparrot_training - Step 49215: {'lr': 3.2989145064518665e-07, 'samples': 25198592, 'steps': 49215, 'loss/train': 1.3481743335723877}
+02/26/2022 17:19:33 - INFO - codeparrot_training - Step 49216: {'lr': 3.290516825224521e-07, 'samples': 25199104, 'steps': 49216, 'loss/train': 0.9960803389549255}
+02/26/2022 17:19:41 - INFO - codeparrot_training - Step 49217: {'lr': 3.2821298391080835e-07, 'samples': 25199616, 'steps': 49217, 'loss/train': 1.0293229818344116}
+02/26/2022 17:19:44 - INFO - codeparrot_training - Step 49218: {'lr': 3.2737535481380807e-07, 'samples': 25200128, 'steps': 49218, 'loss/train': 1.533575177192688}
+02/26/2022 17:19:50 - INFO - codeparrot_training - Step 49219: {'lr': 3.265387952350318e-07, 'samples': 25200640, 'steps': 49219, 'loss/train': 0.6177148222923279}
+02/26/2022 17:19:53 - INFO - codeparrot_training - Step 49220: {'lr': 3.2570330517811555e-07, 'samples': 25201152, 'steps': 49220, 'loss/train': 4.056741237640381}
+02/26/2022 17:19:59 - INFO - codeparrot_training - Step 49221: {'lr': 3.2486888464658413e-07, 'samples': 25201664, 'steps': 49221, 'loss/train': 1.9567835330963135}
+02/26/2022 17:20:02 - INFO - codeparrot_training - Step 49222: {'lr': 3.240355336440459e-07, 'samples': 25202176, 'steps': 49222, 'loss/train': 1.888270616531372}
+02/26/2022 17:20:08 - INFO - codeparrot_training - Step 49223: {'lr': 3.2320325217405354e-07, 'samples': 25202688, 'steps': 49223, 'loss/train': 1.583760380744934}
+02/26/2022 17:20:11 - INFO - codeparrot_training - Step 49224: {'lr': 3.2237204024015976e-07, 'samples': 25203200, 'steps': 49224, 'loss/train': 2.663724184036255}
+02/26/2022 17:20:17 - INFO - codeparrot_training - Step 49225: {'lr': 3.215418978459728e-07, 'samples': 25203712, 'steps': 49225, 'loss/train': 2.2093865871429443}
+02/26/2022 17:20:20 - INFO - codeparrot_training - Step 49226: {'lr': 3.207128249949898e-07, 'samples': 25204224, 'steps': 49226, 'loss/train': 0.3780570924282074}
+02/26/2022 17:20:27 - INFO - codeparrot_training - Step 49227: {'lr': 3.19884821690819e-07, 'samples': 25204736, 'steps': 49227, 'loss/train': 1.0628679990768433}
+02/26/2022 17:20:31 - INFO - codeparrot_training - Step 49228: {'lr': 3.190578879369577e-07, 'samples': 25205248, 'steps': 49228, 'loss/train': 1.1448765993118286}
+02/26/2022 17:20:36 - INFO - codeparrot_training - Step 49229: {'lr': 3.182320237369862e-07, 'samples': 25205760, 'steps': 49229, 'loss/train': 1.6691434383392334}
+02/26/2022 17:20:40 - INFO - codeparrot_training - Step 49230: {'lr': 3.174072290944019e-07, 'samples': 25206272, 'steps': 49230, 'loss/train': 1.346313714981079}
+02/26/2022 17:20:46 - INFO - codeparrot_training - Step 49231: {'lr': 3.165835040127851e-07, 'samples': 25206784, 'steps': 49231, 'loss/train': 1.961909532546997}
+02/26/2022 17:20:49 - INFO - codeparrot_training - Step 49232: {'lr': 3.1576084849563315e-07, 'samples': 25207296, 'steps': 49232, 'loss/train': 1.9196834564208984}
+02/26/2022 17:20:55 - INFO - codeparrot_training - Step 49233: {'lr': 3.149392625464709e-07, 'samples': 25207808, 'steps': 49233, 'loss/train': 0.5348854660987854}
+02/26/2022 17:20:58 - INFO - codeparrot_training - Step 49234: {'lr': 3.141187461688233e-07, 'samples': 25208320, 'steps': 49234, 'loss/train': 1.8401069641113281}
+02/26/2022 17:21:04 - INFO - codeparrot_training - Step 49235: {'lr': 3.1329929936621536e-07, 'samples': 25208832, 'steps': 49235, 'loss/train': 2.233855962753296}
+02/26/2022 17:21:08 - INFO - codeparrot_training - Step 49236: {'lr': 3.1248092214217206e-07, 'samples': 25209344, 'steps': 49236, 'loss/train': 2.24180006980896}
+02/26/2022 17:21:15 - INFO - codeparrot_training - Step 49237: {'lr': 3.11663614500135e-07, 'samples': 25209856, 'steps': 49237, 'loss/train': 1.7753978967666626}
+02/26/2022 17:21:18 - INFO - codeparrot_training - Step 49238: {'lr': 3.1084737644365703e-07, 'samples': 25210368, 'steps': 49238, 'loss/train': 1.365486741065979}
+02/26/2022 17:21:22 - INFO - codeparrot_training - Step 49239: {'lr': 3.100322079762352e-07, 'samples': 25210880, 'steps': 49239, 'loss/train': 2.2111904621124268}
+02/26/2022 17:21:27 - INFO - codeparrot_training - Step 49240: {'lr': 3.0921810910133906e-07, 'samples': 25211392, 'steps': 49240, 'loss/train': 1.0475375652313232}
+02/26/2022 17:21:31 - INFO - codeparrot_training - Step 49241: {'lr': 3.0840507982249354e-07, 'samples': 25211904, 'steps': 49241, 'loss/train': 1.9657444953918457}
+02/26/2022 17:21:36 - INFO - codeparrot_training - Step 49242: {'lr': 3.075931201431126e-07, 'samples': 25212416, 'steps': 49242, 'loss/train': 1.1375079154968262}
+02/26/2022 17:21:40 - INFO - codeparrot_training - Step 49243: {'lr': 3.067822300667489e-07, 'samples': 25212928, 'steps': 49243, 'loss/train': 1.6300870180130005}
+02/26/2022 17:21:45 - INFO - codeparrot_training - Step 49244: {'lr': 3.059724095968441e-07, 'samples': 25213440, 'steps': 49244, 'loss/train': 3.051119089126587}
+02/26/2022 17:21:49 - INFO - codeparrot_training - Step 49245: {'lr': 3.0516365873684004e-07, 'samples': 25213952, 'steps': 49245, 'loss/train': 1.5433343648910522}
+02/26/2022 17:21:54 - INFO - codeparrot_training - Step 49246: {'lr': 3.043559774902338e-07, 'samples': 25214464, 'steps': 49246, 'loss/train': 0.4882158041000366}
+02/26/2022 17:21:58 - INFO - codeparrot_training - Step 49247: {'lr': 3.0354936586049486e-07, 'samples': 25214976, 'steps': 49247, 'loss/train': 1.7533336877822876}
+02/26/2022 17:22:03 - INFO - codeparrot_training - Step 49248: {'lr': 3.027438238510649e-07, 'samples': 25215488, 'steps': 49248, 'loss/train': 1.7116378545761108}
+02/26/2022 17:22:07 - INFO - codeparrot_training - Step 49249: {'lr': 3.019393514653857e-07, 'samples': 25216000, 'steps': 49249, 'loss/train': 2.323669195175171}
+02/26/2022 17:22:12 - INFO - codeparrot_training - Step 49250: {'lr': 3.0113594870689873e-07, 'samples': 25216512, 'steps': 49250, 'loss/train': 1.2697666883468628}
+02/26/2022 17:22:20 - INFO - codeparrot_training - Step 49251: {'lr': 3.003336155790737e-07, 'samples': 25217024, 'steps': 49251, 'loss/train': 2.723059892654419}
+02/26/2022 17:22:23 - INFO - codeparrot_training - Step 49252: {'lr': 2.9953235208532437e-07, 'samples': 25217536, 'steps': 49252, 'loss/train': 1.6497389078140259}
+02/26/2022 17:22:28 - INFO - codeparrot_training - Step 49253: {'lr': 2.987321582290925e-07, 'samples': 25218048, 'steps': 49253, 'loss/train': 2.370401382446289}
+02/26/2022 17:22:32 - INFO - codeparrot_training - Step 49254: {'lr': 2.97933034013792e-07, 'samples': 25218560, 'steps': 49254, 'loss/train': 1.3297098875045776}
+02/26/2022 17:22:37 - INFO - codeparrot_training - Step 49255: {'lr': 2.971349794428646e-07, 'samples': 25219072, 'steps': 49255, 'loss/train': 1.72870934009552}
+02/26/2022 17:22:41 - INFO - codeparrot_training - Step 49256: {'lr': 2.963379945197242e-07, 'samples': 25219584, 'steps': 49256, 'loss/train': 1.5735093355178833}
+02/26/2022 17:22:47 - INFO - codeparrot_training - Step 49257: {'lr': 2.9554207924778475e-07, 'samples': 25220096, 'steps': 49257, 'loss/train': 1.419816017150879}
+02/26/2022 17:22:50 - INFO - codeparrot_training - Step 49258: {'lr': 2.947472336304324e-07, 'samples': 25220608, 'steps': 49258, 'loss/train': 0.5545970797538757}
+02/26/2022 17:22:56 - INFO - codeparrot_training - Step 49259: {'lr': 2.939534576711089e-07, 'samples': 25221120, 'steps': 49259, 'loss/train': 2.7279176712036133}
+02/26/2022 17:22:59 - INFO - codeparrot_training - Step 49260: {'lr': 2.931607513732004e-07, 'samples': 25221632, 'steps': 49260, 'loss/train': 1.7415560483932495}
+02/26/2022 17:23:03 - INFO - codeparrot_training - Step 49261: {'lr': 2.923691147400931e-07, 'samples': 25222144, 'steps': 49261, 'loss/train': 0.4381670355796814}
+02/26/2022 17:23:08 - INFO - codeparrot_training - Step 49262: {'lr': 2.915785477752009e-07, 'samples': 25222656, 'steps': 49262, 'loss/train': 2.497663736343384}
+02/26/2022 17:23:12 - INFO - codeparrot_training - Step 49263: {'lr': 2.907890504818822e-07, 'samples': 25223168, 'steps': 49263, 'loss/train': 1.6097497940063477}
+02/26/2022 17:23:19 - INFO - codeparrot_training - Step 49264: {'lr': 2.9000062286352326e-07, 'samples': 25223680, 'steps': 49264, 'loss/train': 1.1878303289413452}
+02/26/2022 17:23:22 - INFO - codeparrot_training - Step 49265: {'lr': 2.892132649235102e-07, 'samples': 25224192, 'steps': 49265, 'loss/train': 1.3831366300582886}
+02/26/2022 17:23:28 - INFO - codeparrot_training - Step 49266: {'lr': 2.884269766652292e-07, 'samples': 25224704, 'steps': 49266, 'loss/train': 2.0943593978881836}
+02/26/2022 17:23:31 - INFO - codeparrot_training - Step 49267: {'lr': 2.8764175809203873e-07, 'samples': 25225216, 'steps': 49267, 'loss/train': 2.1908011436462402}
+02/26/2022 17:23:37 - INFO - codeparrot_training - Step 49268: {'lr': 2.868576092072972e-07, 'samples': 25225728, 'steps': 49268, 'loss/train': 1.7559328079223633}
+02/26/2022 17:23:41 - INFO - codeparrot_training - Step 49269: {'lr': 2.860745300143353e-07, 'samples': 25226240, 'steps': 49269, 'loss/train': 1.9213505983352661}
+02/26/2022 17:23:46 - INFO - codeparrot_training - Step 49270: {'lr': 2.852925205165668e-07, 'samples': 25226752, 'steps': 49270, 'loss/train': 3.552980661392212}
+02/26/2022 17:23:50 - INFO - codeparrot_training - Step 49271: {'lr': 2.845115807172949e-07, 'samples': 25227264, 'steps': 49271, 'loss/train': 1.8593324422836304}
+02/26/2022 17:23:55 - INFO - codeparrot_training - Step 49272: {'lr': 2.8373171061990553e-07, 'samples': 25227776, 'steps': 49272, 'loss/train': 1.9358209371566772}
+02/26/2022 17:23:59 - INFO - codeparrot_training - Step 49273: {'lr': 2.829529102277018e-07, 'samples': 25228288, 'steps': 49273, 'loss/train': 0.28513431549072266}
+02/26/2022 17:24:06 - INFO - codeparrot_training - Step 49274: {'lr': 2.8217517954404194e-07, 'samples': 25228800, 'steps': 49274, 'loss/train': 2.081965684890747}
+02/26/2022 17:24:09 - INFO - codeparrot_training - Step 49275: {'lr': 2.81398518572229e-07, 'samples': 25229312, 'steps': 49275, 'loss/train': 1.3904247283935547}
+02/26/2022 17:24:15 - INFO - codeparrot_training - Step 49276: {'lr': 2.806229273156213e-07, 'samples': 25229824, 'steps': 49276, 'loss/train': 1.8676384687423706}
+02/26/2022 17:24:18 - INFO - codeparrot_training - Step 49277: {'lr': 2.798484057775219e-07, 'samples': 25230336, 'steps': 49277, 'loss/train': 0.11385231465101242}
+02/26/2022 17:24:24 - INFO - codeparrot_training - Step 49278: {'lr': 2.7907495396126137e-07, 'samples': 25230848, 'steps': 49278, 'loss/train': 1.4199883937835693}
+02/26/2022 17:24:28 - INFO - codeparrot_training - Step 49279: {'lr': 2.783025718701704e-07, 'samples': 25231360, 'steps': 49279, 'loss/train': 0.138031467795372}
+02/26/2022 17:24:33 - INFO - codeparrot_training - Step 49280: {'lr': 2.7753125950752413e-07, 'samples': 25231872, 'steps': 49280, 'loss/train': 1.5217351913452148}
+02/26/2022 17:24:37 - INFO - codeparrot_training - Step 49281: {'lr': 2.7676101687662545e-07, 'samples': 25232384, 'steps': 49281, 'loss/train': 2.1831204891204834}
+02/26/2022 17:24:42 - INFO - codeparrot_training - Step 49282: {'lr': 2.7599184398077737e-07, 'samples': 25232896, 'steps': 49282, 'loss/train': 0.7931455969810486}
+02/26/2022 17:24:46 - INFO - codeparrot_training - Step 49283: {'lr': 2.752237408233105e-07, 'samples': 25233408, 'steps': 49283, 'loss/train': 2.4793100357055664}
+02/26/2022 17:24:53 - INFO - codeparrot_training - Step 49284: {'lr': 2.744567074074722e-07, 'samples': 25233920, 'steps': 49284, 'loss/train': 0.9266887903213501}
+02/26/2022 17:24:56 - INFO - codeparrot_training - Step 49285: {'lr': 2.7369074373656546e-07, 'samples': 25234432, 'steps': 49285, 'loss/train': 1.0195916891098022}
+02/26/2022 17:25:02 - INFO - codeparrot_training - Step 49286: {'lr': 2.729258498138654e-07, 'samples': 25234944, 'steps': 49286, 'loss/train': 1.062586784362793}
+02/26/2022 17:25:06 - INFO - codeparrot_training - Step 49287: {'lr': 2.721620256426749e-07, 'samples': 25235456, 'steps': 49287, 'loss/train': 2.073833465576172}
+02/26/2022 17:25:11 - INFO - codeparrot_training - Step 49288: {'lr': 2.713992712262414e-07, 'samples': 25235968, 'steps': 49288, 'loss/train': 1.8696194887161255}
+02/26/2022 17:25:15 - INFO - codeparrot_training - Step 49289: {'lr': 2.706375865678123e-07, 'samples': 25236480, 'steps': 49289, 'loss/train': 1.149975061416626}
+02/26/2022 17:25:20 - INFO - codeparrot_training - Step 49290: {'lr': 2.698769716706906e-07, 'samples': 25236992, 'steps': 49290, 'loss/train': 1.6504911184310913}
+02/26/2022 17:25:24 - INFO - codeparrot_training - Step 49291: {'lr': 2.691174265381235e-07, 'samples': 25237504, 'steps': 49291, 'loss/train': 1.89723801612854}
+02/26/2022 17:25:29 - INFO - codeparrot_training - Step 49292: {'lr': 2.6835895117335864e-07, 'samples': 25238016, 'steps': 49292, 'loss/train': 1.4044740200042725}
+02/26/2022 17:25:33 - INFO - codeparrot_training - Step 49293: {'lr': 2.6760154557961547e-07, 'samples': 25238528, 'steps': 49293, 'loss/train': 1.5652474164962769}
+02/26/2022 17:25:38 - INFO - codeparrot_training - Step 49294: {'lr': 2.66845209760197e-07, 'samples': 25239040, 'steps': 49294, 'loss/train': 0.8309801816940308}
+02/26/2022 17:25:42 - INFO - codeparrot_training - Step 49295: {'lr': 2.660899437182951e-07, 'samples': 25239552, 'steps': 49295, 'loss/train': 2.2028732299804688}
+02/26/2022 17:25:47 - INFO - codeparrot_training - Step 49296: {'lr': 2.653357474571849e-07, 'samples': 25240064, 'steps': 49296, 'loss/train': 1.136959195137024}
+02/26/2022 17:25:51 - INFO - codeparrot_training - Step 49297: {'lr': 2.6458262098005834e-07, 'samples': 25240576, 'steps': 49297, 'loss/train': 1.6235889196395874}
+02/26/2022 17:25:56 - INFO - codeparrot_training - Step 49298: {'lr': 2.638305642901906e-07, 'samples': 25241088, 'steps': 49298, 'loss/train': 0.2817689776420593}
+02/26/2022 17:26:00 - INFO - codeparrot_training - Step 49299: {'lr': 2.630795773907457e-07, 'samples': 25241600, 'steps': 49299, 'loss/train': 0.5102606415748596}
+02/26/2022 17:26:07 - INFO - codeparrot_training - Step 49300: {'lr': 2.623296602849712e-07, 'samples': 25242112, 'steps': 49300, 'loss/train': 1.696257472038269}
+02/26/2022 17:26:10 - INFO - codeparrot_training - Step 49301: {'lr': 2.615808129760866e-07, 'samples': 25242624, 'steps': 49301, 'loss/train': 1.798613429069519}
+02/26/2022 17:26:16 - INFO - codeparrot_training - Step 49302: {'lr': 2.6083303546728387e-07, 'samples': 25243136, 'steps': 49302, 'loss/train': 0.9105786681175232}
+02/26/2022 17:26:19 - INFO - codeparrot_training - Step 49303: {'lr': 2.600863277617549e-07, 'samples': 25243648, 'steps': 49303, 'loss/train': 1.6797343492507935}
+02/26/2022 17:26:25 - INFO - codeparrot_training - Step 49304: {'lr': 2.593406898627193e-07, 'samples': 25244160, 'steps': 49304, 'loss/train': 1.3931041955947876}
+02/26/2022 17:26:29 - INFO - codeparrot_training - Step 49305: {'lr': 2.5859612177339673e-07, 'samples': 25244672, 'steps': 49305, 'loss/train': 1.4529887437820435}
+02/26/2022 17:26:34 - INFO - codeparrot_training - Step 49306: {'lr': 2.578526234969236e-07, 'samples': 25245184, 'steps': 49306, 'loss/train': 1.996252179145813}
+02/26/2022 17:26:38 - INFO - codeparrot_training - Step 49307: {'lr': 2.571101950364918e-07, 'samples': 25245696, 'steps': 49307, 'loss/train': 1.5781975984573364}
+02/26/2022 17:26:43 - INFO - codeparrot_training - Step 49308: {'lr': 2.5636883639532094e-07, 'samples': 25246208, 'steps': 49308, 'loss/train': 1.5069892406463623}
+02/26/2022 17:26:47 - INFO - codeparrot_training - Step 49309: {'lr': 2.5562854757657516e-07, 'samples': 25246720, 'steps': 49309, 'loss/train': 1.6497517824172974}
+02/26/2022 17:26:54 - INFO - codeparrot_training - Step 49310: {'lr': 2.5488932858339084e-07, 'samples': 25247232, 'steps': 49310, 'loss/train': 1.9937832355499268}
+02/26/2022 17:26:58 - INFO - codeparrot_training - Step 49311: {'lr': 2.5415117941895994e-07, 'samples': 25247744, 'steps': 49311, 'loss/train': 1.3809044361114502}
+02/26/2022 17:27:04 - INFO - codeparrot_training - Step 49312: {'lr': 2.534141000864465e-07, 'samples': 25248256, 'steps': 49312, 'loss/train': 1.4051517248153687}
+02/26/2022 17:27:07 - INFO - codeparrot_training - Step 49313: {'lr': 2.526780905889869e-07, 'samples': 25248768, 'steps': 49313, 'loss/train': 2.258091449737549}
+02/26/2022 17:27:11 - INFO - codeparrot_training - Step 49314: {'lr': 2.5194315092977315e-07, 'samples': 25249280, 'steps': 49314, 'loss/train': 3.379939079284668}
+02/26/2022 17:27:16 - INFO - codeparrot_training - Step 49315: {'lr': 2.512092811119415e-07, 'samples': 25249792, 'steps': 49315, 'loss/train': 0.4511565864086151}
+02/26/2022 17:27:20 - INFO - codeparrot_training - Step 49316: {'lr': 2.504764811386007e-07, 'samples': 25250304, 'steps': 49316, 'loss/train': 1.4937771558761597}
+02/26/2022 17:27:25 - INFO - codeparrot_training - Step 49317: {'lr': 2.497447510129147e-07, 'samples': 25250816, 'steps': 49317, 'loss/train': 1.8493415117263794}
+02/26/2022 17:27:29 - INFO - codeparrot_training - Step 49318: {'lr': 2.4901409073802006e-07, 'samples': 25251328, 'steps': 49318, 'loss/train': 1.8940860033035278}
+02/26/2022 17:27:34 - INFO - codeparrot_training - Step 49319: {'lr': 2.482845003170531e-07, 'samples': 25251840, 'steps': 49319, 'loss/train': 1.6199268102645874}
+02/26/2022 17:27:42 - INFO - codeparrot_training - Step 49320: {'lr': 2.475559797531224e-07, 'samples': 25252352, 'steps': 49320, 'loss/train': 0.8242297172546387}
+02/26/2022 17:27:45 - INFO - codeparrot_training - Step 49321: {'lr': 2.4682852904936434e-07, 'samples': 25252864, 'steps': 49321, 'loss/train': 2.161670684814453}
+02/26/2022 17:27:51 - INFO - codeparrot_training - Step 49322: {'lr': 2.4610214820888767e-07, 'samples': 25253376, 'steps': 49322, 'loss/train': 2.2019267082214355}
+02/26/2022 17:27:54 - INFO - codeparrot_training - Step 49323: {'lr': 2.4537683723480087e-07, 'samples': 25253888, 'steps': 49323, 'loss/train': 0.9363074898719788}
+02/26/2022 17:27:59 - INFO - codeparrot_training - Step 49324: {'lr': 2.4465259613021263e-07, 'samples': 25254400, 'steps': 49324, 'loss/train': 1.4704328775405884}
+02/26/2022 17:28:03 - INFO - codeparrot_training - Step 49325: {'lr': 2.439294248982038e-07, 'samples': 25254912, 'steps': 49325, 'loss/train': 2.071948766708374}
+02/26/2022 17:28:08 - INFO - codeparrot_training - Step 49326: {'lr': 2.4320732354191076e-07, 'samples': 25255424, 'steps': 49326, 'loss/train': 2.5521204471588135}
+02/26/2022 17:28:12 - INFO - codeparrot_training - Step 49327: {'lr': 2.4248629206444215e-07, 'samples': 25255936, 'steps': 49327, 'loss/train': 1.24164617061615}
+02/26/2022 17:28:17 - INFO - codeparrot_training - Step 49328: {'lr': 2.4176633046882337e-07, 'samples': 25256448, 'steps': 49328, 'loss/train': 1.9267997741699219}
+02/26/2022 17:28:21 - INFO - codeparrot_training - Step 49329: {'lr': 2.410474387581629e-07, 'samples': 25256960, 'steps': 49329, 'loss/train': 2.0199813842773438}
+02/26/2022 17:28:28 - INFO - codeparrot_training - Step 49330: {'lr': 2.4032961693556956e-07, 'samples': 25257472, 'steps': 49330, 'loss/train': 2.0545132160186768}
+02/26/2022 17:28:32 - INFO - codeparrot_training - Step 49331: {'lr': 2.396128650040963e-07, 'samples': 25257984, 'steps': 49331, 'loss/train': 1.6450552940368652}
+02/26/2022 17:28:37 - INFO - codeparrot_training - Step 49332: {'lr': 2.388971829667963e-07, 'samples': 25258496, 'steps': 49332, 'loss/train': 1.4598097801208496}
+02/26/2022 17:28:40 - INFO - codeparrot_training - Step 49333: {'lr': 2.3818257082677818e-07, 'samples': 25259008, 'steps': 49333, 'loss/train': 1.7282825708389282}
+02/26/2022 17:28:46 - INFO - codeparrot_training - Step 49334: {'lr': 2.3746902858703956e-07, 'samples': 25259520, 'steps': 49334, 'loss/train': 0.3359510004520416}
+02/26/2022 17:28:49 - INFO - codeparrot_training - Step 49335: {'lr': 2.3675655625068905e-07, 'samples': 25260032, 'steps': 49335, 'loss/train': 2.07143235206604}
+02/26/2022 17:28:55 - INFO - codeparrot_training - Step 49336: {'lr': 2.3604515382077974e-07, 'samples': 25260544, 'steps': 49336, 'loss/train': 0.24383077025413513}
+02/26/2022 17:28:58 - INFO - codeparrot_training - Step 49337: {'lr': 2.3533482130033702e-07, 'samples': 25261056, 'steps': 49337, 'loss/train': 2.0739169120788574}
+02/26/2022 17:29:04 - INFO - codeparrot_training - Step 49338: {'lr': 2.3462555869241398e-07, 'samples': 25261568, 'steps': 49338, 'loss/train': 0.6827960014343262}
+02/26/2022 17:29:08 - INFO - codeparrot_training - Step 49339: {'lr': 2.33917366000036e-07, 'samples': 25262080, 'steps': 49339, 'loss/train': 1.594533920288086}
+02/26/2022 17:29:13 - INFO - codeparrot_training - Step 49340: {'lr': 2.3321024322625618e-07, 'samples': 25262592, 'steps': 49340, 'loss/train': 1.655191421508789}
+02/26/2022 17:29:17 - INFO - codeparrot_training - Step 49341: {'lr': 2.3250419037409986e-07, 'samples': 25263104, 'steps': 49341, 'loss/train': 1.4917610883712769}
+02/26/2022 17:29:22 - INFO - codeparrot_training - Step 49342: {'lr': 2.317992074465647e-07, 'samples': 25263616, 'steps': 49342, 'loss/train': 1.2126142978668213}
+02/26/2022 17:29:26 - INFO - codeparrot_training - Step 49343: {'lr': 2.3109529444670374e-07, 'samples': 25264128, 'steps': 49343, 'loss/train': 1.5893272161483765}
+02/26/2022 17:29:31 - INFO - codeparrot_training - Step 49344: {'lr': 2.3039245137751463e-07, 'samples': 25264640, 'steps': 49344, 'loss/train': 1.7388670444488525}
+02/26/2022 17:29:35 - INFO - codeparrot_training - Step 49345: {'lr': 2.296906782420227e-07, 'samples': 25265152, 'steps': 49345, 'loss/train': 1.2293654680252075}
+02/26/2022 17:29:42 - INFO - codeparrot_training - Step 49346: {'lr': 2.289899750432256e-07, 'samples': 25265664, 'steps': 49346, 'loss/train': 2.2173004150390625}
+02/26/2022 17:29:45 - INFO - codeparrot_training - Step 49347: {'lr': 2.2829034178412088e-07, 'samples': 25266176, 'steps': 49347, 'loss/train': 1.1115319728851318}
+02/26/2022 17:29:51 - INFO - codeparrot_training - Step 49348: {'lr': 2.275917784677062e-07, 'samples': 25266688, 'steps': 49348, 'loss/train': 1.7598215341567993}
+02/26/2022 17:29:54 - INFO - codeparrot_training - Step 49349: {'lr': 2.2689428509697907e-07, 'samples': 25267200, 'steps': 49349, 'loss/train': 2.1307475566864014}
+02/26/2022 17:30:00 - INFO - codeparrot_training - Step 49350: {'lr': 2.2619786167490942e-07, 'samples': 25267712, 'steps': 49350, 'loss/train': 2.045980215072632}
+02/26/2022 17:30:03 - INFO - codeparrot_training - Step 49351: {'lr': 2.2550250820452256e-07, 'samples': 25268224, 'steps': 49351, 'loss/train': 1.6557023525238037}
+02/26/2022 17:30:09 - INFO - codeparrot_training - Step 49352: {'lr': 2.248082246887606e-07, 'samples': 25268736, 'steps': 49352, 'loss/train': 2.380180835723877}
+02/26/2022 17:30:12 - INFO - codeparrot_training - Step 49353: {'lr': 2.241150111305934e-07, 'samples': 25269248, 'steps': 49353, 'loss/train': 2.416721820831299}
+02/26/2022 17:30:18 - INFO - codeparrot_training - Step 49354: {'lr': 2.2342286753301856e-07, 'samples': 25269760, 'steps': 49354, 'loss/train': 2.3425889015197754}
+02/26/2022 17:30:21 - INFO - codeparrot_training - Step 49355: {'lr': 2.2273179389897812e-07, 'samples': 25270272, 'steps': 49355, 'loss/train': 0.30048707127571106}
+02/26/2022 17:30:29 - INFO - codeparrot_training - Step 49356: {'lr': 2.22041790231442e-07, 'samples': 25270784, 'steps': 49356, 'loss/train': 0.8866140246391296}
+02/26/2022 17:30:32 - INFO - codeparrot_training - Step 49357: {'lr': 2.2135285653338e-07, 'samples': 25271296, 'steps': 49357, 'loss/train': 2.2440567016601562}
+02/26/2022 17:30:38 - INFO - codeparrot_training - Step 49358: {'lr': 2.2066499280773421e-07, 'samples': 25271808, 'steps': 49358, 'loss/train': 1.9294134378433228}
+02/26/2022 17:30:41 - INFO - codeparrot_training - Step 49359: {'lr': 2.19978199057419e-07, 'samples': 25272320, 'steps': 49359, 'loss/train': 1.5684888362884521}
+02/26/2022 17:30:47 - INFO - codeparrot_training - Step 49360: {'lr': 2.1929247528540418e-07, 'samples': 25272832, 'steps': 49360, 'loss/train': 0.646882951259613}
+02/26/2022 17:30:50 - INFO - codeparrot_training - Step 49361: {'lr': 2.1860782149463188e-07, 'samples': 25273344, 'steps': 49361, 'loss/train': 0.8152244687080383}
+02/26/2022 17:30:56 - INFO - codeparrot_training - Step 49362: {'lr': 2.1792423768804415e-07, 'samples': 25273856, 'steps': 49362, 'loss/train': 1.682790756225586}
+02/26/2022 17:30:59 - INFO - codeparrot_training - Step 49363: {'lr': 2.1724172386855535e-07, 'samples': 25274368, 'steps': 49363, 'loss/train': 1.901334524154663}
+02/26/2022 17:31:05 - INFO - codeparrot_training - Step 49364: {'lr': 2.1656028003907978e-07, 'samples': 25274880, 'steps': 49364, 'loss/train': 1.6426023244857788}
+02/26/2022 17:31:08 - INFO - codeparrot_training - Step 49365: {'lr': 2.1587990620253185e-07, 'samples': 25275392, 'steps': 49365, 'loss/train': 1.0824851989746094}
+02/26/2022 17:31:15 - INFO - codeparrot_training - Step 49366: {'lr': 2.1520060236185357e-07, 'samples': 25275904, 'steps': 49366, 'loss/train': 1.577638864517212}
+02/26/2022 17:31:19 - INFO - codeparrot_training - Step 49367: {'lr': 2.1452236851993157e-07, 'samples': 25276416, 'steps': 49367, 'loss/train': 0.9240710735321045}
+02/26/2022 17:31:24 - INFO - codeparrot_training - Step 49368: {'lr': 2.1384520467968017e-07, 'samples': 25276928, 'steps': 49368, 'loss/train': 2.516341209411621}
+02/26/2022 17:31:28 - INFO - codeparrot_training - Step 49369: {'lr': 2.131691108440137e-07, 'samples': 25277440, 'steps': 49369, 'loss/train': 1.1944133043289185}
+02/26/2022 17:31:33 - INFO - codeparrot_training - Step 49370: {'lr': 2.1249408701581873e-07, 'samples': 25277952, 'steps': 49370, 'loss/train': 2.0850064754486084}
+02/26/2022 17:31:37 - INFO - codeparrot_training - Step 49371: {'lr': 2.1182013319795412e-07, 'samples': 25278464, 'steps': 49371, 'loss/train': 0.7400929927825928}
+02/26/2022 17:31:43 - INFO - codeparrot_training - Step 49372: {'lr': 2.1114724939336195e-07, 'samples': 25278976, 'steps': 49372, 'loss/train': 2.0787765979766846}
+02/26/2022 17:31:46 - INFO - codeparrot_training - Step 49373: {'lr': 2.1047543560487327e-07, 'samples': 25279488, 'steps': 49373, 'loss/train': 2.363260507583618}
+02/26/2022 17:31:52 - INFO - codeparrot_training - Step 49374: {'lr': 2.098046918354024e-07, 'samples': 25280000, 'steps': 49374, 'loss/train': 1.6996179819107056}
+02/26/2022 17:31:55 - INFO - codeparrot_training - Step 49375: {'lr': 2.091350180878082e-07, 'samples': 25280512, 'steps': 49375, 'loss/train': 2.3067574501037598}
+02/26/2022 17:32:01 - INFO - codeparrot_training - Step 49376: {'lr': 2.0846641436497726e-07, 'samples': 25281024, 'steps': 49376, 'loss/train': 0.6636629700660706}
+02/26/2022 17:32:04 - INFO - codeparrot_training - Step 49377: {'lr': 2.0779888066974062e-07, 'samples': 25281536, 'steps': 49377, 'loss/train': 2.244915723800659}
+02/26/2022 17:32:11 - INFO - codeparrot_training - Step 49378: {'lr': 2.0713241700498485e-07, 'samples': 25282048, 'steps': 49378, 'loss/train': 2.468876838684082}
+02/26/2022 17:32:15 - INFO - codeparrot_training - Step 49379: {'lr': 2.0646702337354108e-07, 'samples': 25282560, 'steps': 49379, 'loss/train': 1.5205780267715454}
+02/26/2022 17:32:20 - INFO - codeparrot_training - Step 49380: {'lr': 2.058026997782958e-07, 'samples': 25283072, 'steps': 49380, 'loss/train': 2.1277313232421875}
+02/26/2022 17:32:24 - INFO - codeparrot_training - Step 49381: {'lr': 2.0513944622205239e-07, 'samples': 25283584, 'steps': 49381, 'loss/train': 0.9164019227027893}
+02/26/2022 17:32:29 - INFO - codeparrot_training - Step 49382: {'lr': 2.0447726270769739e-07, 'samples': 25284096, 'steps': 49382, 'loss/train': 1.305641770362854}
+02/26/2022 17:32:33 - INFO - codeparrot_training - Step 49383: {'lr': 2.0381614923803417e-07, 'samples': 25284608, 'steps': 49383, 'loss/train': 1.5540956258773804}
+02/26/2022 17:32:38 - INFO - codeparrot_training - Step 49384: {'lr': 2.0315610581589372e-07, 'samples': 25285120, 'steps': 49384, 'loss/train': 0.8872753977775574}
+02/26/2022 17:32:42 - INFO - codeparrot_training - Step 49385: {'lr': 2.0249713244410718e-07, 'samples': 25285632, 'steps': 49385, 'loss/train': 1.963981032371521}
+02/26/2022 17:32:47 - INFO - codeparrot_training - Step 49386: {'lr': 2.0183922912553332e-07, 'samples': 25286144, 'steps': 49386, 'loss/train': 2.091745376586914}
+02/26/2022 17:32:51 - INFO - codeparrot_training - Step 49387: {'lr': 2.0118239586292e-07, 'samples': 25286656, 'steps': 49387, 'loss/train': 2.2686331272125244}
+02/26/2022 17:32:56 - INFO - codeparrot_training - Step 49388: {'lr': 2.0052663265915373e-07, 'samples': 25287168, 'steps': 49388, 'loss/train': 0.8810884952545166}
+02/26/2022 17:33:00 - INFO - codeparrot_training - Step 49389: {'lr': 1.9987193951698235e-07, 'samples': 25287680, 'steps': 49389, 'loss/train': 1.7230876684188843}
+02/26/2022 17:33:05 - INFO - codeparrot_training - Step 49390: {'lr': 1.9921831643926468e-07, 'samples': 25288192, 'steps': 49390, 'loss/train': 2.2715725898742676}
+02/26/2022 17:33:09 - INFO - codeparrot_training - Step 49391: {'lr': 1.985657634287763e-07, 'samples': 25288704, 'steps': 49391, 'loss/train': 1.448055386543274}
+02/26/2022 17:33:16 - INFO - codeparrot_training - Step 49392: {'lr': 1.9791428048829275e-07, 'samples': 25289216, 'steps': 49392, 'loss/train': 1.7186641693115234}
+02/26/2022 17:33:20 - INFO - codeparrot_training - Step 49393: {'lr': 1.9726386762064508e-07, 'samples': 25289728, 'steps': 49393, 'loss/train': 1.3137915134429932}
+02/26/2022 17:33:25 - INFO - codeparrot_training - Step 49394: {'lr': 1.966145248285811e-07, 'samples': 25290240, 'steps': 49394, 'loss/train': 0.8090012669563293}
+02/26/2022 17:33:28 - INFO - codeparrot_training - Step 49395: {'lr': 1.9596625211490416e-07, 'samples': 25290752, 'steps': 49395, 'loss/train': 1.6962679624557495}
+02/26/2022 17:33:34 - INFO - codeparrot_training - Step 49396: {'lr': 1.9531904948236202e-07, 'samples': 25291264, 'steps': 49396, 'loss/train': 1.3675588369369507}
+02/26/2022 17:33:37 - INFO - codeparrot_training - Step 49397: {'lr': 1.9467291693378574e-07, 'samples': 25291776, 'steps': 49397, 'loss/train': 0.5769745707511902}
+02/26/2022 17:33:43 - INFO - codeparrot_training - Step 49398: {'lr': 1.9402785447189542e-07, 'samples': 25292288, 'steps': 49398, 'loss/train': 2.389784812927246}
+02/26/2022 17:33:47 - INFO - codeparrot_training - Step 49399: {'lr': 1.9338386209946656e-07, 'samples': 25292800, 'steps': 49399, 'loss/train': 2.068979263305664}
+02/26/2022 17:33:52 - INFO - codeparrot_training - Step 49400: {'lr': 1.9274093981927476e-07, 'samples': 25293312, 'steps': 49400, 'loss/train': 2.364450693130493}
+02/26/2022 17:33:55 - INFO - codeparrot_training - Step 49401: {'lr': 1.9209908763404006e-07, 'samples': 25293824, 'steps': 49401, 'loss/train': 1.3425612449645996}
+02/26/2022 17:34:03 - INFO - codeparrot_training - Step 49402: {'lr': 1.91458305546538e-07, 'samples': 25294336, 'steps': 49402, 'loss/train': 1.4773885011672974}
+02/26/2022 17:34:06 - INFO - codeparrot_training - Step 49403: {'lr': 1.908185935594886e-07, 'samples': 25294848, 'steps': 49403, 'loss/train': 1.7944095134735107}
+02/26/2022 17:34:12 - INFO - codeparrot_training - Step 49404: {'lr': 1.901799516756675e-07, 'samples': 25295360, 'steps': 49404, 'loss/train': 2.3042733669281006}
+02/26/2022 17:34:15 - INFO - codeparrot_training - Step 49405: {'lr': 1.8954237989779467e-07, 'samples': 25295872, 'steps': 49405, 'loss/train': 1.5094069242477417}
+02/26/2022 17:34:21 - INFO - codeparrot_training - Step 49406: {'lr': 1.8890587822859018e-07, 'samples': 25296384, 'steps': 49406, 'loss/train': 1.779155969619751}
+02/26/2022 17:34:24 - INFO - codeparrot_training - Step 49407: {'lr': 1.8827044667077408e-07, 'samples': 25296896, 'steps': 49407, 'loss/train': 2.8769690990448}
+02/26/2022 17:34:30 - INFO - codeparrot_training - Step 49408: {'lr': 1.8763608522709418e-07, 'samples': 25297408, 'steps': 49408, 'loss/train': 1.1164543628692627}
+02/26/2022 17:34:33 - INFO - codeparrot_training - Step 49409: {'lr': 1.8700279390027052e-07, 'samples': 25297920, 'steps': 49409, 'loss/train': 1.7498164176940918}
+02/26/2022 17:34:39 - INFO - codeparrot_training - Step 49410: {'lr': 1.863705726929954e-07, 'samples': 25298432, 'steps': 49410, 'loss/train': 0.7097488641738892}
+02/26/2022 17:34:42 - INFO - codeparrot_training - Step 49411: {'lr': 1.8573942160798885e-07, 'samples': 25298944, 'steps': 49411, 'loss/train': 1.2397820949554443}
+02/26/2022 17:34:50 - INFO - codeparrot_training - Step 49412: {'lr': 1.8510934064791542e-07, 'samples': 25299456, 'steps': 49412, 'loss/train': 2.2480454444885254}
+02/26/2022 17:34:54 - INFO - codeparrot_training - Step 49413: {'lr': 1.8448032981555064e-07, 'samples': 25299968, 'steps': 49413, 'loss/train': 2.2325031757354736}
+02/26/2022 17:34:57 - INFO - codeparrot_training - Step 49414: {'lr': 1.838523891135313e-07, 'samples': 25300480, 'steps': 49414, 'loss/train': 1.7891182899475098}
+02/26/2022 17:35:03 - INFO - codeparrot_training - Step 49415: {'lr': 1.8322551854454973e-07, 'samples': 25300992, 'steps': 49415, 'loss/train': 1.816938877105713}
+02/26/2022 17:35:06 - INFO - codeparrot_training - Step 49416: {'lr': 1.8259971811132593e-07, 'samples': 25301504, 'steps': 49416, 'loss/train': 1.3228381872177124}
+02/26/2022 17:35:11 - INFO - codeparrot_training - Step 49417: {'lr': 1.8197498781652443e-07, 'samples': 25302016, 'steps': 49417, 'loss/train': 1.1276450157165527}
+02/26/2022 17:35:15 - INFO - codeparrot_training - Step 49418: {'lr': 1.8135132766278207e-07, 'samples': 25302528, 'steps': 49418, 'loss/train': 1.9976853132247925}
+02/26/2022 17:35:20 - INFO - codeparrot_training - Step 49419: {'lr': 1.8072873765281883e-07, 'samples': 25303040, 'steps': 49419, 'loss/train': 1.4852638244628906}
+02/26/2022 17:35:24 - INFO - codeparrot_training - Step 49420: {'lr': 1.801072177892993e-07, 'samples': 25303552, 'steps': 49420, 'loss/train': 1.1721638441085815}
+02/26/2022 17:35:29 - INFO - codeparrot_training - Step 49421: {'lr': 1.794867680748602e-07, 'samples': 25304064, 'steps': 49421, 'loss/train': 1.2996807098388672}
+02/26/2022 17:35:33 - INFO - codeparrot_training - Step 49422: {'lr': 1.7886738851219388e-07, 'samples': 25304576, 'steps': 49422, 'loss/train': 1.7551120519638062}
+02/26/2022 17:35:38 - INFO - codeparrot_training - Step 49423: {'lr': 1.782490791039093e-07, 'samples': 25305088, 'steps': 49423, 'loss/train': 1.8207831382751465}
+02/26/2022 17:35:42 - INFO - codeparrot_training - Step 49424: {'lr': 1.7763183985269881e-07, 'samples': 25305600, 'steps': 49424, 'loss/train': 1.570170521736145}
+02/26/2022 17:35:49 - INFO - codeparrot_training - Step 49425: {'lr': 1.770156707611714e-07, 'samples': 25306112, 'steps': 49425, 'loss/train': 1.4419302940368652}
+02/26/2022 17:35:53 - INFO - codeparrot_training - Step 49426: {'lr': 1.7640057183196389e-07, 'samples': 25306624, 'steps': 49426, 'loss/train': 1.6194907426834106}
+02/26/2022 17:35:58 - INFO - codeparrot_training - Step 49427: {'lr': 1.7578654306776853e-07, 'samples': 25307136, 'steps': 49427, 'loss/train': 0.6367307901382446}
+02/26/2022 17:36:02 - INFO - codeparrot_training - Step 49428: {'lr': 1.7517358447113884e-07, 'samples': 25307648, 'steps': 49428, 'loss/train': 1.4055144786834717}
+02/26/2022 17:36:07 - INFO - codeparrot_training - Step 49429: {'lr': 1.7456169604476713e-07, 'samples': 25308160, 'steps': 49429, 'loss/train': 0.07487159967422485}
+02/26/2022 17:36:11 - INFO - codeparrot_training - Step 49430: {'lr': 1.7395087779123465e-07, 'samples': 25308672, 'steps': 49430, 'loss/train': 1.3172026872634888}
+02/26/2022 17:36:16 - INFO - codeparrot_training - Step 49431: {'lr': 1.7334112971315042e-07, 'samples': 25309184, 'steps': 49431, 'loss/train': 1.3774311542510986}
+02/26/2022 17:36:20 - INFO - codeparrot_training - Step 49432: {'lr': 1.7273245181315122e-07, 'samples': 25309696, 'steps': 49432, 'loss/train': 1.0256197452545166}
+02/26/2022 17:36:25 - INFO - codeparrot_training - Step 49433: {'lr': 1.721248440938461e-07, 'samples': 25310208, 'steps': 49433, 'loss/train': 1.517964482307434}
+02/26/2022 17:36:28 - INFO - codeparrot_training - Step 49434: {'lr': 1.7151830655784405e-07, 'samples': 25310720, 'steps': 49434, 'loss/train': 0.5323280096054077}
+02/26/2022 17:36:36 - INFO - codeparrot_training - Step 49435: {'lr': 1.7091283920769863e-07, 'samples': 25311232, 'steps': 49435, 'loss/train': 2.2097764015197754}
+02/26/2022 17:36:39 - INFO - codeparrot_training - Step 49436: {'lr': 1.7030844204604657e-07, 'samples': 25311744, 'steps': 49436, 'loss/train': 2.1547555923461914}
+02/26/2022 17:36:45 - INFO - codeparrot_training - Step 49437: {'lr': 1.6970511507546914e-07, 'samples': 25312256, 'steps': 49437, 'loss/train': 0.35932597517967224}
+02/26/2022 17:36:50 - INFO - codeparrot_training - Step 49438: {'lr': 1.691028582985199e-07, 'samples': 25312768, 'steps': 49438, 'loss/train': 1.6998181343078613}
+02/26/2022 17:36:54 - INFO - codeparrot_training - Step 49439: {'lr': 1.685016717178356e-07, 'samples': 25313280, 'steps': 49439, 'loss/train': 2.5360960960388184}
+02/26/2022 17:36:59 - INFO - codeparrot_training - Step 49440: {'lr': 1.67901555335942e-07, 'samples': 25313792, 'steps': 49440, 'loss/train': 2.00775408744812}
+02/26/2022 17:37:03 - INFO - codeparrot_training - Step 49441: {'lr': 1.6730250915542035e-07, 'samples': 25314304, 'steps': 49441, 'loss/train': 0.01932586543262005}
+02/26/2022 17:37:08 - INFO - codeparrot_training - Step 49442: {'lr': 1.6670453317885192e-07, 'samples': 25314816, 'steps': 49442, 'loss/train': 2.3165156841278076}
+02/26/2022 17:37:12 - INFO - codeparrot_training - Step 49443: {'lr': 1.6610762740879027e-07, 'samples': 25315328, 'steps': 49443, 'loss/train': 2.51115083694458}
+02/26/2022 17:37:17 - INFO - codeparrot_training - Step 49444: {'lr': 1.6551179184781663e-07, 'samples': 25315840, 'steps': 49444, 'loss/train': 2.125347137451172}
+02/26/2022 17:37:21 - INFO - codeparrot_training - Step 49445: {'lr': 1.64917026498429e-07, 'samples': 25316352, 'steps': 49445, 'loss/train': 1.4960134029388428}
+02/26/2022 17:37:26 - INFO - codeparrot_training - Step 49446: {'lr': 1.6432333136320865e-07, 'samples': 25316864, 'steps': 49446, 'loss/train': 0.8772303462028503}
+02/26/2022 17:37:30 - INFO - codeparrot_training - Step 49447: {'lr': 1.637307064447091e-07, 'samples': 25317376, 'steps': 49447, 'loss/train': 1.2380704879760742}
+02/26/2022 17:37:37 - INFO - codeparrot_training - Step 49448: {'lr': 1.6313915174542836e-07, 'samples': 25317888, 'steps': 49448, 'loss/train': 1.084328532218933}
+02/26/2022 17:37:40 - INFO - codeparrot_training - Step 49449: {'lr': 1.625486672679477e-07, 'samples': 25318400, 'steps': 49449, 'loss/train': 2.2731661796569824}
+02/26/2022 17:37:46 - INFO - codeparrot_training - Step 49450: {'lr': 1.6195925301479287e-07, 'samples': 25318912, 'steps': 49450, 'loss/train': 2.20371150970459}
+02/26/2022 17:37:49 - INFO - codeparrot_training - Step 49451: {'lr': 1.613709089884341e-07, 'samples': 25319424, 'steps': 49451, 'loss/train': 1.5246758460998535}
+02/26/2022 17:37:55 - INFO - codeparrot_training - Step 49452: {'lr': 1.6078363519145268e-07, 'samples': 25319936, 'steps': 49452, 'loss/train': 2.318956136703491}
+02/26/2022 17:37:58 - INFO - codeparrot_training - Step 49453: {'lr': 1.601974316263466e-07, 'samples': 25320448, 'steps': 49453, 'loss/train': 1.7194854021072388}
+02/26/2022 17:38:04 - INFO - codeparrot_training - Step 49454: {'lr': 1.5961229829561385e-07, 'samples': 25320960, 'steps': 49454, 'loss/train': 1.0852497816085815}
+02/26/2022 17:38:07 - INFO - codeparrot_training - Step 49455: {'lr': 1.5902823520175247e-07, 'samples': 25321472, 'steps': 49455, 'loss/train': 2.0783467292785645}
+02/26/2022 17:38:13 - INFO - codeparrot_training - Step 49456: {'lr': 1.5844524234728818e-07, 'samples': 25321984, 'steps': 49456, 'loss/train': 0.9691670536994934}
+02/26/2022 17:38:16 - INFO - codeparrot_training - Step 49457: {'lr': 1.57863319734719e-07, 'samples': 25322496, 'steps': 49457, 'loss/train': 2.3410816192626953}
+02/26/2022 17:38:20 - INFO - codeparrot_training - Step 49458: {'lr': 1.5728246736654294e-07, 'samples': 25323008, 'steps': 49458, 'loss/train': 2.033514976501465}
+02/26/2022 17:38:26 - INFO - codeparrot_training - Step 49459: {'lr': 1.5670268524520248e-07, 'samples': 25323520, 'steps': 49459, 'loss/train': 2.0586495399475098}
+02/26/2022 17:38:29 - INFO - codeparrot_training - Step 49460: {'lr': 1.5612397337325114e-07, 'samples': 25324032, 'steps': 49460, 'loss/train': 0.4165288507938385}
+02/26/2022 17:38:36 - INFO - codeparrot_training - Step 49461: {'lr': 1.5554633175310363e-07, 'samples': 25324544, 'steps': 49461, 'loss/train': 1.7726829051971436}
+02/26/2022 17:38:40 - INFO - codeparrot_training - Step 49462: {'lr': 1.5496976038725795e-07, 'samples': 25325056, 'steps': 49462, 'loss/train': 1.6395699977874756}
+02/26/2022 17:38:45 - INFO - codeparrot_training - Step 49463: {'lr': 1.5439425927821217e-07, 'samples': 25325568, 'steps': 49463, 'loss/train': 1.2925454378128052}
+02/26/2022 17:38:49 - INFO - codeparrot_training - Step 49464: {'lr': 1.5381982842838093e-07, 'samples': 25326080, 'steps': 49464, 'loss/train': 1.6523593664169312}
+02/26/2022 17:38:54 - INFO - codeparrot_training - Step 49465: {'lr': 1.5324646784026232e-07, 'samples': 25326592, 'steps': 49465, 'loss/train': 0.07220093905925751}
+02/26/2022 17:38:58 - INFO - codeparrot_training - Step 49466: {'lr': 1.5267417751629876e-07, 'samples': 25327104, 'steps': 49466, 'loss/train': 1.0161285400390625}
+02/26/2022 17:39:03 - INFO - codeparrot_training - Step 49467: {'lr': 1.5210295745893277e-07, 'samples': 25327616, 'steps': 49467, 'loss/train': 1.5583438873291016}
+02/26/2022 17:39:07 - INFO - codeparrot_training - Step 49468: {'lr': 1.515328076706346e-07, 'samples': 25328128, 'steps': 49468, 'loss/train': 1.0967252254486084}
+02/26/2022 17:39:13 - INFO - codeparrot_training - Step 49469: {'lr': 1.5096372815384675e-07, 'samples': 25328640, 'steps': 49469, 'loss/train': 1.5874531269073486}
+02/26/2022 17:39:16 - INFO - codeparrot_training - Step 49470: {'lr': 1.503957189109839e-07, 'samples': 25329152, 'steps': 49470, 'loss/train': 1.2299761772155762}
+02/26/2022 17:39:23 - INFO - codeparrot_training - Step 49471: {'lr': 1.4982877994448862e-07, 'samples': 25329664, 'steps': 49471, 'loss/train': 0.9899556636810303}
+02/26/2022 17:39:27 - INFO - codeparrot_training - Step 49472: {'lr': 1.492629112567756e-07, 'samples': 25330176, 'steps': 49472, 'loss/train': 2.01727032661438}
+02/26/2022 17:39:32 - INFO - codeparrot_training - Step 49473: {'lr': 1.4869811285028734e-07, 'samples': 25330688, 'steps': 49473, 'loss/train': 2.082319974899292}
+02/26/2022 17:39:36 - INFO - codeparrot_training - Step 49474: {'lr': 1.4813438472746632e-07, 'samples': 25331200, 'steps': 49474, 'loss/train': 2.282454013824463}
+02/26/2022 17:39:41 - INFO - codeparrot_training - Step 49475: {'lr': 1.4757172689067177e-07, 'samples': 25331712, 'steps': 49475, 'loss/train': 2.0538759231567383}
+02/26/2022 17:39:45 - INFO - codeparrot_training - Step 49476: {'lr': 1.470101393423462e-07, 'samples': 25332224, 'steps': 49476, 'loss/train': 1.5414130687713623}
+02/26/2022 17:39:50 - INFO - codeparrot_training - Step 49477: {'lr': 1.464496220849043e-07, 'samples': 25332736, 'steps': 49477, 'loss/train': 1.9200726747512817}
+02/26/2022 17:39:54 - INFO - codeparrot_training - Step 49478: {'lr': 1.458901751207331e-07, 'samples': 25333248, 'steps': 49478, 'loss/train': 1.1467262506484985}
+02/26/2022 17:39:59 - INFO - codeparrot_training - Step 49479: {'lr': 1.4533179845221956e-07, 'samples': 25333760, 'steps': 49479, 'loss/train': 1.0757455825805664}
+02/26/2022 17:40:03 - INFO - codeparrot_training - Step 49480: {'lr': 1.447744920817784e-07, 'samples': 25334272, 'steps': 49480, 'loss/train': 1.1312575340270996}
+02/26/2022 17:40:10 - INFO - codeparrot_training - Step 49481: {'lr': 1.442182560117966e-07, 'samples': 25334784, 'steps': 49481, 'loss/train': 1.7154914140701294}
+02/26/2022 17:40:14 - INFO - codeparrot_training - Step 49482: {'lr': 1.4366309024463343e-07, 'samples': 25335296, 'steps': 49482, 'loss/train': 0.4336872696876526}
+02/26/2022 17:40:19 - INFO - codeparrot_training - Step 49483: {'lr': 1.431089947826758e-07, 'samples': 25335808, 'steps': 49483, 'loss/train': 1.353283405303955}
+02/26/2022 17:40:23 - INFO - codeparrot_training - Step 49484: {'lr': 1.42555969628283e-07, 'samples': 25336320, 'steps': 49484, 'loss/train': 1.3421893119812012}
+02/26/2022 17:40:29 - INFO - codeparrot_training - Step 49485: {'lr': 1.420040147838697e-07, 'samples': 25336832, 'steps': 49485, 'loss/train': 2.022498846054077}
+02/26/2022 17:40:33 - INFO - codeparrot_training - Step 49486: {'lr': 1.4145313025176743e-07, 'samples': 25337344, 'steps': 49486, 'loss/train': 2.103182315826416}
+02/26/2022 17:40:38 - INFO - codeparrot_training - Step 49487: {'lr': 1.4090331603433536e-07, 'samples': 25337856, 'steps': 49487, 'loss/train': 1.5370337963104248}
+02/26/2022 17:40:42 - INFO - codeparrot_training - Step 49488: {'lr': 1.4035457213393277e-07, 'samples': 25338368, 'steps': 49488, 'loss/train': 2.414919853210449}
+02/26/2022 17:40:47 - INFO - codeparrot_training - Step 49489: {'lr': 1.3980689855291884e-07, 'samples': 25338880, 'steps': 49489, 'loss/train': 2.0758466720581055}
+02/26/2022 17:40:51 - INFO - codeparrot_training - Step 49490: {'lr': 1.3926029529362506e-07, 'samples': 25339392, 'steps': 49490, 'loss/train': 2.3458642959594727}
+02/26/2022 17:40:56 - INFO - codeparrot_training - Step 49491: {'lr': 1.3871476235841063e-07, 'samples': 25339904, 'steps': 49491, 'loss/train': 2.408580780029297}
+02/26/2022 17:41:00 - INFO - codeparrot_training - Step 49492: {'lr': 1.381702997495793e-07, 'samples': 25340416, 'steps': 49492, 'loss/train': 1.6963386535644531}
+02/26/2022 17:41:05 - INFO - codeparrot_training - Step 49493: {'lr': 1.37626907469518e-07, 'samples': 25340928, 'steps': 49493, 'loss/train': 0.04213282838463783}
+02/26/2022 17:41:09 - INFO - codeparrot_training - Step 49494: {'lr': 1.3708458552050274e-07, 'samples': 25341440, 'steps': 49494, 'loss/train': 1.2074458599090576}
+02/26/2022 17:41:16 - INFO - codeparrot_training - Step 49495: {'lr': 1.3654333390486494e-07, 'samples': 25341952, 'steps': 49495, 'loss/train': 0.6862615942955017}
+02/26/2022 17:41:19 - INFO - codeparrot_training - Step 49496: {'lr': 1.3600315262496388e-07, 'samples': 25342464, 'steps': 49496, 'loss/train': 1.3673888444900513}
+02/26/2022 17:41:25 - INFO - codeparrot_training - Step 49497: {'lr': 1.3546404168307547e-07, 'samples': 25342976, 'steps': 49497, 'loss/train': 1.2132560014724731}
+02/26/2022 17:41:29 - INFO - codeparrot_training - Step 49498: {'lr': 1.3492600108150345e-07, 'samples': 25343488, 'steps': 49498, 'loss/train': 2.350691556930542}
+02/26/2022 17:41:34 - INFO - codeparrot_training - Step 49499: {'lr': 1.3438903082257924e-07, 'samples': 25344000, 'steps': 49499, 'loss/train': 1.8677647113800049}
+02/26/2022 17:41:37 - INFO - codeparrot_training - Step 49500: {'lr': 1.3385313090857886e-07, 'samples': 25344512, 'steps': 49500, 'loss/train': 1.8952529430389404}
+02/26/2022 17:41:43 - INFO - codeparrot_training - Step 49501: {'lr': 1.3331830134183376e-07, 'samples': 25345024, 'steps': 49501, 'loss/train': 1.6151928901672363}
+02/26/2022 17:41:47 - INFO - codeparrot_training - Step 49502: {'lr': 1.3278454212459213e-07, 'samples': 25345536, 'steps': 49502, 'loss/train': 1.4087148904800415}
+02/26/2022 17:41:52 - INFO - codeparrot_training - Step 49503: {'lr': 1.3225185325918542e-07, 'samples': 25346048, 'steps': 49503, 'loss/train': 1.5115303993225098}
+02/26/2022 17:41:56 - INFO - codeparrot_training - Step 49504: {'lr': 1.3172023474783413e-07, 'samples': 25346560, 'steps': 49504, 'loss/train': 1.823445439338684}
+02/26/2022 17:42:03 - INFO - codeparrot_training - Step 49505: {'lr': 1.3118968659286967e-07, 'samples': 25347072, 'steps': 49505, 'loss/train': 2.0007190704345703}
+02/26/2022 17:42:07 - INFO - codeparrot_training - Step 49506: {'lr': 1.3066020879654027e-07, 'samples': 25347584, 'steps': 49506, 'loss/train': 1.6916680335998535}
+02/26/2022 17:42:12 - INFO - codeparrot_training - Step 49507: {'lr': 1.301318013611219e-07, 'samples': 25348096, 'steps': 49507, 'loss/train': 2.3506176471710205}
+02/26/2022 17:42:16 - INFO - codeparrot_training - Step 49508: {'lr': 1.296044642888905e-07, 'samples': 25348608, 'steps': 49508, 'loss/train': 2.174691677093506}
+02/26/2022 17:42:21 - INFO - codeparrot_training - Step 49509: {'lr': 1.2907819758206652e-07, 'samples': 25349120, 'steps': 49509, 'loss/train': 2.3295016288757324}
+02/26/2022 17:42:25 - INFO - codeparrot_training - Step 49510: {'lr': 1.2855300124295367e-07, 'samples': 25349632, 'steps': 49510, 'loss/train': 1.8093880414962769}
+02/26/2022 17:42:31 - INFO - codeparrot_training - Step 49511: {'lr': 1.280288752737724e-07, 'samples': 25350144, 'steps': 49511, 'loss/train': 2.0852577686309814}
+02/26/2022 17:42:34 - INFO - codeparrot_training - Step 49512: {'lr': 1.2750581967674312e-07, 'samples': 25350656, 'steps': 49512, 'loss/train': 2.1863291263580322}
+02/26/2022 17:42:37 - INFO - codeparrot_training - Step 49513: {'lr': 1.2698383445416962e-07, 'samples': 25351168, 'steps': 49513, 'loss/train': 0.9368155598640442}
+02/26/2022 17:42:43 - INFO - codeparrot_training - Step 49514: {'lr': 1.2646291960824452e-07, 'samples': 25351680, 'steps': 49514, 'loss/train': 1.6686816215515137}
+02/26/2022 17:42:46 - INFO - codeparrot_training - Step 49515: {'lr': 1.259430751411883e-07, 'samples': 25352192, 'steps': 49515, 'loss/train': 0.6560873985290527}
+02/26/2022 17:42:52 - INFO - codeparrot_training - Step 49516: {'lr': 1.2542430105527693e-07, 'samples': 25352704, 'steps': 49516, 'loss/train': 0.5743892192840576}
+02/26/2022 17:42:55 - INFO - codeparrot_training - Step 49517: {'lr': 1.2490659735267529e-07, 'samples': 25353216, 'steps': 49517, 'loss/train': 0.031562063843011856}
+02/26/2022 17:43:03 - INFO - codeparrot_training - Step 49518: {'lr': 1.243899640356594e-07, 'samples': 25353728, 'steps': 49518, 'loss/train': 1.756022572517395}
+02/26/2022 17:43:06 - INFO - codeparrot_training - Step 49519: {'lr': 1.2387440110639415e-07, 'samples': 25354240, 'steps': 49519, 'loss/train': 1.9736648797988892}
+02/26/2022 17:43:12 - INFO - codeparrot_training - Step 49520: {'lr': 1.233599085671e-07, 'samples': 25354752, 'steps': 49520, 'loss/train': 1.8910934925079346}
+02/26/2022 17:43:15 - INFO - codeparrot_training - Step 49521: {'lr': 1.228464864199974e-07, 'samples': 25355264, 'steps': 49521, 'loss/train': 1.1839945316314697}
+02/26/2022 17:43:21 - INFO - codeparrot_training - Step 49522: {'lr': 1.2233413466727905e-07, 'samples': 25355776, 'steps': 49522, 'loss/train': 1.8331290483474731}
+02/26/2022 17:43:24 - INFO - codeparrot_training - Step 49523: {'lr': 1.2182285331113764e-07, 'samples': 25356288, 'steps': 49523, 'loss/train': 1.3405741453170776}
+02/26/2022 17:43:30 - INFO - codeparrot_training - Step 49524: {'lr': 1.2131264235376582e-07, 'samples': 25356800, 'steps': 49524, 'loss/train': 0.5912362337112427}
+02/26/2022 17:43:33 - INFO - codeparrot_training - Step 49525: {'lr': 1.2080350179732857e-07, 'samples': 25357312, 'steps': 49525, 'loss/train': 2.036501169204712}
+02/26/2022 17:43:39 - INFO - codeparrot_training - Step 49526: {'lr': 1.2029543164404633e-07, 'samples': 25357824, 'steps': 49526, 'loss/train': 0.03968025743961334}
+02/26/2022 17:43:42 - INFO - codeparrot_training - Step 49527: {'lr': 1.1978843189605626e-07, 'samples': 25358336, 'steps': 49527, 'loss/train': 1.017698049545288}
+02/26/2022 17:43:50 - INFO - codeparrot_training - Step 49528: {'lr': 1.1928250255555106e-07, 'samples': 25358848, 'steps': 49528, 'loss/train': 2.15105938911438}
+02/26/2022 17:43:53 - INFO - codeparrot_training - Step 49529: {'lr': 1.1877764362466792e-07, 'samples': 25359360, 'steps': 49529, 'loss/train': 2.18550181388855}
+02/26/2022 17:43:59 - INFO - codeparrot_training - Step 49530: {'lr': 1.1827385510562727e-07, 'samples': 25359872, 'steps': 49530, 'loss/train': 1.3240455389022827}
+02/26/2022 17:44:02 - INFO - codeparrot_training - Step 49531: {'lr': 1.1777113700053854e-07, 'samples': 25360384, 'steps': 49531, 'loss/train': 1.775473713874817}
+02/26/2022 17:44:08 - INFO - codeparrot_training - Step 49532: {'lr': 1.1726948931159443e-07, 'samples': 25360896, 'steps': 49532, 'loss/train': 1.051789402961731}
+02/26/2022 17:44:11 - INFO - codeparrot_training - Step 49533: {'lr': 1.1676891204090434e-07, 'samples': 25361408, 'steps': 49533, 'loss/train': 1.8594964742660522}
+02/26/2022 17:44:17 - INFO - codeparrot_training - Step 49534: {'lr': 1.1626940519063323e-07, 'samples': 25361920, 'steps': 49534, 'loss/train': 2.3752570152282715}
+02/26/2022 17:44:20 - INFO - codeparrot_training - Step 49535: {'lr': 1.1577096876289051e-07, 'samples': 25362432, 'steps': 49535, 'loss/train': 1.1005746126174927}
+02/26/2022 17:44:26 - INFO - codeparrot_training - Step 49536: {'lr': 1.1527360275986887e-07, 'samples': 25362944, 'steps': 49536, 'loss/train': 2.6564674377441406}
+02/26/2022 17:44:29 - INFO - codeparrot_training - Step 49537: {'lr': 1.1477730718367774e-07, 'samples': 25363456, 'steps': 49537, 'loss/train': 2.443544387817383}
+02/26/2022 17:44:35 - INFO - codeparrot_training - Step 49538: {'lr': 1.1428208203639878e-07, 'samples': 25363968, 'steps': 49538, 'loss/train': 2.4494733810424805}
+02/26/2022 17:44:38 - INFO - codeparrot_training - Step 49539: {'lr': 1.1378792732019694e-07, 'samples': 25364480, 'steps': 49539, 'loss/train': 2.5757603645324707}
+02/26/2022 17:44:44 - INFO - codeparrot_training - Step 49540: {'lr': 1.1329484303720939e-07, 'samples': 25364992, 'steps': 49540, 'loss/train': 1.4004231691360474}
+02/26/2022 17:44:47 - INFO - codeparrot_training - Step 49541: {'lr': 1.1280282918949003e-07, 'samples': 25365504, 'steps': 49541, 'loss/train': 1.7510607242584229}
+02/26/2022 17:44:54 - INFO - codeparrot_training - Step 49542: {'lr': 1.1231188577917607e-07, 'samples': 25366016, 'steps': 49542, 'loss/train': 1.7925915718078613}
+02/26/2022 17:44:58 - INFO - codeparrot_training - Step 49543: {'lr': 1.1182201280837689e-07, 'samples': 25366528, 'steps': 49543, 'loss/train': 1.5280436277389526}
+02/26/2022 17:45:03 - INFO - codeparrot_training - Step 49544: {'lr': 1.113332102791742e-07, 'samples': 25367040, 'steps': 49544, 'loss/train': 0.8311648368835449}
+02/26/2022 17:45:07 - INFO - codeparrot_training - Step 49545: {'lr': 1.1084547819370516e-07, 'samples': 25367552, 'steps': 49545, 'loss/train': 2.5234060287475586}
+02/26/2022 17:45:12 - INFO - codeparrot_training - Step 49546: {'lr': 1.1035881655399593e-07, 'samples': 25368064, 'steps': 49546, 'loss/train': 1.327113151550293}
+02/26/2022 17:45:16 - INFO - codeparrot_training - Step 49547: {'lr': 1.0987322536215594e-07, 'samples': 25368576, 'steps': 49547, 'loss/train': 2.671031951904297}
+02/26/2022 17:45:22 - INFO - codeparrot_training - Step 49548: {'lr': 1.0938870462026685e-07, 'samples': 25369088, 'steps': 49548, 'loss/train': 2.4321107864379883}
+02/26/2022 17:45:25 - INFO - codeparrot_training - Step 49549: {'lr': 1.0890525433043808e-07, 'samples': 25369600, 'steps': 49549, 'loss/train': 0.5166988372802734}
+02/26/2022 17:45:31 - INFO - codeparrot_training - Step 49550: {'lr': 1.0842287449469579e-07, 'samples': 25370112, 'steps': 49550, 'loss/train': 1.7249177694320679}
+02/26/2022 17:45:34 - INFO - codeparrot_training - Step 49551: {'lr': 1.079415651150939e-07, 'samples': 25370624, 'steps': 49551, 'loss/train': 2.3032150268554688}
+02/26/2022 17:45:41 - INFO - codeparrot_training - Step 49552: {'lr': 1.0746132619374182e-07, 'samples': 25371136, 'steps': 49552, 'loss/train': 1.8407658338546753}
+02/26/2022 17:45:45 - INFO - codeparrot_training - Step 49553: {'lr': 1.0698215773269349e-07, 'samples': 25371648, 'steps': 49553, 'loss/train': 2.513997793197632}
+02/26/2022 17:45:50 - INFO - codeparrot_training - Step 49554: {'lr': 1.0650405973397504e-07, 'samples': 25372160, 'steps': 49554, 'loss/train': 1.606418490409851}
+02/26/2022 17:45:54 - INFO - codeparrot_training - Step 49555: {'lr': 1.0602703219961263e-07, 'samples': 25372672, 'steps': 49555, 'loss/train': 1.2359482049942017}
+02/26/2022 17:45:59 - INFO - codeparrot_training - Step 49556: {'lr': 1.0555107513171569e-07, 'samples': 25373184, 'steps': 49556, 'loss/train': 1.5361590385437012}
+02/26/2022 17:46:03 - INFO - codeparrot_training - Step 49557: {'lr': 1.0507618853225487e-07, 'samples': 25373696, 'steps': 49557, 'loss/train': 0.9107977747917175}
+02/26/2022 17:46:08 - INFO - codeparrot_training - Step 49558: {'lr': 1.0460237240333959e-07, 'samples': 25374208, 'steps': 49558, 'loss/train': 3.649322271347046}
+02/26/2022 17:46:12 - INFO - codeparrot_training - Step 49559: {'lr': 1.0412962674691273e-07, 'samples': 25374720, 'steps': 49559, 'loss/train': 1.5797438621520996}
+02/26/2022 17:46:17 - INFO - codeparrot_training - Step 49560: {'lr': 1.0365795156508373e-07, 'samples': 25375232, 'steps': 49560, 'loss/train': 0.07243501394987106}
+02/26/2022 17:46:21 - INFO - codeparrot_training - Step 49561: {'lr': 1.0318734685979548e-07, 'samples': 25375744, 'steps': 49561, 'loss/train': 1.9132602214813232}
+02/26/2022 17:46:26 - INFO - codeparrot_training - Step 49562: {'lr': 1.0271781263312963e-07, 'samples': 25376256, 'steps': 49562, 'loss/train': 1.1305632591247559}
+02/26/2022 17:46:30 - INFO - codeparrot_training - Step 49563: {'lr': 1.0224934888705684e-07, 'samples': 25376768, 'steps': 49563, 'loss/train': 1.9545669555664062}
+02/26/2022 17:46:37 - INFO - codeparrot_training - Step 49564: {'lr': 1.0178195562360326e-07, 'samples': 25377280, 'steps': 49564, 'loss/train': 0.019778305664658546}
+02/26/2022 17:46:40 - INFO - codeparrot_training - Step 49565: {'lr': 1.013156328447673e-07, 'samples': 25377792, 'steps': 49565, 'loss/train': 1.5807524919509888}
+02/26/2022 17:46:46 - INFO - codeparrot_training - Step 49566: {'lr': 1.008503805525196e-07, 'samples': 25378304, 'steps': 49566, 'loss/train': 1.3367489576339722}
+02/26/2022 17:46:49 - INFO - codeparrot_training - Step 49567: {'lr': 1.0038619874888633e-07, 'samples': 25378816, 'steps': 49567, 'loss/train': 2.774674654006958}
+02/26/2022 17:46:55 - INFO - codeparrot_training - Step 49568: {'lr': 9.992308743586587e-08, 'samples': 25379328, 'steps': 49568, 'loss/train': 1.610655426979065}
+02/26/2022 17:46:58 - INFO - codeparrot_training - Step 49569: {'lr': 9.946104661540112e-08, 'samples': 25379840, 'steps': 49569, 'loss/train': 1.7148969173431396}
+02/26/2022 17:47:04 - INFO - codeparrot_training - Step 49570: {'lr': 9.900007628951824e-08, 'samples': 25380352, 'steps': 49570, 'loss/train': 1.8496601581573486}
+02/26/2022 17:47:07 - INFO - codeparrot_training - Step 49571: {'lr': 9.854017646013236e-08, 'samples': 25380864, 'steps': 49571, 'loss/train': 1.7861485481262207}
+02/26/2022 17:47:13 - INFO - codeparrot_training - Step 49572: {'lr': 9.808134712926963e-08, 'samples': 25381376, 'steps': 49572, 'loss/train': 2.089369535446167}
+02/26/2022 17:47:16 - INFO - codeparrot_training - Step 49573: {'lr': 9.762358829887297e-08, 'samples': 25381888, 'steps': 49573, 'loss/train': 1.2879939079284668}
+02/26/2022 17:47:24 - INFO - codeparrot_training - Step 49574: {'lr': 9.716689997091299e-08, 'samples': 25382400, 'steps': 49574, 'loss/train': 2.5727152824401855}
+02/26/2022 17:47:27 - INFO - codeparrot_training - Step 49575: {'lr': 9.671128214730484e-08, 'samples': 25382912, 'steps': 49575, 'loss/train': 0.9690009355545044}
+02/26/2022 17:47:33 - INFO - codeparrot_training - Step 49576: {'lr': 9.625673483007469e-08, 'samples': 25383424, 'steps': 49576, 'loss/train': 1.8034100532531738}
+02/26/2022 17:47:36 - INFO - codeparrot_training - Step 49577: {'lr': 9.580325802108214e-08, 'samples': 25383936, 'steps': 49577, 'loss/train': 0.6103371977806091}
+02/26/2022 17:47:42 - INFO - codeparrot_training - Step 49578: {'lr': 9.535085172235336e-08, 'samples': 25384448, 'steps': 49578, 'loss/train': 1.089733600616455}
+02/26/2022 17:47:45 - INFO - codeparrot_training - Step 49579: {'lr': 9.489951593577573e-08, 'samples': 25384960, 'steps': 49579, 'loss/train': 1.1412036418914795}
+02/26/2022 17:47:51 - INFO - codeparrot_training - Step 49580: {'lr': 9.444925066329213e-08, 'samples': 25385472, 'steps': 49580, 'loss/train': 1.5451005697250366}
+02/26/2022 17:47:54 - INFO - codeparrot_training - Step 49581: {'lr': 9.40000559068177e-08, 'samples': 25385984, 'steps': 49581, 'loss/train': 2.1165547370910645}
+02/26/2022 17:48:00 - INFO - codeparrot_training - Step 49582: {'lr': 9.35519316683231e-08, 'samples': 25386496, 'steps': 49582, 'loss/train': 1.5899916887283325}
+02/26/2022 17:48:03 - INFO - codeparrot_training - Step 49583: {'lr': 9.310487794966793e-08, 'samples': 25387008, 'steps': 49583, 'loss/train': 1.745990514755249}
+02/26/2022 17:48:09 - INFO - codeparrot_training - Step 49584: {'lr': 9.265889475282285e-08, 'samples': 25387520, 'steps': 49584, 'loss/train': 1.5534600019454956}
+02/26/2022 17:48:12 - INFO - codeparrot_training - Step 49585: {'lr': 9.221398207964748e-08, 'samples': 25388032, 'steps': 49585, 'loss/train': 1.450218677520752}
+02/26/2022 17:48:18 - INFO - codeparrot_training - Step 49586: {'lr': 9.177013993208471e-08, 'samples': 25388544, 'steps': 49586, 'loss/train': 1.0296509265899658}
+02/26/2022 17:48:21 - INFO - codeparrot_training - Step 49587: {'lr': 9.132736831202193e-08, 'samples': 25389056, 'steps': 49587, 'loss/train': 1.8154584169387817}
+02/26/2022 17:48:27 - INFO - codeparrot_training - Step 49588: {'lr': 9.088566722137425e-08, 'samples': 25389568, 'steps': 49588, 'loss/train': 1.367182970046997}
+02/26/2022 17:48:30 - INFO - codeparrot_training - Step 49589: {'lr': 9.044503666200132e-08, 'samples': 25390080, 'steps': 49589, 'loss/train': 1.768072247505188}
+02/26/2022 17:48:38 - INFO - codeparrot_training - Step 49590: {'lr': 9.000547663581826e-08, 'samples': 25390592, 'steps': 49590, 'loss/train': 1.7978105545043945}
+02/26/2022 17:48:41 - INFO - codeparrot_training - Step 49591: {'lr': 8.956698714468469e-08, 'samples': 25391104, 'steps': 49591, 'loss/train': 0.29134321212768555}
+02/26/2022 17:48:47 - INFO - codeparrot_training - Step 49592: {'lr': 8.912956819048801e-08, 'samples': 25391616, 'steps': 49592, 'loss/train': 2.074617862701416}
+02/26/2022 17:48:50 - INFO - codeparrot_training - Step 49593: {'lr': 8.869321977511557e-08, 'samples': 25392128, 'steps': 49593, 'loss/train': 1.7583087682724}
+02/26/2022 17:48:57 - INFO - codeparrot_training - Step 49594: {'lr': 8.825794190042703e-08, 'samples': 25392640, 'steps': 49594, 'loss/train': 1.9845067262649536}
+02/26/2022 17:49:00 - INFO - codeparrot_training - Step 49595: {'lr': 8.782373456828197e-08, 'samples': 25393152, 'steps': 49595, 'loss/train': 0.9412076473236084}
+02/26/2022 17:49:04 - INFO - codeparrot_training - Step 49596: {'lr': 8.739059778054004e-08, 'samples': 25393664, 'steps': 49596, 'loss/train': 2.2576260566711426}
+02/26/2022 17:49:09 - INFO - codeparrot_training - Step 49597: {'lr': 8.695853153906086e-08, 'samples': 25394176, 'steps': 49597, 'loss/train': 1.793504238128662}
+02/26/2022 17:49:13 - INFO - codeparrot_training - Step 49598: {'lr': 8.652753584567629e-08, 'samples': 25394688, 'steps': 49598, 'loss/train': 1.7318885326385498}
+02/26/2022 17:49:18 - INFO - codeparrot_training - Step 49599: {'lr': 8.609761070227373e-08, 'samples': 25395200, 'steps': 49599, 'loss/train': 2.2207977771759033}
+02/26/2022 17:49:22 - INFO - codeparrot_training - Step 49600: {'lr': 8.566875611068504e-08, 'samples': 25395712, 'steps': 49600, 'loss/train': 1.7379605770111084}
+02/26/2022 17:49:29 - INFO - codeparrot_training - Step 49601: {'lr': 8.52409720727143e-08, 'samples': 25396224, 'steps': 49601, 'loss/train': 1.3713390827178955}
+02/26/2022 17:49:33 - INFO - codeparrot_training - Step 49602: {'lr': 8.481425859022118e-08, 'samples': 25396736, 'steps': 49602, 'loss/train': 2.2123289108276367}
+02/26/2022 17:49:38 - INFO - codeparrot_training - Step 49603: {'lr': 8.438861566503752e-08, 'samples': 25397248, 'steps': 49603, 'loss/train': 2.246407985687256}
+02/26/2022 17:49:42 - INFO - codeparrot_training - Step 49604: {'lr': 8.396404329893969e-08, 'samples': 25397760, 'steps': 49604, 'loss/train': 2.2773749828338623}
+02/26/2022 17:49:47 - INFO - codeparrot_training - Step 49605: {'lr': 8.354054149381507e-08, 'samples': 25398272, 'steps': 49605, 'loss/train': 2.126040458679199}
+02/26/2022 17:49:51 - INFO - codeparrot_training - Step 49606: {'lr': 8.311811025144e-08, 'samples': 25398784, 'steps': 49606, 'loss/train': 2.6375536918640137}
+02/26/2022 17:49:56 - INFO - codeparrot_training - Step 49607: {'lr': 8.269674957361862e-08, 'samples': 25399296, 'steps': 49607, 'loss/train': 1.9615708589553833}
+02/26/2022 17:50:00 - INFO - codeparrot_training - Step 49608: {'lr': 8.227645946218276e-08, 'samples': 25399808, 'steps': 49608, 'loss/train': 1.1938278675079346}
+02/26/2022 17:50:05 - INFO - codeparrot_training - Step 49609: {'lr': 8.185723991890881e-08, 'samples': 25400320, 'steps': 49609, 'loss/train': 2.205256462097168}
+02/26/2022 17:50:09 - INFO - codeparrot_training - Step 49610: {'lr': 8.143909094560088e-08, 'samples': 25400832, 'steps': 49610, 'loss/train': 1.3597317934036255}
+02/26/2022 17:50:16 - INFO - codeparrot_training - Step 49611: {'lr': 8.10220125440353e-08, 'samples': 25401344, 'steps': 49611, 'loss/train': 2.3371031284332275}
+02/26/2022 17:50:20 - INFO - codeparrot_training - Step 49612: {'lr': 8.060600471604396e-08, 'samples': 25401856, 'steps': 49612, 'loss/train': 1.0389764308929443}
+02/26/2022 17:50:25 - INFO - codeparrot_training - Step 49613: {'lr': 8.01910674633477e-08, 'samples': 25402368, 'steps': 49613, 'loss/train': 3.1536383628845215}
+02/26/2022 17:50:28 - INFO - codeparrot_training - Step 49614: {'lr': 7.97772007877784e-08, 'samples': 25402880, 'steps': 49614, 'loss/train': 0.0577298179268837}
+02/26/2022 17:50:34 - INFO - codeparrot_training - Step 49615: {'lr': 7.936440469105688e-08, 'samples': 25403392, 'steps': 49615, 'loss/train': 0.6977536678314209}
+02/26/2022 17:50:37 - INFO - codeparrot_training - Step 49616: {'lr': 7.895267917501503e-08, 'samples': 25403904, 'steps': 49616, 'loss/train': 1.7199970483779907}
+02/26/2022 17:50:43 - INFO - codeparrot_training - Step 49617: {'lr': 7.854202424134593e-08, 'samples': 25404416, 'steps': 49617, 'loss/train': 1.774612307548523}
+02/26/2022 17:50:46 - INFO - codeparrot_training - Step 49618: {'lr': 7.81324398918537e-08, 'samples': 25404928, 'steps': 49618, 'loss/train': 2.8690364360809326}
+02/26/2022 17:50:52 - INFO - codeparrot_training - Step 49619: {'lr': 7.772392612831469e-08, 'samples': 25405440, 'steps': 49619, 'loss/train': 0.31925228238105774}
+02/26/2022 17:50:55 - INFO - codeparrot_training - Step 49620: {'lr': 7.7316482952422e-08, 'samples': 25405952, 'steps': 49620, 'loss/train': 1.7292675971984863}
+02/26/2022 17:51:02 - INFO - codeparrot_training - Step 49621: {'lr': 7.691011036592421e-08, 'samples': 25406464, 'steps': 49621, 'loss/train': 2.1880085468292236}
+02/26/2022 17:51:06 - INFO - codeparrot_training - Step 49622: {'lr': 7.650480837062546e-08, 'samples': 25406976, 'steps': 49622, 'loss/train': 1.6229488849639893}
+02/26/2022 17:51:11 - INFO - codeparrot_training - Step 49623: {'lr': 7.610057696819106e-08, 'samples': 25407488, 'steps': 49623, 'loss/train': 1.0848151445388794}
+02/26/2022 17:51:15 - INFO - codeparrot_training - Step 49624: {'lr': 7.569741616039738e-08, 'samples': 25408000, 'steps': 49624, 'loss/train': 1.605403184890747}
+02/26/2022 17:51:20 - INFO - codeparrot_training - Step 49625: {'lr': 7.529532594893751e-08, 'samples': 25408512, 'steps': 49625, 'loss/train': 3.2828586101531982}
+02/26/2022 17:51:24 - INFO - codeparrot_training - Step 49626: {'lr': 7.489430633556005e-08, 'samples': 25409024, 'steps': 49626, 'loss/train': 1.3865106105804443}
+02/26/2022 17:51:29 - INFO - codeparrot_training - Step 49627: {'lr': 7.449435732198584e-08, 'samples': 25409536, 'steps': 49627, 'loss/train': 1.3934231996536255}
+02/26/2022 17:51:33 - INFO - codeparrot_training - Step 49628: {'lr': 7.409547890993573e-08, 'samples': 25410048, 'steps': 49628, 'loss/train': 1.8808860778808594}
+02/26/2022 17:51:39 - INFO - codeparrot_training - Step 49629: {'lr': 7.369767110107507e-08, 'samples': 25410560, 'steps': 49629, 'loss/train': 2.0115225315093994}
+02/26/2022 17:51:42 - INFO - codeparrot_training - Step 49630: {'lr': 7.330093389712466e-08, 'samples': 25411072, 'steps': 49630, 'loss/train': 2.3045902252197266}
+02/26/2022 17:51:46 - INFO - codeparrot_training - Step 49631: {'lr': 7.29052672998054e-08, 'samples': 25411584, 'steps': 49631, 'loss/train': 1.2370697259902954}
+02/26/2022 17:51:51 - INFO - codeparrot_training - Step 49632: {'lr': 7.25106713107826e-08, 'samples': 25412096, 'steps': 49632, 'loss/train': 1.484943151473999}
+02/26/2022 17:51:55 - INFO - codeparrot_training - Step 49633: {'lr': 7.211714593177709e-08, 'samples': 25412608, 'steps': 49633, 'loss/train': 2.5384371280670166}
+02/26/2022 17:52:00 - INFO - codeparrot_training - Step 49634: {'lr': 7.172469116448199e-08, 'samples': 25413120, 'steps': 49634, 'loss/train': 1.9349697828292847}
+02/26/2022 17:52:04 - INFO - codeparrot_training - Step 49635: {'lr': 7.133330701050711e-08, 'samples': 25413632, 'steps': 49635, 'loss/train': 2.10516619682312}
+02/26/2022 17:52:11 - INFO - codeparrot_training - Step 49636: {'lr': 7.094299347162881e-08, 'samples': 25414144, 'steps': 49636, 'loss/train': 1.5242670774459839}
+02/26/2022 17:52:14 - INFO - codeparrot_training - Step 49637: {'lr': 7.055375054942914e-08, 'samples': 25414656, 'steps': 49637, 'loss/train': 0.8418490886688232}
+02/26/2022 17:52:20 - INFO - codeparrot_training - Step 49638: {'lr': 7.016557824565672e-08, 'samples': 25415168, 'steps': 49638, 'loss/train': 1.4000853300094604}
+02/26/2022 17:52:23 - INFO - codeparrot_training - Step 49639: {'lr': 6.977847656189362e-08, 'samples': 25415680, 'steps': 49639, 'loss/train': 1.5796009302139282}
+02/26/2022 17:52:29 - INFO - codeparrot_training - Step 49640: {'lr': 6.939244549986068e-08, 'samples': 25416192, 'steps': 49640, 'loss/train': 1.6645773649215698}
+02/26/2022 17:52:33 - INFO - codeparrot_training - Step 49641: {'lr': 6.900748506119547e-08, 'samples': 25416704, 'steps': 49641, 'loss/train': 0.4310552179813385}
+02/26/2022 17:52:38 - INFO - codeparrot_training - Step 49642: {'lr': 6.862359524750782e-08, 'samples': 25417216, 'steps': 49642, 'loss/train': 1.2219706773757935}
+02/26/2022 17:52:41 - INFO - codeparrot_training - Step 49643: {'lr': 6.824077606049083e-08, 'samples': 25417728, 'steps': 49643, 'loss/train': 1.665999174118042}
+02/26/2022 17:52:47 - INFO - codeparrot_training - Step 49644: {'lr': 6.785902750178208e-08, 'samples': 25418240, 'steps': 49644, 'loss/train': 1.4190175533294678}
+02/26/2022 17:52:50 - INFO - codeparrot_training - Step 49645: {'lr': 6.747834957299137e-08, 'samples': 25418752, 'steps': 49645, 'loss/train': 1.5200834274291992}
+02/26/2022 17:52:57 - INFO - codeparrot_training - Step 49646: {'lr': 6.70987422757563e-08, 'samples': 25419264, 'steps': 49646, 'loss/train': 1.9355356693267822}
+02/26/2022 17:53:00 - INFO - codeparrot_training - Step 49647: {'lr': 6.672020561171443e-08, 'samples': 25419776, 'steps': 49647, 'loss/train': 2.010352849960327}
+02/26/2022 17:53:06 - INFO - codeparrot_training - Step 49648: {'lr': 6.63427395824756e-08, 'samples': 25420288, 'steps': 49648, 'loss/train': 1.577196717262268}
+02/26/2022 17:53:09 - INFO - codeparrot_training - Step 49649: {'lr': 6.59663441896774e-08, 'samples': 25420800, 'steps': 49649, 'loss/train': 2.2887322902679443}
+02/26/2022 17:53:15 - INFO - codeparrot_training - Step 49650: {'lr': 6.559101943490186e-08, 'samples': 25421312, 'steps': 49650, 'loss/train': 1.492795467376709}
+02/26/2022 17:53:18 - INFO - codeparrot_training - Step 49651: {'lr': 6.52167653197866e-08, 'samples': 25421824, 'steps': 49651, 'loss/train': 1.6093878746032715}
+02/26/2022 17:53:24 - INFO - codeparrot_training - Step 49652: {'lr': 6.48435818458859e-08, 'samples': 25422336, 'steps': 49652, 'loss/train': 2.2008275985717773}
+02/26/2022 17:53:27 - INFO - codeparrot_training - Step 49653: {'lr': 6.447146901486511e-08, 'samples': 25422848, 'steps': 49653, 'loss/train': 1.3510240316390991}
+02/26/2022 17:53:33 - INFO - codeparrot_training - Step 49654: {'lr': 6.410042682827855e-08, 'samples': 25423360, 'steps': 49654, 'loss/train': 1.2156535387039185}
+02/26/2022 17:53:36 - INFO - codeparrot_training - Step 49655: {'lr': 6.373045528770826e-08, 'samples': 25423872, 'steps': 49655, 'loss/train': 2.296717882156372}
+02/26/2022 17:53:42 - INFO - codeparrot_training - Step 49656: {'lr': 6.33615543947641e-08, 'samples': 25424384, 'steps': 49656, 'loss/train': 1.5778707265853882}
+02/26/2022 17:53:46 - INFO - codeparrot_training - Step 49657: {'lr': 6.299372415100036e-08, 'samples': 25424896, 'steps': 49657, 'loss/train': 1.9722676277160645}
+02/26/2022 17:53:51 - INFO - codeparrot_training - Step 49658: {'lr': 6.262696455802685e-08, 'samples': 25425408, 'steps': 49658, 'loss/train': 2.2931156158447266}
+02/26/2022 17:53:55 - INFO - codeparrot_training - Step 49659: {'lr': 6.226127561739792e-08, 'samples': 25425920, 'steps': 49659, 'loss/train': 2.960163116455078}
+02/26/2022 17:54:00 - INFO - codeparrot_training - Step 49660: {'lr': 6.189665733066785e-08, 'samples': 25426432, 'steps': 49660, 'loss/train': 1.4128910303115845}
+02/26/2022 17:54:04 - INFO - codeparrot_training - Step 49661: {'lr': 6.153310969939096e-08, 'samples': 25426944, 'steps': 49661, 'loss/train': 0.6558352708816528}
+02/26/2022 17:54:09 - INFO - codeparrot_training - Step 49662: {'lr': 6.11706327251771e-08, 'samples': 25427456, 'steps': 49662, 'loss/train': 0.646000862121582}
+02/26/2022 17:54:13 - INFO - codeparrot_training - Step 49663: {'lr': 6.080922640952502e-08, 'samples': 25427968, 'steps': 49663, 'loss/train': 2.069631814956665}
+02/26/2022 17:54:19 - INFO - codeparrot_training - Step 49664: {'lr': 6.044889075398908e-08, 'samples': 25428480, 'steps': 49664, 'loss/train': 1.1431056261062622}
+02/26/2022 17:54:23 - INFO - codeparrot_training - Step 49665: {'lr': 6.008962576015132e-08, 'samples': 25428992, 'steps': 49665, 'loss/train': 1.4981911182403564}
+02/26/2022 17:54:28 - INFO - codeparrot_training - Step 49666: {'lr': 5.973143142951054e-08, 'samples': 25429504, 'steps': 49666, 'loss/train': 1.0248603820800781}
+02/26/2022 17:54:32 - INFO - codeparrot_training - Step 49667: {'lr': 5.9374307763621075e-08, 'samples': 25430016, 'steps': 49667, 'loss/train': 0.1982443928718567}
+02/26/2022 17:54:37 - INFO - codeparrot_training - Step 49668: {'lr': 5.9018254763981705e-08, 'samples': 25430528, 'steps': 49668, 'loss/train': 1.9074077606201172}
+02/26/2022 17:54:41 - INFO - codeparrot_training - Step 49669: {'lr': 5.866327243217451e-08, 'samples': 25431040, 'steps': 49669, 'loss/train': 2.692355155944824}
+02/26/2022 17:54:46 - INFO - codeparrot_training - Step 49670: {'lr': 5.830936076967053e-08, 'samples': 25431552, 'steps': 49670, 'loss/train': 1.570583701133728}
+02/26/2022 17:54:50 - INFO - codeparrot_training - Step 49671: {'lr': 5.795651977802408e-08, 'samples': 25432064, 'steps': 49671, 'loss/train': 1.241619348526001}
+02/26/2022 17:54:55 - INFO - codeparrot_training - Step 49672: {'lr': 5.7604749458706194e-08, 'samples': 25432576, 'steps': 49672, 'loss/train': 1.0222573280334473}
+02/26/2022 17:54:59 - INFO - codeparrot_training - Step 49673: {'lr': 5.725404981324345e-08, 'samples': 25433088, 'steps': 49673, 'loss/train': 0.6399295926094055}
+02/26/2022 17:55:04 - INFO - codeparrot_training - Step 49674: {'lr': 5.690442084313463e-08, 'samples': 25433600, 'steps': 49674, 'loss/train': 2.2612533569335938}
+02/26/2022 17:55:08 - INFO - codeparrot_training - Step 49675: {'lr': 5.65558625499063e-08, 'samples': 25434112, 'steps': 49675, 'loss/train': 1.8178939819335938}
+02/26/2022 17:55:14 - INFO - codeparrot_training - Step 49676: {'lr': 5.620837493500175e-08, 'samples': 25434624, 'steps': 49676, 'loss/train': 2.5089809894561768}
+02/26/2022 17:55:17 - INFO - codeparrot_training - Step 49677: {'lr': 5.586195799994753e-08, 'samples': 25435136, 'steps': 49677, 'loss/train': 1.1936006546020508}
+02/26/2022 17:55:23 - INFO - codeparrot_training - Step 49678: {'lr': 5.55166117462147e-08, 'samples': 25435648, 'steps': 49678, 'loss/train': 9.683402061462402}
+02/26/2022 17:55:27 - INFO - codeparrot_training - Step 49679: {'lr': 5.5172336175274285e-08, 'samples': 25436160, 'steps': 49679, 'loss/train': 1.7998840808868408}
+02/26/2022 17:55:32 - INFO - codeparrot_training - Step 49680: {'lr': 5.48291312886251e-08, 'samples': 25436672, 'steps': 49680, 'loss/train': 1.4408538341522217}
+02/26/2022 17:55:36 - INFO - codeparrot_training - Step 49681: {'lr': 5.448699708771043e-08, 'samples': 25437184, 'steps': 49681, 'loss/train': 1.694828987121582}
+02/26/2022 17:55:42 - INFO - codeparrot_training - Step 49682: {'lr': 5.414593357400133e-08, 'samples': 25437696, 'steps': 49682, 'loss/train': 1.576622724533081}
+02/26/2022 17:55:45 - INFO - codeparrot_training - Step 49683: {'lr': 5.380594074896883e-08, 'samples': 25438208, 'steps': 49683, 'loss/train': 2.116130828857422}
+02/26/2022 17:55:51 - INFO - codeparrot_training - Step 49684: {'lr': 5.346701861408398e-08, 'samples': 25438720, 'steps': 49684, 'loss/train': 1.9564499855041504}
+02/26/2022 17:55:54 - INFO - codeparrot_training - Step 49685: {'lr': 5.312916717076233e-08, 'samples': 25439232, 'steps': 49685, 'loss/train': 1.2669018507003784}
+02/26/2022 17:56:00 - INFO - codeparrot_training - Step 49686: {'lr': 5.27923864204749e-08, 'samples': 25439744, 'steps': 49686, 'loss/train': 1.9539506435394287}
+02/26/2022 17:56:03 - INFO - codeparrot_training - Step 49687: {'lr': 5.245667636463725e-08, 'samples': 25440256, 'steps': 49687, 'loss/train': 1.2277711629867554}
+02/26/2022 17:56:09 - INFO - codeparrot_training - Step 49688: {'lr': 5.21220370047204e-08, 'samples': 25440768, 'steps': 49688, 'loss/train': 2.543898105621338}
+02/26/2022 17:56:12 - INFO - codeparrot_training - Step 49689: {'lr': 5.1788468342139905e-08, 'samples': 25441280, 'steps': 49689, 'loss/train': 1.5499975681304932}
+02/26/2022 17:56:18 - INFO - codeparrot_training - Step 49690: {'lr': 5.145597037833904e-08, 'samples': 25441792, 'steps': 49690, 'loss/train': 1.2460930347442627}
+02/26/2022 17:56:21 - INFO - codeparrot_training - Step 49691: {'lr': 5.112454311473336e-08, 'samples': 25442304, 'steps': 49691, 'loss/train': 1.6725564002990723}
+02/26/2022 17:56:27 - INFO - codeparrot_training - Step 49692: {'lr': 5.0794186552738375e-08, 'samples': 25442816, 'steps': 49692, 'loss/train': 1.2821602821350098}
+02/26/2022 17:56:31 - INFO - codeparrot_training - Step 49693: {'lr': 5.046490069379739e-08, 'samples': 25443328, 'steps': 49693, 'loss/train': 2.3357479572296143}
+02/26/2022 17:56:36 - INFO - codeparrot_training - Step 49694: {'lr': 5.0136685539270424e-08, 'samples': 25443840, 'steps': 49694, 'loss/train': 0.6887544989585876}
+02/26/2022 17:56:40 - INFO - codeparrot_training - Step 49695: {'lr': 4.9809541090573005e-08, 'samples': 25444352, 'steps': 49695, 'loss/train': 3.190800189971924}
+02/26/2022 17:56:45 - INFO - codeparrot_training - Step 49696: {'lr': 4.948346734914844e-08, 'samples': 25444864, 'steps': 49696, 'loss/train': 1.0308183431625366}
+02/26/2022 17:56:49 - INFO - codeparrot_training - Step 49697: {'lr': 4.9158464316384486e-08, 'samples': 25445376, 'steps': 49697, 'loss/train': 2.798590898513794}
+02/26/2022 17:56:54 - INFO - codeparrot_training - Step 49698: {'lr': 4.883453199361343e-08, 'samples': 25445888, 'steps': 49698, 'loss/train': 1.6407665014266968}
+02/26/2022 17:56:58 - INFO - codeparrot_training - Step 49699: {'lr': 4.851167038230631e-08, 'samples': 25446400, 'steps': 49699, 'loss/train': 2.0579278469085693}
+02/26/2022 17:57:03 - INFO - codeparrot_training - Step 49700: {'lr': 4.818987948379538e-08, 'samples': 25446912, 'steps': 49700, 'loss/train': 2.0435235500335693}
+02/26/2022 17:57:07 - INFO - codeparrot_training - Step 49701: {'lr': 4.7869159299468443e-08, 'samples': 25447424, 'steps': 49701, 'loss/train': 3.1509881019592285}
+02/26/2022 17:57:13 - INFO - codeparrot_training - Step 49702: {'lr': 4.7549509830685514e-08, 'samples': 25447936, 'steps': 49702, 'loss/train': 1.106390118598938}
+02/26/2022 17:57:17 - INFO - codeparrot_training - Step 49703: {'lr': 4.723093107883436e-08, 'samples': 25448448, 'steps': 49703, 'loss/train': 1.3935894966125488}
+02/26/2022 17:57:22 - INFO - codeparrot_training - Step 49704: {'lr': 4.6913423045302773e-08, 'samples': 25448960, 'steps': 49704, 'loss/train': 1.944941520690918}
+02/26/2022 17:57:26 - INFO - codeparrot_training - Step 49705: {'lr': 4.659698573139526e-08, 'samples': 25449472, 'steps': 49705, 'loss/train': 1.6275675296783447}
+02/26/2022 17:57:31 - INFO - codeparrot_training - Step 49706: {'lr': 4.628161913852735e-08, 'samples': 25449984, 'steps': 49706, 'loss/train': 1.9497686624526978}
+02/26/2022 17:57:35 - INFO - codeparrot_training - Step 49707: {'lr': 4.596732326800357e-08, 'samples': 25450496, 'steps': 49707, 'loss/train': 1.7830393314361572}
+02/26/2022 17:57:40 - INFO - codeparrot_training - Step 49708: {'lr': 4.565409812118393e-08, 'samples': 25451008, 'steps': 49708, 'loss/train': 2.2010281085968018}
+02/26/2022 17:57:44 - INFO - codeparrot_training - Step 49709: {'lr': 4.53419436994007e-08, 'samples': 25451520, 'steps': 49709, 'loss/train': 0.8849744200706482}
+02/26/2022 17:57:50 - INFO - codeparrot_training - Step 49710: {'lr': 4.503086000404166e-08, 'samples': 25452032, 'steps': 49710, 'loss/train': 2.3053812980651855}
+02/26/2022 17:57:53 - INFO - codeparrot_training - Step 49711: {'lr': 4.4720847036355814e-08, 'samples': 25452544, 'steps': 49711, 'loss/train': 0.9687795042991638}
+02/26/2022 17:57:59 - INFO - codeparrot_training - Step 49712: {'lr': 4.4411904797758695e-08, 'samples': 25453056, 'steps': 49712, 'loss/train': 2.7189462184906006}
+02/26/2022 17:58:03 - INFO - codeparrot_training - Step 49713: {'lr': 4.41040332894993e-08, 'samples': 25453568, 'steps': 49713, 'loss/train': 0.98175448179245}
+02/26/2022 17:58:08 - INFO - codeparrot_training - Step 49714: {'lr': 4.3797232512937655e-08, 'samples': 25454080, 'steps': 49714, 'loss/train': 1.117993950843811}
+02/26/2022 17:58:12 - INFO - codeparrot_training - Step 49715: {'lr': 4.3491502469378275e-08, 'samples': 25454592, 'steps': 49715, 'loss/train': 1.2794795036315918}
+02/26/2022 17:58:17 - INFO - codeparrot_training - Step 49716: {'lr': 4.3186843160125664e-08, 'samples': 25455104, 'steps': 49716, 'loss/train': 1.9968023300170898}
+02/26/2022 17:58:21 - INFO - codeparrot_training - Step 49717: {'lr': 4.28832545865121e-08, 'samples': 25455616, 'steps': 49717, 'loss/train': 2.224591016769409}
+02/26/2022 17:58:27 - INFO - codeparrot_training - Step 49718: {'lr': 4.2580736749786575e-08, 'samples': 25456128, 'steps': 49718, 'loss/train': 2.2531254291534424}
+02/26/2022 17:58:31 - INFO - codeparrot_training - Step 49719: {'lr': 4.227928965130912e-08, 'samples': 25456640, 'steps': 49719, 'loss/train': 1.970960021018982}
+02/26/2022 17:58:34 - INFO - codeparrot_training - Step 49720: {'lr': 4.197891329230097e-08, 'samples': 25457152, 'steps': 49720, 'loss/train': 2.8190581798553467}
+02/26/2022 17:58:40 - INFO - codeparrot_training - Step 49721: {'lr': 4.16796076740944e-08, 'samples': 25457664, 'steps': 49721, 'loss/train': 1.9687682390213013}
+02/26/2022 17:58:43 - INFO - codeparrot_training - Step 49722: {'lr': 4.138137279799392e-08, 'samples': 25458176, 'steps': 49722, 'loss/train': 2.0611155033111572}
+02/26/2022 17:58:49 - INFO - codeparrot_training - Step 49723: {'lr': 4.1084208665193025e-08, 'samples': 25458688, 'steps': 49723, 'loss/train': 1.9255492687225342}
+02/26/2022 17:58:53 - INFO - codeparrot_training - Step 49724: {'lr': 4.0788115277051727e-08, 'samples': 25459200, 'steps': 49724, 'loss/train': 1.390259027481079}
+02/26/2022 17:58:59 - INFO - codeparrot_training - Step 49725: {'lr': 4.0493092634791286e-08, 'samples': 25459712, 'steps': 49725, 'loss/train': 1.708520770072937}
+02/26/2022 17:59:02 - INFO - codeparrot_training - Step 49726: {'lr': 4.019914073968844e-08, 'samples': 25460224, 'steps': 49726, 'loss/train': 2.6933388710021973}
+02/26/2022 17:59:07 - INFO - codeparrot_training - Step 49727: {'lr': 3.990625959301997e-08, 'samples': 25460736, 'steps': 49727, 'loss/train': 1.3498625755310059}
+02/26/2022 17:59:11 - INFO - codeparrot_training - Step 49728: {'lr': 3.96144491960071e-08, 'samples': 25461248, 'steps': 49728, 'loss/train': 1.728865623474121}
+02/26/2022 17:59:16 - INFO - codeparrot_training - Step 49729: {'lr': 3.932370954989884e-08, 'samples': 25461760, 'steps': 49729, 'loss/train': 1.798843502998352}
+02/26/2022 17:59:20 - INFO - codeparrot_training - Step 49730: {'lr': 3.903404065597194e-08, 'samples': 25462272, 'steps': 49730, 'loss/train': 1.8355815410614014}
+02/26/2022 17:59:25 - INFO - codeparrot_training - Step 49731: {'lr': 3.874544251544765e-08, 'samples': 25462784, 'steps': 49731, 'loss/train': 1.0345803499221802}
+02/26/2022 17:59:29 - INFO - codeparrot_training - Step 49732: {'lr': 3.845791512957497e-08, 'samples': 25463296, 'steps': 49732, 'loss/train': 1.8129682540893555}
+02/26/2022 17:59:34 - INFO - codeparrot_training - Step 49733: {'lr': 3.8171458499575154e-08, 'samples': 25463808, 'steps': 49733, 'loss/train': 1.964986801147461}
+02/26/2022 17:59:38 - INFO - codeparrot_training - Step 49734: {'lr': 3.7886072626669434e-08, 'samples': 25464320, 'steps': 49734, 'loss/train': 1.252440094947815}
+02/26/2022 17:59:43 - INFO - codeparrot_training - Step 49735: {'lr': 3.760175751207906e-08, 'samples': 25464832, 'steps': 49735, 'loss/train': 1.0094990730285645}
+02/26/2022 17:59:47 - INFO - codeparrot_training - Step 49736: {'lr': 3.7318513157053036e-08, 'samples': 25465344, 'steps': 49736, 'loss/train': 1.6370432376861572}
+02/26/2022 17:59:52 - INFO - codeparrot_training - Step 49737: {'lr': 3.703633956275709e-08, 'samples': 25465856, 'steps': 49737, 'loss/train': 0.5853754878044128}
+02/26/2022 17:59:56 - INFO - codeparrot_training - Step 49738: {'lr': 3.6755236730440235e-08, 'samples': 25466368, 'steps': 49738, 'loss/train': 1.8676743507385254}
+02/26/2022 18:00:02 - INFO - codeparrot_training - Step 49739: {'lr': 3.6475204661295946e-08, 'samples': 25466880, 'steps': 49739, 'loss/train': 0.8303935527801514}
+02/26/2022 18:00:05 - INFO - codeparrot_training - Step 49740: {'lr': 3.619624335651772e-08, 'samples': 25467392, 'steps': 49740, 'loss/train': 0.33742159605026245}
+02/26/2022 18:00:11 - INFO - codeparrot_training - Step 49741: {'lr': 3.591835281729905e-08, 'samples': 25467904, 'steps': 49741, 'loss/train': 2.738327980041504}
+02/26/2022 18:00:14 - INFO - codeparrot_training - Step 49742: {'lr': 3.5641533044833417e-08, 'samples': 25468416, 'steps': 49742, 'loss/train': 1.4374414682388306}
+02/26/2022 18:00:20 - INFO - codeparrot_training - Step 49743: {'lr': 3.536578404031432e-08, 'samples': 25468928, 'steps': 49743, 'loss/train': 1.5922343730926514}
+02/26/2022 18:00:23 - INFO - codeparrot_training - Step 49744: {'lr': 3.509110580490749e-08, 'samples': 25469440, 'steps': 49744, 'loss/train': 0.9981396794319153}
+02/26/2022 18:00:29 - INFO - codeparrot_training - Step 49745: {'lr': 3.481749833977865e-08, 'samples': 25469952, 'steps': 49745, 'loss/train': 1.3962358236312866}
+02/26/2022 18:00:32 - INFO - codeparrot_training - Step 49746: {'lr': 3.4544961646149065e-08, 'samples': 25470464, 'steps': 49746, 'loss/train': 2.002027988433838}
+02/26/2022 18:00:38 - INFO - codeparrot_training - Step 49747: {'lr': 3.42734957251567e-08, 'samples': 25470976, 'steps': 49747, 'loss/train': 0.1796267330646515}
+02/26/2022 18:00:41 - INFO - codeparrot_training - Step 49748: {'lr': 3.4003100577939536e-08, 'samples': 25471488, 'steps': 49748, 'loss/train': 1.6250418424606323}
+02/26/2022 18:00:48 - INFO - codeparrot_training - Step 49749: {'lr': 3.373377620569107e-08, 'samples': 25472000, 'steps': 49749, 'loss/train': 1.3040080070495605}
+02/26/2022 18:00:51 - INFO - codeparrot_training - Step 49750: {'lr': 3.346552260954927e-08, 'samples': 25472512, 'steps': 49750, 'loss/train': 1.5120213031768799}
+02/26/2022 18:00:57 - INFO - codeparrot_training - Step 49751: {'lr': 3.3198339790679874e-08, 'samples': 25473024, 'steps': 49751, 'loss/train': 1.7201801538467407}
+02/26/2022 18:01:00 - INFO - codeparrot_training - Step 49752: {'lr': 3.293222775022087e-08, 'samples': 25473536, 'steps': 49752, 'loss/train': 2.538609027862549}
+02/26/2022 18:01:06 - INFO - codeparrot_training - Step 49753: {'lr': 3.2667186489282465e-08, 'samples': 25474048, 'steps': 49753, 'loss/train': 1.6552835702896118}
+02/26/2022 18:01:09 - INFO - codeparrot_training - Step 49754: {'lr': 3.24032160090304e-08, 'samples': 25474560, 'steps': 49754, 'loss/train': 1.060267686843872}
+02/26/2022 18:01:15 - INFO - codeparrot_training - Step 49755: {'lr': 3.214031631060266e-08, 'samples': 25475072, 'steps': 49755, 'loss/train': 1.078352689743042}
+02/26/2022 18:01:18 - INFO - codeparrot_training - Step 49756: {'lr': 3.187848739510946e-08, 'samples': 25475584, 'steps': 49756, 'loss/train': 1.0409449338912964}
+02/26/2022 18:01:24 - INFO - codeparrot_training - Step 49757: {'lr': 3.161772926366102e-08, 'samples': 25476096, 'steps': 49757, 'loss/train': 1.6763296127319336}
+02/26/2022 18:01:27 - INFO - codeparrot_training - Step 49758: {'lr': 3.135804191739533e-08, 'samples': 25476608, 'steps': 49758, 'loss/train': 2.109692096710205}
+02/26/2022 18:01:34 - INFO - codeparrot_training - Step 49759: {'lr': 3.1099425357422604e-08, 'samples': 25477120, 'steps': 49759, 'loss/train': 0.2643895447254181}
+02/26/2022 18:01:37 - INFO - codeparrot_training - Step 49760: {'lr': 3.084187958485307e-08, 'samples': 25477632, 'steps': 49760, 'loss/train': 2.084852695465088}
+02/26/2022 18:01:43 - INFO - codeparrot_training - Step 49761: {'lr': 3.05854046007692e-08, 'samples': 25478144, 'steps': 49761, 'loss/train': 2.0043864250183105}
+02/26/2022 18:01:47 - INFO - codeparrot_training - Step 49762: {'lr': 3.03300004062812e-08, 'samples': 25478656, 'steps': 49762, 'loss/train': 1.6485490798950195}
+02/26/2022 18:01:52 - INFO - codeparrot_training - Step 49763: {'lr': 3.007566700249931e-08, 'samples': 25479168, 'steps': 49763, 'loss/train': 1.4275230169296265}
+02/26/2022 18:01:56 - INFO - codeparrot_training - Step 49764: {'lr': 2.982240439047823e-08, 'samples': 25479680, 'steps': 49764, 'loss/train': 1.6674201488494873}
+02/26/2022 18:02:01 - INFO - codeparrot_training - Step 49765: {'lr': 2.9570212571355947e-08, 'samples': 25480192, 'steps': 49765, 'loss/train': 1.0222281217575073}
+02/26/2022 18:02:05 - INFO - codeparrot_training - Step 49766: {'lr': 2.9319091546159415e-08, 'samples': 25480704, 'steps': 49766, 'loss/train': 1.992708444595337}
+02/26/2022 18:02:11 - INFO - codeparrot_training - Step 49767: {'lr': 2.90690413159711e-08, 'samples': 25481216, 'steps': 49767, 'loss/train': 0.5209033489227295}
+02/26/2022 18:02:14 - INFO - codeparrot_training - Step 49768: {'lr': 2.8820061881901226e-08, 'samples': 25481728, 'steps': 49768, 'loss/train': 1.1297340393066406}
+02/26/2022 18:02:19 - INFO - codeparrot_training - Step 49769: {'lr': 2.857215324497675e-08, 'samples': 25482240, 'steps': 49769, 'loss/train': 0.6178969144821167}
+02/26/2022 18:02:23 - INFO - codeparrot_training - Step 49770: {'lr': 2.832531540628014e-08, 'samples': 25482752, 'steps': 49770, 'loss/train': 1.3847692012786865}
+02/26/2022 18:02:30 - INFO - codeparrot_training - Step 49771: {'lr': 2.8079548366866105e-08, 'samples': 25483264, 'steps': 49771, 'loss/train': 2.3196327686309814}
+02/26/2022 18:02:33 - INFO - codeparrot_training - Step 49772: {'lr': 2.7834852127789356e-08, 'samples': 25483776, 'steps': 49772, 'loss/train': 2.1752512454986572}
+02/26/2022 18:02:39 - INFO - codeparrot_training - Step 49773: {'lr': 2.7591226690076853e-08, 'samples': 25484288, 'steps': 49773, 'loss/train': 2.07833194732666}
+02/26/2022 18:02:42 - INFO - codeparrot_training - Step 49774: {'lr': 2.734867205478331e-08, 'samples': 25484800, 'steps': 49774, 'loss/train': 1.5006252527236938}
+02/26/2022 18:02:48 - INFO - codeparrot_training - Step 49775: {'lr': 2.7107188222991187e-08, 'samples': 25485312, 'steps': 49775, 'loss/train': 1.5359556674957275}
+02/26/2022 18:02:51 - INFO - codeparrot_training - Step 49776: {'lr': 2.686677519564418e-08, 'samples': 25485824, 'steps': 49776, 'loss/train': 1.9262195825576782}
+02/26/2022 18:02:57 - INFO - codeparrot_training - Step 49777: {'lr': 2.6627432973852506e-08, 'samples': 25486336, 'steps': 49777, 'loss/train': 0.4731055796146393}
+02/26/2022 18:03:00 - INFO - codeparrot_training - Step 49778: {'lr': 2.638916155861537e-08, 'samples': 25486848, 'steps': 49778, 'loss/train': 2.6146676540374756}
+02/26/2022 18:03:06 - INFO - codeparrot_training - Step 49779: {'lr': 2.615196095095973e-08, 'samples': 25487360, 'steps': 49779, 'loss/train': 2.111182928085327}
+02/26/2022 18:03:09 - INFO - codeparrot_training - Step 49780: {'lr': 2.5915831151884784e-08, 'samples': 25487872, 'steps': 49780, 'loss/train': 1.5207401514053345}
+02/26/2022 18:03:15 - INFO - codeparrot_training - Step 49781: {'lr': 2.5680772162389732e-08, 'samples': 25488384, 'steps': 49781, 'loss/train': 2.062026262283325}
+02/26/2022 18:03:18 - INFO - codeparrot_training - Step 49782: {'lr': 2.5446783983529286e-08, 'samples': 25488896, 'steps': 49782, 'loss/train': 1.3513418436050415}
+02/26/2022 18:03:24 - INFO - codeparrot_training - Step 49783: {'lr': 2.5213866616274895e-08, 'samples': 25489408, 'steps': 49783, 'loss/train': 1.4412165880203247}
+02/26/2022 18:03:27 - INFO - codeparrot_training - Step 49784: {'lr': 2.4982020061625754e-08, 'samples': 25489920, 'steps': 49784, 'loss/train': 0.5724613070487976}
+02/26/2022 18:03:33 - INFO - codeparrot_training - Step 49785: {'lr': 2.475124432058107e-08, 'samples': 25490432, 'steps': 49785, 'loss/train': 1.6561940908432007}
+02/26/2022 18:03:37 - INFO - codeparrot_training - Step 49786: {'lr': 2.452153939411228e-08, 'samples': 25490944, 'steps': 49786, 'loss/train': 2.0832836627960205}
+02/26/2022 18:03:42 - INFO - codeparrot_training - Step 49787: {'lr': 2.4292905283218592e-08, 'samples': 25491456, 'steps': 49787, 'loss/train': 1.4292668104171753}
+02/26/2022 18:03:46 - INFO - codeparrot_training - Step 49788: {'lr': 2.4065341988899204e-08, 'samples': 25491968, 'steps': 49788, 'loss/train': 1.3877793550491333}
+02/26/2022 18:03:51 - INFO - codeparrot_training - Step 49789: {'lr': 2.383884951207005e-08, 'samples': 25492480, 'steps': 49789, 'loss/train': 0.8354439735412598}
+02/26/2022 18:03:55 - INFO - codeparrot_training - Step 49790: {'lr': 2.3613427853758086e-08, 'samples': 25492992, 'steps': 49790, 'loss/train': 1.7699052095413208}
+02/26/2022 18:04:00 - INFO - codeparrot_training - Step 49791: {'lr': 2.3389077014934756e-08, 'samples': 25493504, 'steps': 49791, 'loss/train': 0.6568465232849121}
+02/26/2022 18:04:04 - INFO - codeparrot_training - Step 49792: {'lr': 2.3165796996515997e-08, 'samples': 25494016, 'steps': 49792, 'loss/train': 1.5680688619613647}
+02/26/2022 18:04:09 - INFO - codeparrot_training - Step 49793: {'lr': 2.2943587799473253e-08, 'samples': 25494528, 'steps': 49793, 'loss/train': 1.324320912361145}
+02/26/2022 18:04:13 - INFO - codeparrot_training - Step 49794: {'lr': 2.2722449424750214e-08, 'samples': 25495040, 'steps': 49794, 'loss/train': 2.320014238357544}
+02/26/2022 18:04:18 - INFO - codeparrot_training - Step 49795: {'lr': 2.250238187334608e-08, 'samples': 25495552, 'steps': 49795, 'loss/train': 2.499485969543457}
+02/26/2022 18:04:22 - INFO - codeparrot_training - Step 49796: {'lr': 2.2283385146121272e-08, 'samples': 25496064, 'steps': 49796, 'loss/train': 2.381624698638916}
+02/26/2022 18:04:28 - INFO - codeparrot_training - Step 49797: {'lr': 2.2065459244074992e-08, 'samples': 25496576, 'steps': 49797, 'loss/train': 2.4942405223846436}
+02/26/2022 18:04:31 - INFO - codeparrot_training - Step 49798: {'lr': 2.184860416815093e-08, 'samples': 25497088, 'steps': 49798, 'loss/train': 0.09525682032108307}
+02/26/2022 18:04:37 - INFO - codeparrot_training - Step 49799: {'lr': 2.1632819919209513e-08, 'samples': 25497600, 'steps': 49799, 'loss/train': 2.514915943145752}
+02/26/2022 18:04:40 - INFO - codeparrot_training - Step 49800: {'lr': 2.1418106498249933e-08, 'samples': 25498112, 'steps': 49800, 'loss/train': 1.5702953338623047}
+02/26/2022 18:04:46 - INFO - codeparrot_training - Step 49801: {'lr': 2.120446390613262e-08, 'samples': 25498624, 'steps': 49801, 'loss/train': 1.5474028587341309}
+02/26/2022 18:04:49 - INFO - codeparrot_training - Step 49802: {'lr': 2.0991892143801262e-08, 'samples': 25499136, 'steps': 49802, 'loss/train': 2.387803792953491}
+02/26/2022 18:04:55 - INFO - codeparrot_training - Step 49803: {'lr': 2.078039121217179e-08, 'samples': 25499648, 'steps': 49803, 'loss/train': 0.5508601665496826}
+02/26/2022 18:04:58 - INFO - codeparrot_training - Step 49804: {'lr': 2.0569961112132385e-08, 'samples': 25500160, 'steps': 49804, 'loss/train': 2.0543103218078613}
+02/26/2022 18:05:04 - INFO - codeparrot_training - Step 49805: {'lr': 2.0360601844626736e-08, 'samples': 25500672, 'steps': 49805, 'loss/train': 2.405283212661743}
+02/26/2022 18:05:07 - INFO - codeparrot_training - Step 49806: {'lr': 2.015231341048751e-08, 'samples': 25501184, 'steps': 49806, 'loss/train': 2.0754432678222656}
+02/26/2022 18:05:13 - INFO - codeparrot_training - Step 49807: {'lr': 1.9945095810630642e-08, 'samples': 25501696, 'steps': 49807, 'loss/train': 0.9231384992599487}
+02/26/2022 18:05:17 - INFO - codeparrot_training - Step 49808: {'lr': 1.9738949045972064e-08, 'samples': 25502208, 'steps': 49808, 'loss/train': 1.648268699645996}
+02/26/2022 18:05:22 - INFO - codeparrot_training - Step 49809: {'lr': 1.95338731173722e-08, 'samples': 25502720, 'steps': 49809, 'loss/train': 1.5111702680587769}
+02/26/2022 18:05:26 - INFO - codeparrot_training - Step 49810: {'lr': 1.9329868025691478e-08, 'samples': 25503232, 'steps': 49810, 'loss/train': 1.0084084272384644}
+02/26/2022 18:05:31 - INFO - codeparrot_training - Step 49811: {'lr': 1.912693377184582e-08, 'samples': 25503744, 'steps': 49811, 'loss/train': 1.7887688875198364}
+02/26/2022 18:05:35 - INFO - codeparrot_training - Step 49812: {'lr': 1.8925070356667907e-08, 'samples': 25504256, 'steps': 49812, 'loss/train': 1.0794496536254883}
+02/26/2022 18:05:40 - INFO - codeparrot_training - Step 49813: {'lr': 1.8724277781045906e-08, 'samples': 25504768, 'steps': 49813, 'loss/train': 0.8234782814979553}
+02/26/2022 18:05:44 - INFO - codeparrot_training - Step 49814: {'lr': 1.8524556045840245e-08, 'samples': 25505280, 'steps': 49814, 'loss/train': 1.2747784852981567}
+02/26/2022 18:05:49 - INFO - codeparrot_training - Step 49815: {'lr': 1.8325905151883593e-08, 'samples': 25505792, 'steps': 49815, 'loss/train': 2.13350510597229}
+02/26/2022 18:05:53 - INFO - codeparrot_training - Step 49816: {'lr': 1.8128325100036367e-08, 'samples': 25506304, 'steps': 49816, 'loss/train': 2.030944585800171}
+02/26/2022 18:05:59 - INFO - codeparrot_training - Step 49817: {'lr': 1.7931815891158998e-08, 'samples': 25506816, 'steps': 49817, 'loss/train': 4.86380672454834}
+02/26/2022 18:06:02 - INFO - codeparrot_training - Step 49818: {'lr': 1.7736377526084145e-08, 'samples': 25507328, 'steps': 49818, 'loss/train': 2.0606706142425537}
+02/26/2022 18:06:08 - INFO - codeparrot_training - Step 49819: {'lr': 1.754201000564448e-08, 'samples': 25507840, 'steps': 49819, 'loss/train': 2.158308744430542}
+02/26/2022 18:06:12 - INFO - codeparrot_training - Step 49820: {'lr': 1.7348713330672673e-08, 'samples': 25508352, 'steps': 49820, 'loss/train': 1.022890329360962}
+02/26/2022 18:06:17 - INFO - codeparrot_training - Step 49821: {'lr': 1.7156487502001383e-08, 'samples': 25508864, 'steps': 49821, 'loss/train': 1.3499585390090942}
+02/26/2022 18:06:21 - INFO - codeparrot_training - Step 49822: {'lr': 1.6965332520463282e-08, 'samples': 25509376, 'steps': 49822, 'loss/train': 1.7400176525115967}
+02/26/2022 18:06:26 - INFO - codeparrot_training - Step 49823: {'lr': 1.6775248386863283e-08, 'samples': 25509888, 'steps': 49823, 'loss/train': 1.4092994928359985}
+02/26/2022 18:06:30 - INFO - codeparrot_training - Step 49824: {'lr': 1.6586235102006298e-08, 'samples': 25510400, 'steps': 49824, 'loss/train': 2.208158254623413}
+02/26/2022 18:06:35 - INFO - codeparrot_training - Step 49825: {'lr': 1.639829266672499e-08, 'samples': 25510912, 'steps': 49825, 'loss/train': 1.0031169652938843}
+02/26/2022 18:06:39 - INFO - codeparrot_training - Step 49826: {'lr': 1.6211421081824273e-08, 'samples': 25511424, 'steps': 49826, 'loss/train': 1.2313945293426514}
+02/26/2022 18:06:44 - INFO - codeparrot_training - Step 49827: {'lr': 1.6025620348081303e-08, 'samples': 25511936, 'steps': 49827, 'loss/train': 2.046518087387085}
+02/26/2022 18:06:48 - INFO - codeparrot_training - Step 49828: {'lr': 1.5840890466300995e-08, 'samples': 25512448, 'steps': 49828, 'loss/train': 1.7530125379562378}
+02/26/2022 18:06:53 - INFO - codeparrot_training - Step 49829: {'lr': 1.5657231437288257e-08, 'samples': 25512960, 'steps': 49829, 'loss/train': 1.495530128479004}
+02/26/2022 18:06:57 - INFO - codeparrot_training - Step 49830: {'lr': 1.5474643261820243e-08, 'samples': 25513472, 'steps': 49830, 'loss/train': 1.5152678489685059}
+02/26/2022 18:07:02 - INFO - codeparrot_training - Step 49831: {'lr': 1.5293125940701868e-08, 'samples': 25513984, 'steps': 49831, 'loss/train': 0.8266059160232544}
+02/26/2022 18:07:06 - INFO - codeparrot_training - Step 49832: {'lr': 1.511267947465478e-08, 'samples': 25514496, 'steps': 49832, 'loss/train': 1.1114251613616943}
+02/26/2022 18:07:12 - INFO - codeparrot_training - Step 49833: {'lr': 1.4933303864511637e-08, 'samples': 25515008, 'steps': 49833, 'loss/train': 1.2117236852645874}
+02/26/2022 18:07:15 - INFO - codeparrot_training - Step 49834: {'lr': 1.4754999110994094e-08, 'samples': 25515520, 'steps': 49834, 'loss/train': 0.901160478591919}
+02/26/2022 18:07:21 - INFO - codeparrot_training - Step 49835: {'lr': 1.4577765214907056e-08, 'samples': 25516032, 'steps': 49835, 'loss/train': 1.5964235067367554}
+02/26/2022 18:07:24 - INFO - codeparrot_training - Step 49836: {'lr': 1.4401602176972173e-08, 'samples': 25516544, 'steps': 49836, 'loss/train': 1.9681991338729858}
+02/26/2022 18:07:30 - INFO - codeparrot_training - Step 49837: {'lr': 1.4226509997966596e-08, 'samples': 25517056, 'steps': 49837, 'loss/train': 1.6212447881698608}
+02/26/2022 18:07:33 - INFO - codeparrot_training - Step 49838: {'lr': 1.405248867863973e-08, 'samples': 25517568, 'steps': 49838, 'loss/train': 0.8212572336196899}
+02/26/2022 18:07:39 - INFO - codeparrot_training - Step 49839: {'lr': 1.3879538219713217e-08, 'samples': 25518080, 'steps': 49839, 'loss/train': 1.2232691049575806}
+02/26/2022 18:07:42 - INFO - codeparrot_training - Step 49840: {'lr': 1.3707658621964214e-08, 'samples': 25518592, 'steps': 49840, 'loss/train': 1.354844570159912}
+02/26/2022 18:07:48 - INFO - codeparrot_training - Step 49841: {'lr': 1.3536849886114366e-08, 'samples': 25519104, 'steps': 49841, 'loss/train': 1.6901588439941406}
+02/26/2022 18:07:52 - INFO - codeparrot_training - Step 49842: {'lr': 1.3367112012885319e-08, 'samples': 25519616, 'steps': 49842, 'loss/train': 8.419468879699707}
+02/26/2022 18:07:58 - INFO - codeparrot_training - Step 49843: {'lr': 1.3198445003026472e-08, 'samples': 25520128, 'steps': 49843, 'loss/train': 2.186472177505493}
+02/26/2022 18:08:01 - INFO - codeparrot_training - Step 49844: {'lr': 1.3030848857231714e-08, 'samples': 25520640, 'steps': 49844, 'loss/train': 1.437972068786621}
+02/26/2022 18:08:07 - INFO - codeparrot_training - Step 49845: {'lr': 1.2864323576222692e-08, 'samples': 25521152, 'steps': 49845, 'loss/train': 1.9912728071212769}
+02/26/2022 18:08:10 - INFO - codeparrot_training - Step 49846: {'lr': 1.2698869160748806e-08, 'samples': 25521664, 'steps': 49846, 'loss/train': 0.8067148327827454}
+02/26/2022 18:08:16 - INFO - codeparrot_training - Step 49847: {'lr': 1.2534485611476187e-08, 'samples': 25522176, 'steps': 49847, 'loss/train': 1.9356513023376465}
+02/26/2022 18:08:19 - INFO - codeparrot_training - Step 49848: {'lr': 1.2371172929126484e-08, 'samples': 25522688, 'steps': 49848, 'loss/train': 2.290548324584961}
+02/26/2022 18:08:25 - INFO - codeparrot_training - Step 49849: {'lr': 1.2208931114393584e-08, 'samples': 25523200, 'steps': 49849, 'loss/train': 1.7514867782592773}
+02/26/2022 18:08:28 - INFO - codeparrot_training - Step 49850: {'lr': 1.2047760167999133e-08, 'samples': 25523712, 'steps': 49850, 'loss/train': 1.7774622440338135}
+02/26/2022 18:08:34 - INFO - codeparrot_training - Step 49851: {'lr': 1.1887660090609264e-08, 'samples': 25524224, 'steps': 49851, 'loss/train': 1.676344394683838}
+02/26/2022 18:08:37 - INFO - codeparrot_training - Step 49852: {'lr': 1.1728630882890112e-08, 'samples': 25524736, 'steps': 49852, 'loss/train': 2.046475648880005}
+02/26/2022 18:08:44 - INFO - codeparrot_training - Step 49853: {'lr': 1.1570672545563321e-08, 'samples': 25525248, 'steps': 49853, 'loss/train': 1.997458815574646}
+02/26/2022 18:08:47 - INFO - codeparrot_training - Step 49854: {'lr': 1.1413785079267269e-08, 'samples': 25525760, 'steps': 49854, 'loss/train': 1.489466667175293}
+02/26/2022 18:08:53 - INFO - codeparrot_training - Step 49855: {'lr': 1.1257968484723602e-08, 'samples': 25526272, 'steps': 49855, 'loss/train': 2.143293857574463}
+02/26/2022 18:08:56 - INFO - codeparrot_training - Step 49856: {'lr': 1.1103222762542941e-08, 'samples': 25526784, 'steps': 49856, 'loss/train': 1.6462370157241821}
+02/26/2022 18:09:02 - INFO - codeparrot_training - Step 49857: {'lr': 1.0949547913446933e-08, 'samples': 25527296, 'steps': 49857, 'loss/train': 2.2388648986816406}
+02/26/2022 18:09:05 - INFO - codeparrot_training - Step 49858: {'lr': 1.0796943938018445e-08, 'samples': 25527808, 'steps': 49858, 'loss/train': 2.399623155593872}
+02/26/2022 18:09:11 - INFO - codeparrot_training - Step 49859: {'lr': 1.0645410837006875e-08, 'samples': 25528320, 'steps': 49859, 'loss/train': 1.6432126760482788}
+02/26/2022 18:09:14 - INFO - codeparrot_training - Step 49860: {'lr': 1.0494948610967337e-08, 'samples': 25528832, 'steps': 49860, 'loss/train': 1.7454429864883423}
+02/26/2022 18:09:20 - INFO - codeparrot_training - Step 49861: {'lr': 1.034555726059372e-08, 'samples': 25529344, 'steps': 49861, 'loss/train': 2.9228708744049072}
+02/26/2022 18:09:23 - INFO - codeparrot_training - Step 49862: {'lr': 1.0197236786552155e-08, 'samples': 25529856, 'steps': 49862, 'loss/train': 1.7823240756988525}
+02/26/2022 18:09:30 - INFO - codeparrot_training - Step 49863: {'lr': 1.0049987189397758e-08, 'samples': 25530368, 'steps': 49863, 'loss/train': 1.155539870262146}
+02/26/2022 18:09:33 - INFO - codeparrot_training - Step 49864: {'lr': 9.90380846985217e-09, 'samples': 25530880, 'steps': 49864, 'loss/train': 1.9393876791000366}
+02/26/2022 18:09:37 - INFO - codeparrot_training - Step 49865: {'lr': 9.758700628470507e-09, 'samples': 25531392, 'steps': 49865, 'loss/train': 0.8131659030914307}
+02/26/2022 18:09:42 - INFO - codeparrot_training - Step 49866: {'lr': 9.614663665918899e-09, 'samples': 25531904, 'steps': 49866, 'loss/train': 1.6611566543579102}
+02/26/2022 18:09:46 - INFO - codeparrot_training - Step 49867: {'lr': 9.471697582780214e-09, 'samples': 25532416, 'steps': 49867, 'loss/train': 2.2808587551116943}
+02/26/2022 18:09:51 - INFO - codeparrot_training - Step 49868: {'lr': 9.329802379692831e-09, 'samples': 25532928, 'steps': 49868, 'loss/train': 2.0008318424224854}
+02/26/2022 18:09:55 - INFO - codeparrot_training - Step 49869: {'lr': 9.188978057239616e-09, 'samples': 25533440, 'steps': 49869, 'loss/train': 1.3950893878936768}
+02/26/2022 18:10:00 - INFO - codeparrot_training - Step 49870: {'lr': 9.049224616058948e-09, 'samples': 25533952, 'steps': 49870, 'loss/train': 2.2258694171905518}
+02/26/2022 18:10:04 - INFO - codeparrot_training - Step 49871: {'lr': 8.910542056733695e-09, 'samples': 25534464, 'steps': 49871, 'loss/train': 1.170242190361023}
+02/26/2022 18:10:09 - INFO - codeparrot_training - Step 49872: {'lr': 8.772930379846722e-09, 'samples': 25534976, 'steps': 49872, 'loss/train': 2.095367193222046}
+02/26/2022 18:10:13 - INFO - codeparrot_training - Step 49873: {'lr': 8.636389585980897e-09, 'samples': 25535488, 'steps': 49873, 'loss/train': 0.40433958172798157}
+02/26/2022 18:10:19 - INFO - codeparrot_training - Step 49874: {'lr': 8.500919675746843e-09, 'samples': 25536000, 'steps': 49874, 'loss/train': 0.24236752092838287}
+02/26/2022 18:10:22 - INFO - codeparrot_training - Step 49875: {'lr': 8.366520649727427e-09, 'samples': 25536512, 'steps': 49875, 'loss/train': 2.23716402053833}
+02/26/2022 18:10:28 - INFO - codeparrot_training - Step 49876: {'lr': 8.233192508477761e-09, 'samples': 25537024, 'steps': 49876, 'loss/train': 1.3792365789413452}
+02/26/2022 18:10:31 - INFO - codeparrot_training - Step 49877: {'lr': 8.100935252552955e-09, 'samples': 25537536, 'steps': 49877, 'loss/train': 1.2408055067062378}
+02/26/2022 18:10:37 - INFO - codeparrot_training - Step 49878: {'lr': 7.969748882563632e-09, 'samples': 25538048, 'steps': 49878, 'loss/train': 1.9493943452835083}
+02/26/2022 18:10:40 - INFO - codeparrot_training - Step 49879: {'lr': 7.839633399064905e-09, 'samples': 25538560, 'steps': 49879, 'loss/train': 1.695171594619751}
+02/26/2022 18:10:46 - INFO - codeparrot_training - Step 49880: {'lr': 7.710588802584129e-09, 'samples': 25539072, 'steps': 49880, 'loss/train': 2.728811740875244}
+02/26/2022 18:10:49 - INFO - codeparrot_training - Step 49881: {'lr': 7.582615093676415e-09, 'samples': 25539584, 'steps': 49881, 'loss/train': 0.7754898071289062}
+02/26/2022 18:10:55 - INFO - codeparrot_training - Step 49882: {'lr': 7.455712272924632e-09, 'samples': 25540096, 'steps': 49882, 'loss/train': 1.793603539466858}
+02/26/2022 18:10:58 - INFO - codeparrot_training - Step 49883: {'lr': 7.329880340828377e-09, 'samples': 25540608, 'steps': 49883, 'loss/train': 1.2226747274398804}
+02/26/2022 18:11:04 - INFO - codeparrot_training - Step 49884: {'lr': 7.20511929797052e-09, 'samples': 25541120, 'steps': 49884, 'loss/train': 1.2711772918701172}
+02/26/2022 18:11:07 - INFO - codeparrot_training - Step 49885: {'lr': 7.081429144850659e-09, 'samples': 25541632, 'steps': 49885, 'loss/train': 1.431278109550476}
+02/26/2022 18:11:13 - INFO - codeparrot_training - Step 49886: {'lr': 6.958809882023909e-09, 'samples': 25542144, 'steps': 49886, 'loss/train': 2.0990211963653564}
+02/26/2022 18:11:16 - INFO - codeparrot_training - Step 49887: {'lr': 6.837261509989867e-09, 'samples': 25542656, 'steps': 49887, 'loss/train': 1.3828198909759521}
+02/26/2022 18:11:22 - INFO - codeparrot_training - Step 49888: {'lr': 6.716784029303646e-09, 'samples': 25543168, 'steps': 49888, 'loss/train': 2.3065266609191895}
+02/26/2022 18:11:25 - INFO - codeparrot_training - Step 49889: {'lr': 6.597377440437091e-09, 'samples': 25543680, 'steps': 49889, 'loss/train': 2.7996435165405273}
+02/26/2022 18:11:32 - INFO - codeparrot_training - Step 49890: {'lr': 6.479041743945313e-09, 'samples': 25544192, 'steps': 49890, 'loss/train': 0.8790472149848938}
+02/26/2022 18:11:36 - INFO - codeparrot_training - Step 49891: {'lr': 6.361776940300157e-09, 'samples': 25544704, 'steps': 49891, 'loss/train': 1.6156212091445923}
+02/26/2022 18:11:41 - INFO - codeparrot_training - Step 49892: {'lr': 6.2455830300289785e-09, 'samples': 25545216, 'steps': 49892, 'loss/train': 0.49404293298721313}
+02/26/2022 18:11:44 - INFO - codeparrot_training - Step 49893: {'lr': 6.1304600136036225e-09, 'samples': 25545728, 'steps': 49893, 'loss/train': 1.9654839038848877}
+02/26/2022 18:11:50 - INFO - codeparrot_training - Step 49894: {'lr': 6.016407891551445e-09, 'samples': 25546240, 'steps': 49894, 'loss/train': 1.9663312435150146}
+02/26/2022 18:11:54 - INFO - codeparrot_training - Step 49895: {'lr': 5.9034266643442915e-09, 'samples': 25546752, 'steps': 49895, 'loss/train': 1.3407528400421143}
+02/26/2022 18:11:59 - INFO - codeparrot_training - Step 49896: {'lr': 5.791516332454006e-09, 'samples': 25547264, 'steps': 49896, 'loss/train': 1.2587236166000366}
+02/26/2022 18:12:02 - INFO - codeparrot_training - Step 49897: {'lr': 5.680676896380188e-09, 'samples': 25547776, 'steps': 49897, 'loss/train': 2.067148208618164}
+02/26/2022 18:12:08 - INFO - codeparrot_training - Step 49898: {'lr': 5.570908356566929e-09, 'samples': 25548288, 'steps': 49898, 'loss/train': 1.4106074571609497}
+02/26/2022 18:12:11 - INFO - codeparrot_training - Step 49899: {'lr': 5.462210713513826e-09, 'samples': 25548800, 'steps': 49899, 'loss/train': 1.870240569114685}
+02/26/2022 18:12:18 - INFO - codeparrot_training - Step 49900: {'lr': 5.354583967692728e-09, 'samples': 25549312, 'steps': 49900, 'loss/train': 1.6069060564041138}
+02/26/2022 18:12:22 - INFO - codeparrot_training - Step 49901: {'lr': 5.248028119547721e-09, 'samples': 25549824, 'steps': 49901, 'loss/train': 1.6345579624176025}
+02/26/2022 18:12:27 - INFO - codeparrot_training - Step 49902: {'lr': 5.1425431695506504e-09, 'samples': 25550336, 'steps': 49902, 'loss/train': 1.5215963125228882}
+02/26/2022 18:12:30 - INFO - codeparrot_training - Step 49903: {'lr': 5.038129118117851e-09, 'samples': 25550848, 'steps': 49903, 'loss/train': 1.8097896575927734}
+02/26/2022 18:12:36 - INFO - codeparrot_training - Step 49904: {'lr': 4.934785965721167e-09, 'samples': 25551360, 'steps': 49904, 'loss/train': 3.924865961074829}
+02/26/2022 18:12:40 - INFO - codeparrot_training - Step 49905: {'lr': 4.832513712804687e-09, 'samples': 25551872, 'steps': 49905, 'loss/train': 2.7155182361602783}
+02/26/2022 18:12:45 - INFO - codeparrot_training - Step 49906: {'lr': 4.731312359812501e-09, 'samples': 25552384, 'steps': 49906, 'loss/train': 1.0408412218093872}
+02/26/2022 18:12:49 - INFO - codeparrot_training - Step 49907: {'lr': 4.631181907160942e-09, 'samples': 25552896, 'steps': 49907, 'loss/train': 1.5556855201721191}
+02/26/2022 18:12:54 - INFO - codeparrot_training - Step 49908: {'lr': 4.5321223552663436e-09, 'samples': 25553408, 'steps': 49908, 'loss/train': 2.84206223487854}
+02/26/2022 18:12:58 - INFO - codeparrot_training - Step 49909: {'lr': 4.434133704600552e-09, 'samples': 25553920, 'steps': 49909, 'loss/train': 2.1844868659973145}
+02/26/2022 18:13:04 - INFO - codeparrot_training - Step 49910: {'lr': 4.3372159555521426e-09, 'samples': 25554432, 'steps': 49910, 'loss/train': 2.244971513748169}
+02/26/2022 18:13:08 - INFO - codeparrot_training - Step 49911: {'lr': 4.241369108537452e-09, 'samples': 25554944, 'steps': 49911, 'loss/train': 1.7292194366455078}
+02/26/2022 18:13:13 - INFO - codeparrot_training - Step 49912: {'lr': 4.146593163972812e-09, 'samples': 25555456, 'steps': 49912, 'loss/train': 2.6721560955047607}
+02/26/2022 18:13:17 - INFO - codeparrot_training - Step 49913: {'lr': 4.052888122246801e-09, 'samples': 25555968, 'steps': 49913, 'loss/train': 0.9000661373138428}
+02/26/2022 18:13:22 - INFO - codeparrot_training - Step 49914: {'lr': 3.960253983803508e-09, 'samples': 25556480, 'steps': 49914, 'loss/train': 1.0378044843673706}
+02/26/2022 18:13:26 - INFO - codeparrot_training - Step 49915: {'lr': 3.868690749003756e-09, 'samples': 25556992, 'steps': 49915, 'loss/train': 1.2975900173187256}
+02/26/2022 18:13:32 - INFO - codeparrot_training - Step 49916: {'lr': 3.7781984182361226e-09, 'samples': 25557504, 'steps': 49916, 'loss/train': 1.4350013732910156}
+02/26/2022 18:13:35 - INFO - codeparrot_training - Step 49917: {'lr': 3.688776991889187e-09, 'samples': 25558016, 'steps': 49917, 'loss/train': 2.298407793045044}
+02/26/2022 18:13:41 - INFO - codeparrot_training - Step 49918: {'lr': 3.600426470379281e-09, 'samples': 25558528, 'steps': 49918, 'loss/train': 1.7967218160629272}
+02/26/2022 18:13:44 - INFO - codeparrot_training - Step 49919: {'lr': 3.5131468540672283e-09, 'samples': 25559040, 'steps': 49919, 'loss/train': 1.8604594469070435}
+02/26/2022 18:13:49 - INFO - codeparrot_training - Step 49920: {'lr': 3.4269381433138515e-09, 'samples': 25559552, 'steps': 49920, 'loss/train': 2.1783883571624756}
+02/26/2022 18:13:53 - INFO - codeparrot_training - Step 49921: {'lr': 3.341800338479972e-09, 'samples': 25560064, 'steps': 49921, 'loss/train': 0.9122430682182312}
+02/26/2022 18:14:00 - INFO - codeparrot_training - Step 49922: {'lr': 3.2577334399541693e-09, 'samples': 25560576, 'steps': 49922, 'loss/train': 0.28049030900001526}
+02/26/2022 18:14:03 - INFO - codeparrot_training - Step 49923: {'lr': 3.1747374480972647e-09, 'samples': 25561088, 'steps': 49923, 'loss/train': 0.6618022322654724}
+02/26/2022 18:14:09 - INFO - codeparrot_training - Step 49924: {'lr': 3.092812363270081e-09, 'samples': 25561600, 'steps': 49924, 'loss/train': 1.9529082775115967}
+02/26/2022 18:14:12 - INFO - codeparrot_training - Step 49925: {'lr': 3.01195818577793e-09, 'samples': 25562112, 'steps': 49925, 'loss/train': 1.6398659944534302}
+02/26/2022 18:14:18 - INFO - codeparrot_training - Step 49926: {'lr': 2.932174916009389e-09, 'samples': 25562624, 'steps': 49926, 'loss/train': 1.4235341548919678}
+02/26/2022 18:14:21 - INFO - codeparrot_training - Step 49927: {'lr': 2.8534625543252813e-09, 'samples': 25563136, 'steps': 49927, 'loss/train': 1.7917429208755493}
+02/26/2022 18:14:27 - INFO - codeparrot_training - Step 49928: {'lr': 2.775821101003162e-09, 'samples': 25563648, 'steps': 49928, 'loss/train': 2.244459629058838}
+02/26/2022 18:14:30 - INFO - codeparrot_training - Step 49929: {'lr': 2.699250556403854e-09, 'samples': 25564160, 'steps': 49929, 'loss/train': 1.7520989179611206}
+02/26/2022 18:14:36 - INFO - codeparrot_training - Step 49930: {'lr': 2.6237509208881795e-09, 'samples': 25564672, 'steps': 49930, 'loss/train': 1.5223432779312134}
+02/26/2022 18:14:39 - INFO - codeparrot_training - Step 49931: {'lr': 2.549322194733694e-09, 'samples': 25565184, 'steps': 49931, 'loss/train': 1.947978138923645}
+02/26/2022 18:14:45 - INFO - codeparrot_training - Step 49932: {'lr': 2.4759643782734655e-09, 'samples': 25565696, 'steps': 49932, 'loss/train': 0.6739815473556519}
+02/26/2022 18:14:48 - INFO - codeparrot_training - Step 49933: {'lr': 2.403677471812804e-09, 'samples': 25566208, 'steps': 49933, 'loss/train': 2.4989287853240967}
+02/26/2022 18:14:54 - INFO - codeparrot_training - Step 49934: {'lr': 2.3324614756847774e-09, 'samples': 25566720, 'steps': 49934, 'loss/train': 1.0190621614456177}
+02/26/2022 18:14:57 - INFO - codeparrot_training - Step 49935: {'lr': 2.2623163901669407e-09, 'samples': 25567232, 'steps': 49935, 'loss/train': 0.25735238194465637}
+02/26/2022 18:15:04 - INFO - codeparrot_training - Step 49936: {'lr': 2.1932422155923616e-09, 'samples': 25567744, 'steps': 49936, 'loss/train': 0.910692036151886}
+02/26/2022 18:15:07 - INFO - codeparrot_training - Step 49937: {'lr': 2.1252389522108396e-09, 'samples': 25568256, 'steps': 49937, 'loss/train': 1.597078800201416}
+02/26/2022 18:15:13 - INFO - codeparrot_training - Step 49938: {'lr': 2.0583066003831975e-09, 'samples': 25568768, 'steps': 49938, 'loss/train': 2.289205312728882}
+02/26/2022 18:15:16 - INFO - codeparrot_training - Step 49939: {'lr': 1.99244516033148e-09, 'samples': 25569280, 'steps': 49939, 'loss/train': 0.7344565987586975}
+02/26/2022 18:15:20 - INFO - codeparrot_training - Step 49940: {'lr': 1.927654632360998e-09, 'samples': 25569792, 'steps': 49940, 'loss/train': 0.9511494040489197}
+02/26/2022 18:15:25 - INFO - codeparrot_training - Step 49941: {'lr': 1.8639350167493074e-09, 'samples': 25570304, 'steps': 49941, 'loss/train': 1.2779861688613892}
+02/26/2022 18:15:29 - INFO - codeparrot_training - Step 49942: {'lr': 1.8012863137462088e-09, 'samples': 25570816, 'steps': 49942, 'loss/train': 1.6683646440505981}
+02/26/2022 18:15:34 - INFO - codeparrot_training - Step 49943: {'lr': 1.7397085236847687e-09, 'samples': 25571328, 'steps': 49943, 'loss/train': 1.4545079469680786}
+02/26/2022 18:15:38 - INFO - codeparrot_training - Step 49944: {'lr': 1.6792016467592764e-09, 'samples': 25571840, 'steps': 49944, 'loss/train': 1.3636372089385986}
+02/26/2022 18:15:43 - INFO - codeparrot_training - Step 49945: {'lr': 1.6197656832750428e-09, 'samples': 25572352, 'steps': 49945, 'loss/train': 1.4487106800079346}
+02/26/2022 18:15:47 - INFO - codeparrot_training - Step 49946: {'lr': 1.561400633454113e-09, 'samples': 25572864, 'steps': 49946, 'loss/train': 2.730422019958496}
+02/26/2022 18:15:53 - INFO - codeparrot_training - Step 49947: {'lr': 1.5041064975740426e-09, 'samples': 25573376, 'steps': 49947, 'loss/train': 1.2567530870437622}
+02/26/2022 18:15:56 - INFO - codeparrot_training - Step 49948: {'lr': 1.4478832758568762e-09, 'samples': 25573888, 'steps': 49948, 'loss/train': 0.8132756948471069}
+02/26/2022 18:16:02 - INFO - codeparrot_training - Step 49949: {'lr': 1.3927309685524136e-09, 'samples': 25574400, 'steps': 49949, 'loss/train': 1.624342679977417}
+02/26/2022 18:16:06 - INFO - codeparrot_training - Step 49950: {'lr': 1.3386495758827e-09, 'samples': 25574912, 'steps': 49950, 'loss/train': 1.123450756072998}
+02/26/2022 18:16:11 - INFO - codeparrot_training - Step 49951: {'lr': 1.2856390981252907e-09, 'samples': 25575424, 'steps': 49951, 'loss/train': 5.038472652435303}
+02/26/2022 18:16:15 - INFO - codeparrot_training - Step 49952: {'lr': 1.2336995354467196e-09, 'samples': 25575936, 'steps': 49952, 'loss/train': 1.3611881732940674}
+02/26/2022 18:16:20 - INFO - codeparrot_training - Step 49953: {'lr': 1.182830888124542e-09, 'samples': 25576448, 'steps': 49953, 'loss/train': 1.7058833837509155}
+02/26/2022 18:16:24 - INFO - codeparrot_training - Step 49954: {'lr': 1.133033156353047e-09, 'samples': 25576960, 'steps': 49954, 'loss/train': 1.687996745109558}
+02/26/2022 18:16:29 - INFO - codeparrot_training - Step 49955: {'lr': 1.0843063403265242e-09, 'samples': 25577472, 'steps': 49955, 'loss/train': 2.9062206745147705}
+02/26/2022 18:16:33 - INFO - codeparrot_training - Step 49956: {'lr': 1.0366504402947729e-09, 'samples': 25577984, 'steps': 49956, 'loss/train': 0.9279352426528931}
+02/26/2022 18:16:39 - INFO - codeparrot_training - Step 49957: {'lr': 9.900654563965716e-10, 'samples': 25578496, 'steps': 49957, 'loss/train': 0.7256073355674744}
+02/26/2022 18:16:43 - INFO - codeparrot_training - Step 49958: {'lr': 9.445513889094759e-10, 'samples': 25579008, 'steps': 49958, 'loss/train': 0.22750012576580048}
+02/26/2022 18:16:49 - INFO - codeparrot_training - Step 49959: {'lr': 9.001082379722636e-10, 'samples': 25579520, 'steps': 49959, 'loss/train': 1.4443467855453491}
+02/26/2022 18:16:52 - INFO - codeparrot_training - Step 49960: {'lr': 8.567360038069794e-10, 'samples': 25580032, 'steps': 49960, 'loss/train': 1.0324279069900513}
+02/26/2022 18:16:56 - INFO - codeparrot_training - Step 49961: {'lr': 8.144346866079122e-10, 'samples': 25580544, 'steps': 49961, 'loss/train': 1.933714509010315}
+02/26/2022 18:17:01 - INFO - codeparrot_training - Step 49962: {'lr': 7.7320428651384e-10, 'samples': 25581056, 'steps': 49962, 'loss/train': 2.138439178466797}
+02/26/2022 18:17:05 - INFO - codeparrot_training - Step 49963: {'lr': 7.330448037190518e-10, 'samples': 25581568, 'steps': 49963, 'loss/train': 2.0313007831573486}
+02/26/2022 18:17:11 - INFO - codeparrot_training - Step 49964: {'lr': 6.939562383900811e-10, 'samples': 25582080, 'steps': 49964, 'loss/train': 1.2096962928771973}
+02/26/2022 18:17:14 - INFO - codeparrot_training - Step 49965: {'lr': 6.559385907212167e-10, 'samples': 25582592, 'steps': 49965, 'loss/train': 0.6134408712387085}
+02/26/2022 18:17:18 - INFO - codeparrot_training - Step 49966: {'lr': 6.189918608789924e-10, 'samples': 25583104, 'steps': 49966, 'loss/train': 1.0768870115280151}
+02/26/2022 18:17:23 - INFO - codeparrot_training - Step 49967: {'lr': 5.831160489744303e-10, 'samples': 25583616, 'steps': 49967, 'loss/train': 2.3901000022888184}
+02/26/2022 18:17:27 - INFO - codeparrot_training - Step 49968: {'lr': 5.483111551740638e-10, 'samples': 25584128, 'steps': 49968, 'loss/train': 1.9718189239501953}
+02/26/2022 18:17:32 - INFO - codeparrot_training - Step 49969: {'lr': 5.145771796721821e-10, 'samples': 25584640, 'steps': 49969, 'loss/train': 1.3913748264312744}
+02/26/2022 18:17:36 - INFO - codeparrot_training - Step 49970: {'lr': 4.819141225798074e-10, 'samples': 25585152, 'steps': 49970, 'loss/train': 1.8307862281799316}
+02/26/2022 18:17:42 - INFO - codeparrot_training - Step 49971: {'lr': 4.50321984007962e-10, 'samples': 25585664, 'steps': 49971, 'loss/train': 1.5337656736373901}
+02/26/2022 18:17:45 - INFO - codeparrot_training - Step 49972: {'lr': 4.198007641509349e-10, 'samples': 25586176, 'steps': 49972, 'loss/train': 2.0893256664276123}
+02/26/2022 18:17:51 - INFO - codeparrot_training - Step 49973: {'lr': 3.903504631197485e-10, 'samples': 25586688, 'steps': 49973, 'loss/train': 1.5156900882720947}
+02/26/2022 18:17:55 - INFO - codeparrot_training - Step 49974: {'lr': 3.619710809976695e-10, 'samples': 25587200, 'steps': 49974, 'loss/train': 1.4385193586349487}
+02/26/2022 18:18:00 - INFO - codeparrot_training - Step 49975: {'lr': 3.346626179789869e-10, 'samples': 25587712, 'steps': 49975, 'loss/train': 1.695233941078186}
+02/26/2022 18:18:04 - INFO - codeparrot_training - Step 49976: {'lr': 3.0842507411921185e-10, 'samples': 25588224, 'steps': 49976, 'loss/train': 1.9455724954605103}
+02/26/2022 18:18:09 - INFO - codeparrot_training - Step 49977: {'lr': 2.832584495571222e-10, 'samples': 25588736, 'steps': 49977, 'loss/train': 1.5164356231689453}
+02/26/2022 18:18:12 - INFO - codeparrot_training - Step 49978: {'lr': 2.5916274440374034e-10, 'samples': 25589248, 'steps': 49978, 'loss/train': 2.4412038326263428}
+02/26/2022 18:18:18 - INFO - codeparrot_training - Step 49979: {'lr': 2.361379587423329e-10, 'samples': 25589760, 'steps': 49979, 'loss/train': 0.5632081031799316}
+02/26/2022 18:18:21 - INFO - codeparrot_training - Step 49980: {'lr': 2.1418409271167783e-10, 'samples': 25590272, 'steps': 49980, 'loss/train': 2.2746522426605225}
+02/26/2022 18:18:27 - INFO - codeparrot_training - Step 49981: {'lr': 1.9330114636728623e-10, 'samples': 25590784, 'steps': 49981, 'loss/train': 2.490490198135376}
+02/26/2022 18:18:30 - INFO - codeparrot_training - Step 49982: {'lr': 1.734891197924249e-10, 'samples': 25591296, 'steps': 49982, 'loss/train': 1.5169414281845093}
+02/26/2022 18:18:36 - INFO - codeparrot_training - Step 49983: {'lr': 1.5474801309811605e-10, 'samples': 25591808, 'steps': 49983, 'loss/train': 1.9936089515686035}
+02/26/2022 18:18:39 - INFO - codeparrot_training - Step 49984: {'lr': 1.3707782636762644e-10, 'samples': 25592320, 'steps': 49984, 'loss/train': 1.350411057472229}
+02/26/2022 18:18:45 - INFO - codeparrot_training - Step 49985: {'lr': 1.2047855968422282e-10, 'samples': 25592832, 'steps': 49985, 'loss/train': 1.2329177856445312}
+02/26/2022 18:18:48 - INFO - codeparrot_training - Step 49986: {'lr': 1.0495021307566077e-10, 'samples': 25593344, 'steps': 49986, 'loss/train': 0.12649674713611603}
+02/26/2022 18:18:54 - INFO - codeparrot_training - Step 49987: {'lr': 9.049278662520699e-11, 'samples': 25593856, 'steps': 49987, 'loss/train': 1.775930643081665}
+02/26/2022 18:18:58 - INFO - codeparrot_training - Step 49988: {'lr': 7.710628041612822e-11, 'samples': 25594368, 'steps': 49988, 'loss/train': 1.977392554283142}
+02/26/2022 18:19:03 - INFO - codeparrot_training - Step 49989: {'lr': 6.47906945039356e-11, 'samples': 25594880, 'steps': 49989, 'loss/train': 2.5542259216308594}
+02/26/2022 18:19:07 - INFO - codeparrot_training - Step 49990: {'lr': 5.354602891638471e-11, 'samples': 25595392, 'steps': 49990, 'loss/train': 0.2777775526046753}
+02/26/2022 18:19:12 - INFO - codeparrot_training - Step 49991: {'lr': 4.3372283708986714e-11, 'samples': 25595904, 'steps': 49991, 'loss/train': 1.014872431755066}
+02/26/2022 18:19:16 - INFO - codeparrot_training - Step 49992: {'lr': 3.4269458937252754e-11, 'samples': 25596416, 'steps': 49992, 'loss/train': 1.938048005104065}
+02/26/2022 18:19:21 - INFO - codeparrot_training - Step 49993: {'lr': 2.6237554656693975e-11, 'samples': 25596928, 'steps': 49993, 'loss/train': 0.7541128993034363}
+02/26/2022 18:19:25 - INFO - codeparrot_training - Step 49994: {'lr': 1.927657083955481e-11, 'samples': 25597440, 'steps': 49994, 'loss/train': 2.6811628341674805}
+02/26/2022 18:19:31 - INFO - codeparrot_training - Step 49995: {'lr': 1.3386507596857556e-11, 'samples': 25597952, 'steps': 49995, 'loss/train': 1.5248055458068848}
+02/26/2022 18:19:34 - INFO - codeparrot_training - Step 49996: {'lr': 8.567364873091067e-12, 'samples': 25598464, 'steps': 49996, 'loss/train': 1.6005619764328003}
+02/26/2022 18:19:40 - INFO - codeparrot_training - Step 49997: {'lr': 4.819142751522065e-12, 'samples': 25598976, 'steps': 49997, 'loss/train': 1.537976622581482}
+02/26/2022 18:19:44 - INFO - codeparrot_training - Step 49998: {'lr': 2.1418412321505543e-12, 'samples': 25599488, 'steps': 49998, 'loss/train': 2.184359073638916}
+02/26/2022 18:19:49 - INFO - codeparrot_training - Step 49999: {'lr': 5.354603149765325e-13, 'samples': 25600000, 'steps': 49999, 'loss/train': 2.408355236053467}
+02/26/2022 18:19:49 - INFO - codeparrot_training - Evaluating and saving model checkpoint