diff --git "a/log/debug_0.log" "b/log/debug_0.log"
--- "a/log/debug_0.log"
+++ "b/log/debug_0.log"
@@ -44187,3 +44187,2511 @@ Use FP16 precision: False
 01/23/2022 08:08:38 - INFO - codeparrot_training - Evaluating and saving model checkpoint
 01/23/2022 08:10:13 - INFO - codeparrot_training - Evaluation loss: 3.0351104736328125 perplexity: 20.80327606201172
 01/23/2022 08:10:13 - INFO - codeparrot_training - Step 37500: {'loss/eval': 3.0351104736328125, 'perplexity': 20.80327606201172}
+01/23/2022 08:11:08 - WARNING - huggingface_hub.repository - Several commits (16) will be pushed upstream.
+01/23/2022 08:11:08 - WARNING - huggingface_hub.repository - The progress bars may be unreliable.
+01/23/2022 08:13:26 - WARNING - huggingface_hub.repository - To https://huggingface.co/ncoop57/codeparrot-py
+   a4480fd..9362dd8  toasty-breeze-17 -> toasty-breeze-17
+
+01/23/2022 08:13:29 - INFO - codeparrot_training - Step 37501: {'lr': 0.0004903904494207405, 'samples': 900024, 'steps': 6250, 'loss/train': 3.4224889278411865}
+01/23/2022 08:13:29 - INFO - codeparrot_training - Step 37502: {'lr': 0.0004903904494207405, 'samples': 900048, 'steps': 6250, 'loss/train': 1.3186789751052856}
+01/23/2022 08:13:29 - INFO - codeparrot_training - Step 37503: {'lr': 0.0004903904494207405, 'samples': 900072, 'steps': 6250, 'loss/train': 2.0525577068328857}
+01/23/2022 08:13:30 - INFO - codeparrot_training - Step 37504: {'lr': 0.0004903904494207405, 'samples': 900096, 'steps': 6250, 'loss/train': 3.3865487575531006}
+01/23/2022 08:13:30 - INFO - codeparrot_training - Step 37505: {'lr': 0.0004903904494207405, 'samples': 900120, 'steps': 6250, 'loss/train': 3.6378986835479736}
+01/23/2022 08:13:30 - INFO - codeparrot_training - Step 37506: {'lr': 0.0004903904494207405, 'samples': 900144, 'steps': 6250, 'loss/train': 4.499091625213623}
+01/23/2022 08:13:31 - INFO - codeparrot_training - Step 37507: {'lr': 0.000490385955960107, 'samples': 900168, 'steps': 6251, 'loss/train': 1.8081129789352417}
+01/23/2022 08:13:31 - INFO - codeparrot_training - Step 37508: {'lr': 0.000490385955960107, 'samples': 900192, 'steps': 6251, 'loss/train': 3.6283297538757324}
+01/23/2022 08:13:32 - INFO - codeparrot_training - Step 37509: {'lr': 0.000490385955960107, 'samples': 900216, 'steps': 6251, 'loss/train': 2.0077919960021973}
+01/23/2022 08:13:32 - INFO - codeparrot_training - Step 37510: {'lr': 0.000490385955960107, 'samples': 900240, 'steps': 6251, 'loss/train': 5.048254013061523}
+01/23/2022 08:13:32 - INFO - codeparrot_training - Step 37511: {'lr': 0.000490385955960107, 'samples': 900264, 'steps': 6251, 'loss/train': 4.164653301239014}
+01/23/2022 08:13:33 - INFO - codeparrot_training - Step 37512: {'lr': 0.000490385955960107, 'samples': 900288, 'steps': 6251, 'loss/train': 3.3390047550201416}
+01/23/2022 08:13:33 - INFO - codeparrot_training - Step 37513: {'lr': 0.0004903814614697363, 'samples': 900312, 'steps': 6252, 'loss/train': 4.57468318939209}
+01/23/2022 08:13:34 - INFO - codeparrot_training - Step 37514: {'lr': 0.0004903814614697363, 'samples': 900336, 'steps': 6252, 'loss/train': 3.456688642501831}
+01/23/2022 08:13:34 - INFO - codeparrot_training - Step 37515: {'lr': 0.0004903814614697363, 'samples': 900360, 'steps': 6252, 'loss/train': 3.2044308185577393}
+01/23/2022 08:13:34 - INFO - codeparrot_training - Step 37516: {'lr': 0.0004903814614697363, 'samples': 900384, 'steps': 6252, 'loss/train': 4.102293014526367}
+01/23/2022 08:13:35 - INFO - codeparrot_training - Step 37517: {'lr': 0.0004903814614697363, 'samples': 900408, 'steps': 6252, 'loss/train': 1.4503099918365479}
+01/23/2022 08:13:35 - INFO - codeparrot_training - Step 37518: {'lr': 0.0004903814614697363, 'samples': 900432, 'steps': 6252, 'loss/train': 4.025431156158447}
+01/23/2022 08:13:36 - INFO - codeparrot_training - Step 37519: {'lr': 0.0004903769659496478, 'samples': 900456, 'steps': 6253, 'loss/train': 2.314880609512329}
+01/23/2022 08:13:36 - INFO - codeparrot_training - Step 37520: {'lr': 0.0004903769659496478, 'samples': 900480, 'steps': 6253, 'loss/train': 3.221278429031372}
+01/23/2022 08:13:36 - INFO - codeparrot_training - Step 37521: {'lr': 0.0004903769659496478, 'samples': 900504, 'steps': 6253, 'loss/train': 4.11875057220459}
+01/23/2022 08:13:37 - INFO - codeparrot_training - Step 37522: {'lr': 0.0004903769659496478, 'samples': 900528, 'steps': 6253, 'loss/train': 3.7213919162750244}
+01/23/2022 08:13:37 - INFO - codeparrot_training - Step 37523: {'lr': 0.0004903769659496478, 'samples': 900552, 'steps': 6253, 'loss/train': 4.172064781188965}
+01/23/2022 08:13:37 - INFO - codeparrot_training - Step 37524: {'lr': 0.0004903769659496478, 'samples': 900576, 'steps': 6253, 'loss/train': 3.3028979301452637}
+01/23/2022 08:13:38 - INFO - codeparrot_training - Step 37525: {'lr': 0.0004903724693998607, 'samples': 900600, 'steps': 6254, 'loss/train': 3.7406933307647705}
+01/23/2022 08:13:38 - INFO - codeparrot_training - Step 37526: {'lr': 0.0004903724693998607, 'samples': 900624, 'steps': 6254, 'loss/train': 3.6452577114105225}
+01/23/2022 08:13:39 - INFO - codeparrot_training - Step 37527: {'lr': 0.0004903724693998607, 'samples': 900648, 'steps': 6254, 'loss/train': 3.335989475250244}
+01/23/2022 08:13:39 - INFO - codeparrot_training - Step 37528: {'lr': 0.0004903724693998607, 'samples': 900672, 'steps': 6254, 'loss/train': 4.234575271606445}
+01/23/2022 08:13:39 - INFO - codeparrot_training - Step 37529: {'lr': 0.0004903724693998607, 'samples': 900696, 'steps': 6254, 'loss/train': 1.4713222980499268}
+01/23/2022 08:13:40 - INFO - codeparrot_training - Step 37530: {'lr': 0.0004903724693998607, 'samples': 900720, 'steps': 6254, 'loss/train': 1.2963242530822754}
+01/23/2022 08:13:40 - INFO - codeparrot_training - Step 37531: {'lr': 0.0004903679718203942, 'samples': 900744, 'steps': 6255, 'loss/train': 1.7340645790100098}
+01/23/2022 08:13:40 - INFO - codeparrot_training - Step 37532: {'lr': 0.0004903679718203942, 'samples': 900768, 'steps': 6255, 'loss/train': 1.5053272247314453}
+01/23/2022 08:13:41 - INFO - codeparrot_training - Step 37533: {'lr': 0.0004903679718203942, 'samples': 900792, 'steps': 6255, 'loss/train': 2.0631496906280518}
+01/23/2022 08:13:41 - INFO - codeparrot_training - Step 37534: {'lr': 0.0004903679718203942, 'samples': 900816, 'steps': 6255, 'loss/train': 1.8252447843551636}
+01/23/2022 08:13:42 - INFO - codeparrot_training - Step 37535: {'lr': 0.0004903679718203942, 'samples': 900840, 'steps': 6255, 'loss/train': 1.8859456777572632}
+01/23/2022 08:13:42 - INFO - codeparrot_training - Step 37536: {'lr': 0.0004903679718203942, 'samples': 900864, 'steps': 6255, 'loss/train': 2.2249348163604736}
+01/23/2022 08:13:42 - INFO - codeparrot_training - Step 37537: {'lr': 0.0004903634732112678, 'samples': 900888, 'steps': 6256, 'loss/train': 1.6210452318191528}
+01/23/2022 08:13:43 - INFO - codeparrot_training - Step 37538: {'lr': 0.0004903634732112678, 'samples': 900912, 'steps': 6256, 'loss/train': 1.7127647399902344}
+01/23/2022 08:13:43 - INFO - codeparrot_training - Step 37539: {'lr': 0.0004903634732112678, 'samples': 900936, 'steps': 6256, 'loss/train': 1.3490970134735107}
+01/23/2022 08:13:44 - INFO - codeparrot_training - Step 37540: {'lr': 0.0004903634732112678, 'samples': 900960, 'steps': 6256, 'loss/train': 1.6857177019119263}
+01/23/2022 08:13:44 - INFO - codeparrot_training - Step 37541: {'lr': 0.0004903634732112678, 'samples': 900984, 'steps': 6256, 'loss/train': 1.9687196016311646}
+01/23/2022 08:13:44 - INFO - codeparrot_training - Step 37542: {'lr': 0.0004903634732112678, 'samples': 901008, 'steps': 6256, 'loss/train': 1.4564141035079956}
+01/23/2022 08:13:45 - INFO - codeparrot_training - Step 37543: {'lr': 0.0004903589735725004, 'samples': 901032, 'steps': 6257, 'loss/train': 1.4199600219726562}
+01/23/2022 08:13:45 - INFO - codeparrot_training - Step 37544: {'lr': 0.0004903589735725004, 'samples': 901056, 'steps': 6257, 'loss/train': 1.441383957862854}
+01/23/2022 08:13:45 - INFO - codeparrot_training - Step 37545: {'lr': 0.0004903589735725004, 'samples': 901080, 'steps': 6257, 'loss/train': 1.484126329421997}
+01/23/2022 08:13:46 - INFO - codeparrot_training - Step 37546: {'lr': 0.0004903589735725004, 'samples': 901104, 'steps': 6257, 'loss/train': 1.3329545259475708}
+01/23/2022 08:13:46 - INFO - codeparrot_training - Step 37547: {'lr': 0.0004903589735725004, 'samples': 901128, 'steps': 6257, 'loss/train': 1.4931472539901733}
+01/23/2022 08:13:47 - INFO - codeparrot_training - Step 37548: {'lr': 0.0004903589735725004, 'samples': 901152, 'steps': 6257, 'loss/train': 1.4556130170822144}
+01/23/2022 08:13:47 - INFO - codeparrot_training - Step 37549: {'lr': 0.0004903544729041116, 'samples': 901176, 'steps': 6258, 'loss/train': 2.280463933944702}
+01/23/2022 08:13:47 - INFO - codeparrot_training - Step 37550: {'lr': 0.0004903544729041116, 'samples': 901200, 'steps': 6258, 'loss/train': 2.2042903900146484}
+01/23/2022 08:13:48 - INFO - codeparrot_training - Step 37551: {'lr': 0.0004903544729041116, 'samples': 901224, 'steps': 6258, 'loss/train': 2.5973260402679443}
+01/23/2022 08:13:48 - INFO - codeparrot_training - Step 37552: {'lr': 0.0004903544729041116, 'samples': 901248, 'steps': 6258, 'loss/train': 2.1237428188323975}
+01/23/2022 08:13:48 - INFO - codeparrot_training - Step 37553: {'lr': 0.0004903544729041116, 'samples': 901272, 'steps': 6258, 'loss/train': 2.4023711681365967}
+01/23/2022 08:13:49 - INFO - codeparrot_training - Step 37554: {'lr': 0.0004903544729041116, 'samples': 901296, 'steps': 6258, 'loss/train': 1.9494754076004028}
+01/23/2022 08:13:50 - INFO - codeparrot_training - Step 37555: {'lr': 0.0004903499712061206, 'samples': 901320, 'steps': 6259, 'loss/train': 2.7171032428741455}
+01/23/2022 08:13:50 - INFO - codeparrot_training - Step 37556: {'lr': 0.0004903499712061206, 'samples': 901344, 'steps': 6259, 'loss/train': 2.288947105407715}
+01/23/2022 08:13:51 - INFO - codeparrot_training - Step 37557: {'lr': 0.0004903499712061206, 'samples': 901368, 'steps': 6259, 'loss/train': 5.566119194030762}
+01/23/2022 08:13:51 - INFO - codeparrot_training - Step 37558: {'lr': 0.0004903499712061206, 'samples': 901392, 'steps': 6259, 'loss/train': 3.7261993885040283}
+01/23/2022 08:13:51 - INFO - codeparrot_training - Step 37559: {'lr': 0.0004903499712061206, 'samples': 901416, 'steps': 6259, 'loss/train': 3.127708911895752}
+01/23/2022 08:13:52 - INFO - codeparrot_training - Step 37560: {'lr': 0.0004903499712061206, 'samples': 901440, 'steps': 6259, 'loss/train': 1.9046757221221924}
+01/23/2022 08:13:52 - INFO - codeparrot_training - Step 37561: {'lr': 0.0004903454684785465, 'samples': 901464, 'steps': 6260, 'loss/train': 2.0267412662506104}
+01/23/2022 08:13:53 - INFO - codeparrot_training - Step 37562: {'lr': 0.0004903454684785465, 'samples': 901488, 'steps': 6260, 'loss/train': 2.5861926078796387}
+01/23/2022 08:13:53 - INFO - codeparrot_training - Step 37563: {'lr': 0.0004903454684785465, 'samples': 901512, 'steps': 6260, 'loss/train': 3.3678789138793945}
+01/23/2022 08:13:53 - INFO - codeparrot_training - Step 37564: {'lr': 0.0004903454684785465, 'samples': 901536, 'steps': 6260, 'loss/train': 5.299434185028076}
+01/23/2022 08:13:54 - INFO - codeparrot_training - Step 37565: {'lr': 0.0004903454684785465, 'samples': 901560, 'steps': 6260, 'loss/train': 4.989582538604736}
+01/23/2022 08:13:54 - INFO - codeparrot_training - Step 37566: {'lr': 0.0004903454684785465, 'samples': 901584, 'steps': 6260, 'loss/train': 2.3679869174957275}
+01/23/2022 08:13:55 - INFO - codeparrot_training - Step 37567: {'lr': 0.0004903409647214088, 'samples': 901608, 'steps': 6261, 'loss/train': 4.426783084869385}
+01/23/2022 08:13:55 - INFO - codeparrot_training - Step 37568: {'lr': 0.0004903409647214088, 'samples': 901632, 'steps': 6261, 'loss/train': 3.9701647758483887}
+01/23/2022 08:13:55 - INFO - codeparrot_training - Step 37569: {'lr': 0.0004903409647214088, 'samples': 901656, 'steps': 6261, 'loss/train': 3.0339760780334473}
+01/23/2022 08:13:56 - INFO - codeparrot_training - Step 37570: {'lr': 0.0004903409647214088, 'samples': 901680, 'steps': 6261, 'loss/train': 4.2049407958984375}
+01/23/2022 08:13:56 - INFO - codeparrot_training - Step 37571: {'lr': 0.0004903409647214088, 'samples': 901704, 'steps': 6261, 'loss/train': 3.4313740730285645}
+01/23/2022 08:13:56 - INFO - codeparrot_training - Step 37572: {'lr': 0.0004903409647214088, 'samples': 901728, 'steps': 6261, 'loss/train': 1.0982614755630493}
+01/23/2022 08:13:57 - INFO - codeparrot_training - Step 37573: {'lr': 0.0004903364599347268, 'samples': 901752, 'steps': 6262, 'loss/train': 3.8730950355529785}
+01/23/2022 08:13:57 - INFO - codeparrot_training - Step 37574: {'lr': 0.0004903364599347268, 'samples': 901776, 'steps': 6262, 'loss/train': 4.83231782913208}
+01/23/2022 08:13:58 - INFO - codeparrot_training - Step 37575: {'lr': 0.0004903364599347268, 'samples': 901800, 'steps': 6262, 'loss/train': 5.2223639488220215}
+01/23/2022 08:13:58 - INFO - codeparrot_training - Step 37576: {'lr': 0.0004903364599347268, 'samples': 901824, 'steps': 6262, 'loss/train': 3.819498300552368}
+01/23/2022 08:13:58 - INFO - codeparrot_training - Step 37577: {'lr': 0.0004903364599347268, 'samples': 901848, 'steps': 6262, 'loss/train': 4.063527584075928}
+01/23/2022 08:13:59 - INFO - codeparrot_training - Step 37578: {'lr': 0.0004903364599347268, 'samples': 901872, 'steps': 6262, 'loss/train': 2.792623519897461}
+01/23/2022 08:13:59 - INFO - codeparrot_training - Step 37579: {'lr': 0.0004903319541185196, 'samples': 901896, 'steps': 6263, 'loss/train': 2.27655291557312}
+01/23/2022 08:14:00 - INFO - codeparrot_training - Step 37580: {'lr': 0.0004903319541185196, 'samples': 901920, 'steps': 6263, 'loss/train': 3.7601327896118164}
+01/23/2022 08:14:00 - INFO - codeparrot_training - Step 37581: {'lr': 0.0004903319541185196, 'samples': 901944, 'steps': 6263, 'loss/train': 3.565462112426758}
+01/23/2022 08:14:00 - INFO - codeparrot_training - Step 37582: {'lr': 0.0004903319541185196, 'samples': 901968, 'steps': 6263, 'loss/train': 3.7000505924224854}
+01/23/2022 08:14:01 - INFO - codeparrot_training - Step 37583: {'lr': 0.0004903319541185196, 'samples': 901992, 'steps': 6263, 'loss/train': 3.2316994667053223}
+01/23/2022 08:14:01 - INFO - codeparrot_training - Step 37584: {'lr': 0.0004903319541185196, 'samples': 902016, 'steps': 6263, 'loss/train': 2.7217679023742676}
+01/23/2022 08:14:01 - INFO - codeparrot_training - Step 37585: {'lr': 0.0004903274472728067, 'samples': 902040, 'steps': 6264, 'loss/train': 3.9970855712890625}
+01/23/2022 08:14:02 - INFO - codeparrot_training - Step 37586: {'lr': 0.0004903274472728067, 'samples': 902064, 'steps': 6264, 'loss/train': 3.6334378719329834}
+01/23/2022 08:14:02 - INFO - codeparrot_training - Step 37587: {'lr': 0.0004903274472728067, 'samples': 902088, 'steps': 6264, 'loss/train': 2.61781907081604}
+01/23/2022 08:14:03 - INFO - codeparrot_training - Step 37588: {'lr': 0.0004903274472728067, 'samples': 902112, 'steps': 6264, 'loss/train': 3.608214855194092}
+01/23/2022 08:14:03 - INFO - codeparrot_training - Step 37589: {'lr': 0.0004903274472728067, 'samples': 902136, 'steps': 6264, 'loss/train': 3.6204795837402344}
+01/23/2022 08:14:03 - INFO - codeparrot_training - Step 37590: {'lr': 0.0004903274472728067, 'samples': 902160, 'steps': 6264, 'loss/train': 3.2772772312164307}
+01/23/2022 08:14:04 - INFO - codeparrot_training - Step 37591: {'lr': 0.0004903229393976073, 'samples': 902184, 'steps': 6265, 'loss/train': 4.117708206176758}
+01/23/2022 08:14:04 - INFO - codeparrot_training - Step 37592: {'lr': 0.0004903229393976073, 'samples': 902208, 'steps': 6265, 'loss/train': 3.5142672061920166}
+01/23/2022 08:14:07 - INFO - codeparrot_training - Step 37593: {'lr': 0.0004903229393976073, 'samples': 902232, 'steps': 6265, 'loss/train': 4.110105514526367}
+01/23/2022 08:14:07 - INFO - codeparrot_training - Step 37594: {'lr': 0.0004903229393976073, 'samples': 902256, 'steps': 6265, 'loss/train': 4.239628314971924}
+01/23/2022 08:14:08 - INFO - codeparrot_training - Step 37595: {'lr': 0.0004903229393976073, 'samples': 902280, 'steps': 6265, 'loss/train': 3.8482789993286133}
+01/23/2022 08:14:08 - INFO - codeparrot_training - Step 37596: {'lr': 0.0004903229393976073, 'samples': 902304, 'steps': 6265, 'loss/train': 3.257831335067749}
+01/23/2022 08:14:09 - INFO - codeparrot_training - Step 37597: {'lr': 0.0004903184304929408, 'samples': 902328, 'steps': 6266, 'loss/train': 3.442838430404663}
+01/23/2022 08:14:09 - INFO - codeparrot_training - Step 37598: {'lr': 0.0004903184304929408, 'samples': 902352, 'steps': 6266, 'loss/train': 0.7810565829277039}
+01/23/2022 08:14:09 - INFO - codeparrot_training - Step 37599: {'lr': 0.0004903184304929408, 'samples': 902376, 'steps': 6266, 'loss/train': 4.392501354217529}
+01/23/2022 08:14:10 - INFO - codeparrot_training - Step 37600: {'lr': 0.0004903184304929408, 'samples': 902400, 'steps': 6266, 'loss/train': 3.2622876167297363}
+01/23/2022 08:14:10 - INFO - codeparrot_training - Step 37601: {'lr': 0.0004903184304929408, 'samples': 902424, 'steps': 6266, 'loss/train': 3.6535778045654297}
+01/23/2022 08:14:11 - INFO - codeparrot_training - Step 37602: {'lr': 0.0004903184304929408, 'samples': 902448, 'steps': 6266, 'loss/train': 2.3122365474700928}
+01/23/2022 08:14:11 - INFO - codeparrot_training - Step 37603: {'lr': 0.0004903139205588264, 'samples': 902472, 'steps': 6267, 'loss/train': 2.2866344451904297}
+01/23/2022 08:14:11 - INFO - codeparrot_training - Step 37604: {'lr': 0.0004903139205588264, 'samples': 902496, 'steps': 6267, 'loss/train': 2.3332748413085938}
+01/23/2022 08:14:12 - INFO - codeparrot_training - Step 37605: {'lr': 0.0004903139205588264, 'samples': 902520, 'steps': 6267, 'loss/train': 2.290217876434326}
+01/23/2022 08:14:12 - INFO - codeparrot_training - Step 37606: {'lr': 0.0004903139205588264, 'samples': 902544, 'steps': 6267, 'loss/train': 2.38089656829834}
+01/23/2022 08:14:12 - INFO - codeparrot_training - Step 37607: {'lr': 0.0004903139205588264, 'samples': 902568, 'steps': 6267, 'loss/train': 2.386378049850464}
+01/23/2022 08:14:13 - INFO - codeparrot_training - Step 37608: {'lr': 0.0004903139205588264, 'samples': 902592, 'steps': 6267, 'loss/train': 2.2600221633911133}
+01/23/2022 08:14:13 - INFO - codeparrot_training - Step 37609: {'lr': 0.0004903094095952834, 'samples': 902616, 'steps': 6268, 'loss/train': 3.344865083694458}
+01/23/2022 08:14:14 - INFO - codeparrot_training - Step 37610: {'lr': 0.0004903094095952834, 'samples': 902640, 'steps': 6268, 'loss/train': 3.37148380279541}
+01/23/2022 08:14:14 - INFO - codeparrot_training - Step 37611: {'lr': 0.0004903094095952834, 'samples': 902664, 'steps': 6268, 'loss/train': 4.019938945770264}
+01/23/2022 08:14:14 - INFO - codeparrot_training - Step 37612: {'lr': 0.0004903094095952834, 'samples': 902688, 'steps': 6268, 'loss/train': 3.119903326034546}
+01/23/2022 08:14:15 - INFO - codeparrot_training - Step 37613: {'lr': 0.0004903094095952834, 'samples': 902712, 'steps': 6268, 'loss/train': 3.0245606899261475}
+01/23/2022 08:14:15 - INFO - codeparrot_training - Step 37614: {'lr': 0.0004903094095952834, 'samples': 902736, 'steps': 6268, 'loss/train': 4.173507213592529}
+01/23/2022 08:14:16 - INFO - codeparrot_training - Step 37615: {'lr': 0.0004903048976023313, 'samples': 902760, 'steps': 6269, 'loss/train': 3.7592990398406982}
+01/23/2022 08:14:16 - INFO - codeparrot_training - Step 37616: {'lr': 0.0004903048976023313, 'samples': 902784, 'steps': 6269, 'loss/train': 3.533968687057495}
+01/23/2022 08:14:16 - INFO - codeparrot_training - Step 37617: {'lr': 0.0004903048976023313, 'samples': 902808, 'steps': 6269, 'loss/train': 3.4178497791290283}
+01/23/2022 08:14:17 - INFO - codeparrot_training - Step 37618: {'lr': 0.0004903048976023313, 'samples': 902832, 'steps': 6269, 'loss/train': 3.848008394241333}
+01/23/2022 08:14:17 - INFO - codeparrot_training - Step 37619: {'lr': 0.0004903048976023313, 'samples': 902856, 'steps': 6269, 'loss/train': 3.504366159439087}
+01/23/2022 08:14:17 - INFO - codeparrot_training - Step 37620: {'lr': 0.0004903048976023313, 'samples': 902880, 'steps': 6269, 'loss/train': 2.5294606685638428}
+01/23/2022 08:14:18 - INFO - codeparrot_training - Step 37621: {'lr': 0.0004903003845799893, 'samples': 902904, 'steps': 6270, 'loss/train': 3.799109697341919}
+01/23/2022 08:14:18 - INFO - codeparrot_training - Step 37622: {'lr': 0.0004903003845799893, 'samples': 902928, 'steps': 6270, 'loss/train': 3.8772225379943848}
+01/23/2022 08:14:19 - INFO - codeparrot_training - Step 37623: {'lr': 0.0004903003845799893, 'samples': 902952, 'steps': 6270, 'loss/train': 4.155581474304199}
+01/23/2022 08:14:19 - INFO - codeparrot_training - Step 37624: {'lr': 0.0004903003845799893, 'samples': 902976, 'steps': 6270, 'loss/train': 3.4999380111694336}
+01/23/2022 08:14:19 - INFO - codeparrot_training - Step 37625: {'lr': 0.0004903003845799893, 'samples': 903000, 'steps': 6270, 'loss/train': 4.869592666625977}
+01/23/2022 08:14:20 - INFO - codeparrot_training - Step 37626: {'lr': 0.0004903003845799893, 'samples': 903024, 'steps': 6270, 'loss/train': 3.263678550720215}
+01/23/2022 08:14:20 - INFO - codeparrot_training - Step 37627: {'lr': 0.0004902958705282767, 'samples': 903048, 'steps': 6271, 'loss/train': 3.7850239276885986}
+01/23/2022 08:14:20 - INFO - codeparrot_training - Step 37628: {'lr': 0.0004902958705282767, 'samples': 903072, 'steps': 6271, 'loss/train': 4.266970157623291}
+01/23/2022 08:14:21 - INFO - codeparrot_training - Step 37629: {'lr': 0.0004902958705282767, 'samples': 903096, 'steps': 6271, 'loss/train': 2.8915436267852783}
+01/23/2022 08:14:21 - INFO - codeparrot_training - Step 37630: {'lr': 0.0004902958705282767, 'samples': 903120, 'steps': 6271, 'loss/train': 3.8451972007751465}
+01/23/2022 08:14:22 - INFO - codeparrot_training - Step 37631: {'lr': 0.0004902958705282767, 'samples': 903144, 'steps': 6271, 'loss/train': 2.5345876216888428}
+01/23/2022 08:14:22 - INFO - codeparrot_training - Step 37632: {'lr': 0.0004902958705282767, 'samples': 903168, 'steps': 6271, 'loss/train': 3.4366872310638428}
+01/23/2022 08:14:23 - INFO - codeparrot_training - Step 37633: {'lr': 0.000490291355447213, 'samples': 903192, 'steps': 6272, 'loss/train': 3.7665860652923584}
+01/23/2022 08:14:23 - INFO - codeparrot_training - Step 37634: {'lr': 0.000490291355447213, 'samples': 903216, 'steps': 6272, 'loss/train': 3.1111855506896973}
+01/23/2022 08:14:24 - INFO - codeparrot_training - Step 37635: {'lr': 0.000490291355447213, 'samples': 903240, 'steps': 6272, 'loss/train': 2.646123170852661}
+01/23/2022 08:14:24 - INFO - codeparrot_training - Step 37636: {'lr': 0.000490291355447213, 'samples': 903264, 'steps': 6272, 'loss/train': 3.962960958480835}
+01/23/2022 08:14:25 - INFO - codeparrot_training - Step 37637: {'lr': 0.000490291355447213, 'samples': 903288, 'steps': 6272, 'loss/train': 3.225156784057617}
+01/23/2022 08:14:25 - INFO - codeparrot_training - Step 37638: {'lr': 0.000490291355447213, 'samples': 903312, 'steps': 6272, 'loss/train': 1.4839524030685425}
+01/23/2022 08:14:25 - INFO - codeparrot_training - Step 37639: {'lr': 0.0004902868393368174, 'samples': 903336, 'steps': 6273, 'loss/train': 1.8856667280197144}
+01/23/2022 08:14:26 - INFO - codeparrot_training - Step 37640: {'lr': 0.0004902868393368174, 'samples': 903360, 'steps': 6273, 'loss/train': 3.7147183418273926}
+01/23/2022 08:14:26 - INFO - codeparrot_training - Step 37641: {'lr': 0.0004902868393368174, 'samples': 903384, 'steps': 6273, 'loss/train': 1.454135775566101}
+01/23/2022 08:14:27 - INFO - codeparrot_training - Step 37642: {'lr': 0.0004902868393368174, 'samples': 903408, 'steps': 6273, 'loss/train': 3.03653883934021}
+01/23/2022 08:14:27 - INFO - codeparrot_training - Step 37643: {'lr': 0.0004902868393368174, 'samples': 903432, 'steps': 6273, 'loss/train': 2.9382543563842773}
+01/23/2022 08:14:27 - INFO - codeparrot_training - Step 37644: {'lr': 0.0004902868393368174, 'samples': 903456, 'steps': 6273, 'loss/train': 3.783289909362793}
+01/23/2022 08:14:28 - INFO - codeparrot_training - Step 37645: {'lr': 0.0004902823221971092, 'samples': 903480, 'steps': 6274, 'loss/train': 3.38983416557312}
+01/23/2022 08:14:28 - INFO - codeparrot_training - Step 37646: {'lr': 0.0004902823221971092, 'samples': 903504, 'steps': 6274, 'loss/train': 3.5525505542755127}
+01/23/2022 08:14:28 - INFO - codeparrot_training - Step 37647: {'lr': 0.0004902823221971092, 'samples': 903528, 'steps': 6274, 'loss/train': 2.812250852584839}
+01/23/2022 08:14:29 - INFO - codeparrot_training - Step 37648: {'lr': 0.0004902823221971092, 'samples': 903552, 'steps': 6274, 'loss/train': 3.541242837905884}
+01/23/2022 08:14:29 - INFO - codeparrot_training - Step 37649: {'lr': 0.0004902823221971092, 'samples': 903576, 'steps': 6274, 'loss/train': 2.1092965602874756}
+01/23/2022 08:14:30 - INFO - codeparrot_training - Step 37650: {'lr': 0.0004902823221971092, 'samples': 903600, 'steps': 6274, 'loss/train': 3.894580364227295}
+01/23/2022 08:14:30 - INFO - codeparrot_training - Step 37651: {'lr': 0.000490277804028108, 'samples': 903624, 'steps': 6275, 'loss/train': 1.8260362148284912}
+01/23/2022 08:14:30 - INFO - codeparrot_training - Step 37652: {'lr': 0.000490277804028108, 'samples': 903648, 'steps': 6275, 'loss/train': 3.3551127910614014}
+01/23/2022 08:14:31 - INFO - codeparrot_training - Step 37653: {'lr': 0.000490277804028108, 'samples': 903672, 'steps': 6275, 'loss/train': 3.2720091342926025}
+01/23/2022 08:14:31 - INFO - codeparrot_training - Step 37654: {'lr': 0.000490277804028108, 'samples': 903696, 'steps': 6275, 'loss/train': 2.2118167877197266}
+01/23/2022 08:14:31 - INFO - codeparrot_training - Step 37655: {'lr': 0.000490277804028108, 'samples': 903720, 'steps': 6275, 'loss/train': 3.8786520957946777}
+01/23/2022 08:14:32 - INFO - codeparrot_training - Step 37656: {'lr': 0.000490277804028108, 'samples': 903744, 'steps': 6275, 'loss/train': 2.672150135040283}
+01/23/2022 08:14:32 - INFO - codeparrot_training - Step 37657: {'lr': 0.0004902732848298328, 'samples': 903768, 'steps': 6276, 'loss/train': 3.0499513149261475}
+01/23/2022 08:14:33 - INFO - codeparrot_training - Step 37658: {'lr': 0.0004902732848298328, 'samples': 903792, 'steps': 6276, 'loss/train': 2.7271006107330322}
+01/23/2022 08:14:33 - INFO - codeparrot_training - Step 37659: {'lr': 0.0004902732848298328, 'samples': 903816, 'steps': 6276, 'loss/train': 3.100252151489258}
+01/23/2022 08:14:33 - INFO - codeparrot_training - Step 37660: {'lr': 0.0004902732848298328, 'samples': 903840, 'steps': 6276, 'loss/train': 3.3098411560058594}
+01/23/2022 08:14:34 - INFO - codeparrot_training - Step 37661: {'lr': 0.0004902732848298328, 'samples': 903864, 'steps': 6276, 'loss/train': 3.751389503479004}
+01/23/2022 08:14:34 - INFO - codeparrot_training - Step 37662: {'lr': 0.0004902732848298328, 'samples': 903888, 'steps': 6276, 'loss/train': 2.6268961429595947}
+01/23/2022 08:14:35 - INFO - codeparrot_training - Step 37663: {'lr': 0.0004902687646023032, 'samples': 903912, 'steps': 6277, 'loss/train': 4.152556419372559}
+01/23/2022 08:14:35 - INFO - codeparrot_training - Step 37664: {'lr': 0.0004902687646023032, 'samples': 903936, 'steps': 6277, 'loss/train': 3.1295621395111084}
+01/23/2022 08:14:36 - INFO - codeparrot_training - Step 37665: {'lr': 0.0004902687646023032, 'samples': 903960, 'steps': 6277, 'loss/train': 1.7684074640274048}
+01/23/2022 08:14:36 - INFO - codeparrot_training - Step 37666: {'lr': 0.0004902687646023032, 'samples': 903984, 'steps': 6277, 'loss/train': 2.551396608352661}
+01/23/2022 08:14:36 - INFO - codeparrot_training - Step 37667: {'lr': 0.0004902687646023032, 'samples': 904008, 'steps': 6277, 'loss/train': 2.3114256858825684}
+01/23/2022 08:14:37 - INFO - codeparrot_training - Step 37668: {'lr': 0.0004902687646023032, 'samples': 904032, 'steps': 6277, 'loss/train': 3.481093645095825}
+01/23/2022 08:14:37 - INFO - codeparrot_training - Step 37669: {'lr': 0.0004902642433455385, 'samples': 904056, 'steps': 6278, 'loss/train': 1.9966479539871216}
+01/23/2022 08:14:37 - INFO - codeparrot_training - Step 37670: {'lr': 0.0004902642433455385, 'samples': 904080, 'steps': 6278, 'loss/train': 3.8825149536132812}
+01/23/2022 08:14:38 - INFO - codeparrot_training - Step 37671: {'lr': 0.0004902642433455385, 'samples': 904104, 'steps': 6278, 'loss/train': 5.059380531311035}
+01/23/2022 08:14:38 - INFO - codeparrot_training - Step 37672: {'lr': 0.0004902642433455385, 'samples': 904128, 'steps': 6278, 'loss/train': 3.6976728439331055}
+01/23/2022 08:14:39 - INFO - codeparrot_training - Step 37673: {'lr': 0.0004902642433455385, 'samples': 904152, 'steps': 6278, 'loss/train': 2.757147789001465}
+01/23/2022 08:14:40 - INFO - codeparrot_training - Step 37674: {'lr': 0.0004902642433455385, 'samples': 904176, 'steps': 6278, 'loss/train': 2.8882856369018555}
+01/23/2022 08:14:40 - INFO - codeparrot_training - Step 37675: {'lr': 0.0004902597210595581, 'samples': 904200, 'steps': 6279, 'loss/train': 3.1395084857940674}
+01/23/2022 08:14:41 - INFO - codeparrot_training - Step 37676: {'lr': 0.0004902597210595581, 'samples': 904224, 'steps': 6279, 'loss/train': 3.3885509967803955}
+01/23/2022 08:14:41 - INFO - codeparrot_training - Step 37677: {'lr': 0.0004902597210595581, 'samples': 904248, 'steps': 6279, 'loss/train': 2.2416677474975586}
+01/23/2022 08:14:41 - INFO - codeparrot_training - Step 37678: {'lr': 0.0004902597210595581, 'samples': 904272, 'steps': 6279, 'loss/train': 2.4258742332458496}
+01/23/2022 08:14:42 - INFO - codeparrot_training - Step 37679: {'lr': 0.0004902597210595581, 'samples': 904296, 'steps': 6279, 'loss/train': 2.2132132053375244}
+01/23/2022 08:14:42 - INFO - codeparrot_training - Step 37680: {'lr': 0.0004902597210595581, 'samples': 904320, 'steps': 6279, 'loss/train': 2.652662515640259}
+01/23/2022 08:14:42 - INFO - codeparrot_training - Step 37681: {'lr': 0.0004902551977443813, 'samples': 904344, 'steps': 6280, 'loss/train': 2.3196191787719727}
+01/23/2022 08:14:43 - INFO - codeparrot_training - Step 37682: {'lr': 0.0004902551977443813, 'samples': 904368, 'steps': 6280, 'loss/train': 3.809709310531616}
+01/23/2022 08:14:43 - INFO - codeparrot_training - Step 37683: {'lr': 0.0004902551977443813, 'samples': 904392, 'steps': 6280, 'loss/train': 1.554927945137024}
+01/23/2022 08:14:44 - INFO - codeparrot_training - Step 37684: {'lr': 0.0004902551977443813, 'samples': 904416, 'steps': 6280, 'loss/train': 3.5474653244018555}
+01/23/2022 08:14:44 - INFO - codeparrot_training - Step 37685: {'lr': 0.0004902551977443813, 'samples': 904440, 'steps': 6280, 'loss/train': 3.4337799549102783}
+01/23/2022 08:14:44 - INFO - codeparrot_training - Step 37686: {'lr': 0.0004902551977443813, 'samples': 904464, 'steps': 6280, 'loss/train': 3.6570115089416504}
+01/23/2022 08:14:45 - INFO - codeparrot_training - Step 37687: {'lr': 0.0004902506734000276, 'samples': 904488, 'steps': 6281, 'loss/train': 3.1090476512908936}
+01/23/2022 08:14:45 - INFO - codeparrot_training - Step 37688: {'lr': 0.0004902506734000276, 'samples': 904512, 'steps': 6281, 'loss/train': 5.097733020782471}
+01/23/2022 08:14:46 - INFO - codeparrot_training - Step 37689: {'lr': 0.0004902506734000276, 'samples': 904536, 'steps': 6281, 'loss/train': 3.0065667629241943}
+01/23/2022 08:14:46 - INFO - codeparrot_training - Step 37690: {'lr': 0.0004902506734000276, 'samples': 904560, 'steps': 6281, 'loss/train': 2.296613931655884}
+01/23/2022 08:14:46 - INFO - codeparrot_training - Step 37691: {'lr': 0.0004902506734000276, 'samples': 904584, 'steps': 6281, 'loss/train': 3.560441493988037}
+01/23/2022 08:14:47 - INFO - codeparrot_training - Step 37692: {'lr': 0.0004902506734000276, 'samples': 904608, 'steps': 6281, 'loss/train': 2.8691117763519287}
+01/23/2022 08:14:47 - INFO - codeparrot_training - Step 37693: {'lr': 0.0004902461480265163, 'samples': 904632, 'steps': 6282, 'loss/train': 3.713400363922119}
+01/23/2022 08:14:47 - INFO - codeparrot_training - Step 37694: {'lr': 0.0004902461480265163, 'samples': 904656, 'steps': 6282, 'loss/train': 3.4958531856536865}
+01/23/2022 08:14:48 - INFO - codeparrot_training - Step 37695: {'lr': 0.0004902461480265163, 'samples': 904680, 'steps': 6282, 'loss/train': 1.2169078588485718}
+01/23/2022 08:14:48 - INFO - codeparrot_training - Step 37696: {'lr': 0.0004902461480265163, 'samples': 904704, 'steps': 6282, 'loss/train': 1.5273609161376953}
+01/23/2022 08:14:49 - INFO - codeparrot_training - Step 37697: {'lr': 0.0004902461480265163, 'samples': 904728, 'steps': 6282, 'loss/train': 2.4769866466522217}
+01/23/2022 08:14:49 - INFO - codeparrot_training - Step 37698: {'lr': 0.0004902461480265163, 'samples': 904752, 'steps': 6282, 'loss/train': 3.9737329483032227}
+01/23/2022 08:14:49 - INFO - codeparrot_training - Step 37699: {'lr': 0.0004902416216238667, 'samples': 904776, 'steps': 6283, 'loss/train': 2.1553244590759277}
+01/23/2022 08:14:50 - INFO - codeparrot_training - Step 37700: {'lr': 0.0004902416216238667, 'samples': 904800, 'steps': 6283, 'loss/train': 4.176263809204102}
+01/23/2022 08:14:50 - INFO - codeparrot_training - Step 37701: {'lr': 0.0004902416216238667, 'samples': 904824, 'steps': 6283, 'loss/train': 2.588149070739746}
+01/23/2022 08:14:51 - INFO - codeparrot_training - Step 37702: {'lr': 0.0004902416216238667, 'samples': 904848, 'steps': 6283, 'loss/train': 3.5810234546661377}
+01/23/2022 08:14:51 - INFO - codeparrot_training - Step 37703: {'lr': 0.0004902416216238667, 'samples': 904872, 'steps': 6283, 'loss/train': 2.9737820625305176}
+01/23/2022 08:14:51 - INFO - codeparrot_training - Step 37704: {'lr': 0.0004902416216238667, 'samples': 904896, 'steps': 6283, 'loss/train': 3.836458921432495}
+01/23/2022 08:14:52 - INFO - codeparrot_training - Step 37705: {'lr': 0.0004902370941920984, 'samples': 904920, 'steps': 6284, 'loss/train': 3.026304244995117}
+01/23/2022 08:14:52 - INFO - codeparrot_training - Step 37706: {'lr': 0.0004902370941920984, 'samples': 904944, 'steps': 6284, 'loss/train': 2.6805169582366943}
+01/23/2022 08:14:52 - INFO - codeparrot_training - Step 37707: {'lr': 0.0004902370941920984, 'samples': 904968, 'steps': 6284, 'loss/train': 2.3823459148406982}
+01/23/2022 08:14:53 - INFO - codeparrot_training - Step 37708: {'lr': 0.0004902370941920984, 'samples': 904992, 'steps': 6284, 'loss/train': 2.887446165084839}
+01/23/2022 08:14:53 - INFO - codeparrot_training - Step 37709: {'lr': 0.0004902370941920984, 'samples': 905016, 'steps': 6284, 'loss/train': 2.8621132373809814}
+01/23/2022 08:14:54 - INFO - codeparrot_training - Step 37710: {'lr': 0.0004902370941920984, 'samples': 905040, 'steps': 6284, 'loss/train': 2.4248077869415283}
+01/23/2022 08:14:54 - INFO - codeparrot_training - Step 37711: {'lr': 0.0004902325657312306, 'samples': 905064, 'steps': 6285, 'loss/train': 2.2736806869506836}
+01/23/2022 08:14:54 - INFO - codeparrot_training - Step 37712: {'lr': 0.0004902325657312306, 'samples': 905088, 'steps': 6285, 'loss/train': 2.984248399734497}
+01/23/2022 08:14:55 - INFO - codeparrot_training - Step 37713: {'lr': 0.0004902325657312306, 'samples': 905112, 'steps': 6285, 'loss/train': 3.180389404296875}
+01/23/2022 08:14:56 - INFO - codeparrot_training - Step 37714: {'lr': 0.0004902325657312306, 'samples': 905136, 'steps': 6285, 'loss/train': 3.668384313583374}
+01/23/2022 08:14:56 - INFO - codeparrot_training - Step 37715: {'lr': 0.0004902325657312306, 'samples': 905160, 'steps': 6285, 'loss/train': 3.486767530441284}
+01/23/2022 08:14:56 - INFO - codeparrot_training - Step 37716: {'lr': 0.0004902325657312306, 'samples': 905184, 'steps': 6285, 'loss/train': 3.3007867336273193}
+01/23/2022 08:14:57 - INFO - codeparrot_training - Step 37717: {'lr': 0.0004902280362412828, 'samples': 905208, 'steps': 6286, 'loss/train': 3.5221593379974365}
+01/23/2022 08:14:57 - INFO - codeparrot_training - Step 37718: {'lr': 0.0004902280362412828, 'samples': 905232, 'steps': 6286, 'loss/train': 3.0277600288391113}
+01/23/2022 08:14:58 - INFO - codeparrot_training - Step 37719: {'lr': 0.0004902280362412828, 'samples': 905256, 'steps': 6286, 'loss/train': 4.027431488037109}
+01/23/2022 08:14:58 - INFO - codeparrot_training - Step 37720: {'lr': 0.0004902280362412828, 'samples': 905280, 'steps': 6286, 'loss/train': 2.84482741355896}
+01/23/2022 08:14:58 - INFO - codeparrot_training - Step 37721: {'lr': 0.0004902280362412828, 'samples': 905304, 'steps': 6286, 'loss/train': 2.520885467529297}
+01/23/2022 08:14:59 - INFO - codeparrot_training - Step 37722: {'lr': 0.0004902280362412828, 'samples': 905328, 'steps': 6286, 'loss/train': 1.1265738010406494}
+01/23/2022 08:14:59 - INFO - codeparrot_training - Step 37723: {'lr': 0.0004902235057222743, 'samples': 905352, 'steps': 6287, 'loss/train': 2.1344265937805176}
+01/23/2022 08:15:00 - INFO - codeparrot_training - Step 37724: {'lr': 0.0004902235057222743, 'samples': 905376, 'steps': 6287, 'loss/train': 4.99800443649292}
+01/23/2022 08:15:00 - INFO - codeparrot_training - Step 37725: {'lr': 0.0004902235057222743, 'samples': 905400, 'steps': 6287, 'loss/train': 2.317671775817871}
+01/23/2022 08:15:00 - INFO - codeparrot_training - Step 37726: {'lr': 0.0004902235057222743, 'samples': 905424, 'steps': 6287, 'loss/train': 3.0260074138641357}
+01/23/2022 08:15:01 - INFO - codeparrot_training - Step 37727: {'lr': 0.0004902235057222743, 'samples': 905448, 'steps': 6287, 'loss/train': 3.0189433097839355}
+01/23/2022 08:15:01 - INFO - codeparrot_training - Step 37728: {'lr': 0.0004902235057222743, 'samples': 905472, 'steps': 6287, 'loss/train': 2.8333301544189453}
+01/23/2022 08:15:02 - INFO - codeparrot_training - Step 37729: {'lr': 0.0004902189741742246, 'samples': 905496, 'steps': 6288, 'loss/train': 3.8287456035614014}
+01/23/2022 08:15:02 - INFO - codeparrot_training - Step 37730: {'lr': 0.0004902189741742246, 'samples': 905520, 'steps': 6288, 'loss/train': 2.6554510593414307}
+01/23/2022 08:15:02 - INFO - codeparrot_training - Step 37731: {'lr': 0.0004902189741742246, 'samples': 905544, 'steps': 6288, 'loss/train': 2.160921573638916}
+01/23/2022 08:15:03 - INFO - codeparrot_training - Step 37732: {'lr': 0.0004902189741742246, 'samples': 905568, 'steps': 6288, 'loss/train': 1.8292348384857178}
+01/23/2022 08:15:03 - INFO - codeparrot_training - Step 37733: {'lr': 0.0004902189741742246, 'samples': 905592, 'steps': 6288, 'loss/train': 2.8487229347229004}
+01/23/2022 08:15:03 - INFO - codeparrot_training - Step 37734: {'lr': 0.0004902189741742246, 'samples': 905616, 'steps': 6288, 'loss/train': 1.8693689107894897}
+01/23/2022 08:15:04 - INFO - codeparrot_training - Step 37735: {'lr': 0.0004902144415971532, 'samples': 905640, 'steps': 6289, 'loss/train': 3.2635750770568848}
+01/23/2022 08:15:04 - INFO - codeparrot_training - Step 37736: {'lr': 0.0004902144415971532, 'samples': 905664, 'steps': 6289, 'loss/train': 3.33247709274292}
+01/23/2022 08:15:05 - INFO - codeparrot_training - Step 37737: {'lr': 0.0004902144415971532, 'samples': 905688, 'steps': 6289, 'loss/train': 3.5691590309143066}
+01/23/2022 08:15:05 - INFO - codeparrot_training - Step 37738: {'lr': 0.0004902144415971532, 'samples': 905712, 'steps': 6289, 'loss/train': 3.2798454761505127}
+01/23/2022 08:15:05 - INFO - codeparrot_training - Step 37739: {'lr': 0.0004902144415971532, 'samples': 905736, 'steps': 6289, 'loss/train': 4.932941913604736}
+01/23/2022 08:15:06 - INFO - codeparrot_training - Step 37740: {'lr': 0.0004902144415971532, 'samples': 905760, 'steps': 6289, 'loss/train': 3.0339231491088867}
+01/23/2022 08:15:06 - INFO - codeparrot_training - Step 37741: {'lr': 0.0004902099079910794, 'samples': 905784, 'steps': 6290, 'loss/train': 2.4675583839416504}
+01/23/2022 08:15:07 - INFO - codeparrot_training - Step 37742: {'lr': 0.0004902099079910794, 'samples': 905808, 'steps': 6290, 'loss/train': 3.2239809036254883}
+01/23/2022 08:15:07 - INFO - codeparrot_training - Step 37743: {'lr': 0.0004902099079910794, 'samples': 905832, 'steps': 6290, 'loss/train': 3.2205488681793213}
+01/23/2022 08:15:07 - INFO - codeparrot_training - Step 37744: {'lr': 0.0004902099079910794, 'samples': 905856, 'steps': 6290, 'loss/train': 3.2861483097076416}
+01/23/2022 08:15:08 - INFO - codeparrot_training - Step 37745: {'lr': 0.0004902099079910794, 'samples': 905880, 'steps': 6290, 'loss/train': 3.7654635906219482}
+01/23/2022 08:15:08 - INFO - codeparrot_training - Step 37746: {'lr': 0.0004902099079910794, 'samples': 905904, 'steps': 6290, 'loss/train': 2.8685615062713623}
+01/23/2022 08:15:08 - INFO - codeparrot_training - Step 37747: {'lr': 0.0004902053733560225, 'samples': 905928, 'steps': 6291, 'loss/train': 2.4852848052978516}
+01/23/2022 08:15:09 - INFO - codeparrot_training - Step 37748: {'lr': 0.0004902053733560225, 'samples': 905952, 'steps': 6291, 'loss/train': 3.462493896484375}
+01/23/2022 08:15:09 - INFO - codeparrot_training - Step 37749: {'lr': 0.0004902053733560225, 'samples': 905976, 'steps': 6291, 'loss/train': 4.089900493621826}
+01/23/2022 08:15:10 - INFO - codeparrot_training - Step 37750: {'lr': 0.0004902053733560225, 'samples': 906000, 'steps': 6291, 'loss/train': 3.857781410217285}
+01/23/2022 08:15:10 - INFO - codeparrot_training - Step 37751: {'lr': 0.0004902053733560225, 'samples': 906024, 'steps': 6291, 'loss/train': 4.64593505859375}
+01/23/2022 08:15:10 - INFO - codeparrot_training - Step 37752: {'lr': 0.0004902053733560225, 'samples': 906048, 'steps': 6291, 'loss/train': 3.329068183898926}
+01/23/2022 08:15:12 - INFO - codeparrot_training - Step 37753: {'lr': 0.0004902008376920021, 'samples': 906072, 'steps': 6292, 'loss/train': 4.116684913635254}
+01/23/2022 08:15:12 - INFO - codeparrot_training - Step 37754: {'lr': 0.0004902008376920021, 'samples': 906096, 'steps': 6292, 'loss/train': 2.3550291061401367}
+01/23/2022 08:15:12 - INFO - codeparrot_training - Step 37755: {'lr': 0.0004902008376920021, 'samples': 906120, 'steps': 6292, 'loss/train': 4.093232154846191}
+01/23/2022 08:15:13 - INFO - codeparrot_training - Step 37756: {'lr': 0.0004902008376920021, 'samples': 906144, 'steps': 6292, 'loss/train': 3.4960896968841553}
+01/23/2022 08:15:13 - INFO - codeparrot_training - Step 37757: {'lr': 0.0004902008376920021, 'samples': 906168, 'steps': 6292, 'loss/train': 3.441514492034912}
+01/23/2022 08:15:13 - INFO - codeparrot_training - Step 37758: {'lr': 0.0004902008376920021, 'samples': 906192, 'steps': 6292, 'loss/train': 2.0436618328094482}
+01/23/2022 08:15:14 - INFO - codeparrot_training - Step 37759: {'lr': 0.0004901963009990376, 'samples': 906216, 'steps': 6293, 'loss/train': 3.2099061012268066}
+01/23/2022 08:15:14 - INFO - codeparrot_training - Step 37760: {'lr': 0.0004901963009990376, 'samples': 906240, 'steps': 6293, 'loss/train': 1.7677963972091675}
+01/23/2022 08:15:15 - INFO - codeparrot_training - Step 37761: {'lr': 0.0004901963009990376, 'samples': 906264, 'steps': 6293, 'loss/train': 3.5531985759735107}
+01/23/2022 08:15:15 - INFO - codeparrot_training - Step 37762: {'lr': 0.0004901963009990376, 'samples': 906288, 'steps': 6293, 'loss/train': 2.466416835784912}
+01/23/2022 08:15:15 - INFO - codeparrot_training - Step 37763: {'lr': 0.0004901963009990376, 'samples': 906312, 'steps': 6293, 'loss/train': 2.3980906009674072}
+01/23/2022 08:15:16 - INFO - codeparrot_training - Step 37764: {'lr': 0.0004901963009990376, 'samples': 906336, 'steps': 6293, 'loss/train': 3.4464993476867676}
+01/23/2022 08:15:16 - INFO - codeparrot_training - Step 37765: {'lr': 0.0004901917632771485, 'samples': 906360, 'steps': 6294, 'loss/train': 3.092379570007324}
+01/23/2022 08:15:16 - INFO - codeparrot_training - Step 37766: {'lr': 0.0004901917632771485, 'samples': 906384, 'steps': 6294, 'loss/train': 2.82291579246521}
+01/23/2022 08:15:17 - INFO - codeparrot_training - Step 37767: {'lr': 0.0004901917632771485, 'samples': 906408, 'steps': 6294, 'loss/train': 3.5863561630249023}
+01/23/2022 08:15:17 - INFO - codeparrot_training - Step 37768: {'lr': 0.0004901917632771485, 'samples': 906432, 'steps': 6294, 'loss/train': 7.002708435058594}
+01/23/2022 08:15:18 - INFO - codeparrot_training - Step 37769: {'lr': 0.0004901917632771485, 'samples': 906456, 'steps': 6294, 'loss/train': 3.0623538494110107}
+01/23/2022 08:15:18 - INFO - codeparrot_training - Step 37770: {'lr': 0.0004901917632771485, 'samples': 906480, 'steps': 6294, 'loss/train': 3.381479024887085}
+01/23/2022 08:15:18 - INFO - codeparrot_training - Step 37771: {'lr': 0.000490187224526354, 'samples': 906504, 'steps': 6295, 'loss/train': 3.122671365737915}
+01/23/2022 08:15:19 - INFO - codeparrot_training - Step 37772: {'lr': 0.000490187224526354, 'samples': 906528, 'steps': 6295, 'loss/train': 2.3643813133239746}
+01/23/2022 08:15:19 - INFO - codeparrot_training - Step 37773: {'lr': 0.000490187224526354, 'samples': 906552, 'steps': 6295, 'loss/train': 2.782296895980835}
+01/23/2022 08:15:19 - INFO - codeparrot_training - Step 37774: {'lr': 0.000490187224526354, 'samples': 906576, 'steps': 6295, 'loss/train': 3.4656615257263184}
+01/23/2022 08:15:20 - INFO - codeparrot_training - Step 37775: {'lr': 0.000490187224526354, 'samples': 906600, 'steps': 6295, 'loss/train': 3.4442691802978516}
+01/23/2022 08:15:20 - INFO - codeparrot_training - Step 37776: {'lr': 0.000490187224526354, 'samples': 906624, 'steps': 6295, 'loss/train': 2.8341073989868164}
+01/23/2022 08:15:21 - INFO - codeparrot_training - Step 37777: {'lr': 0.0004901826847466738, 'samples': 906648, 'steps': 6296, 'loss/train': 2.1489787101745605}
+01/23/2022 08:15:21 - INFO - codeparrot_training - Step 37778: {'lr': 0.0004901826847466738, 'samples': 906672, 'steps': 6296, 'loss/train': 4.008424282073975}
+01/23/2022 08:15:21 - INFO - codeparrot_training - Step 37779: {'lr': 0.0004901826847466738, 'samples': 906696, 'steps': 6296, 'loss/train': 4.829866886138916}
+01/23/2022 08:15:22 - INFO - codeparrot_training - Step 37780: {'lr': 0.0004901826847466738, 'samples': 906720, 'steps': 6296, 'loss/train': 3.088489532470703}
+01/23/2022 08:15:22 - INFO - codeparrot_training - Step 37781: {'lr': 0.0004901826847466738, 'samples': 906744, 'steps': 6296, 'loss/train': 4.106000900268555}
+01/23/2022 08:15:23 - INFO - codeparrot_training - Step 37782: {'lr': 0.0004901826847466738, 'samples': 906768, 'steps': 6296, 'loss/train': 3.722447156906128}
+01/23/2022 08:15:23 - INFO - codeparrot_training - Step 37783: {'lr': 0.0004901781439381272, 'samples': 906792, 'steps': 6297, 'loss/train': 4.094018459320068}
+01/23/2022 08:15:23 - INFO - codeparrot_training - Step 37784: {'lr': 0.0004901781439381272, 'samples': 906816, 'steps': 6297, 'loss/train': 2.143143653869629}
+01/23/2022 08:15:24 - INFO - codeparrot_training - Step 37785: {'lr': 0.0004901781439381272, 'samples': 906840, 'steps': 6297, 'loss/train': 2.994565010070801}
+01/23/2022 08:15:24 - INFO - codeparrot_training - Step 37786: {'lr': 0.0004901781439381272, 'samples': 906864, 'steps': 6297, 'loss/train': 3.3487725257873535}
+01/23/2022 08:15:24 - INFO - codeparrot_training - Step 37787: {'lr': 0.0004901781439381272, 'samples': 906888, 'steps': 6297, 'loss/train': 3.292073965072632}
+01/23/2022 08:15:25 - INFO - codeparrot_training - Step 37788: {'lr': 0.0004901781439381272, 'samples': 906912, 'steps': 6297, 'loss/train': 3.3860809803009033}
+01/23/2022 08:15:25 - INFO - codeparrot_training - Step 37789: {'lr': 0.0004901736021007337, 'samples': 906936, 'steps': 6298, 'loss/train': 3.9149580001831055}
+01/23/2022 08:15:26 - INFO - codeparrot_training - Step 37790: {'lr': 0.0004901736021007337, 'samples': 906960, 'steps': 6298, 'loss/train': 3.0163094997406006}
+01/23/2022 08:15:26 - INFO - codeparrot_training - Step 37791: {'lr': 0.0004901736021007337, 'samples': 906984, 'steps': 6298, 'loss/train': 2.590177059173584}
+01/23/2022 08:15:29 - INFO - codeparrot_training - Step 37792: {'lr': 0.0004901736021007337, 'samples': 907008, 'steps': 6298, 'loss/train': 3.424577236175537}
+01/23/2022 08:15:29 - INFO - codeparrot_training - Step 37793: {'lr': 0.0004901736021007337, 'samples': 907032, 'steps': 6298, 'loss/train': 2.8133082389831543}
+01/23/2022 08:15:30 - INFO - codeparrot_training - Step 37794: {'lr': 0.0004901736021007337, 'samples': 907056, 'steps': 6298, 'loss/train': 3.7219905853271484}
+01/23/2022 08:15:30 - INFO - codeparrot_training - Step 37795: {'lr': 0.0004901690592345127, 'samples': 907080, 'steps': 6299, 'loss/train': 3.0178351402282715}
+01/23/2022 08:15:30 - INFO - codeparrot_training - Step 37796: {'lr': 0.0004901690592345127, 'samples': 907104, 'steps': 6299, 'loss/train': 2.467684030532837}
+01/23/2022 08:15:31 - INFO - codeparrot_training - Step 37797: {'lr': 0.0004901690592345127, 'samples': 907128, 'steps': 6299, 'loss/train': 2.9338741302490234}
+01/23/2022 08:15:31 - INFO - codeparrot_training - Step 37798: {'lr': 0.0004901690592345127, 'samples': 907152, 'steps': 6299, 'loss/train': 3.2851860523223877}
+01/23/2022 08:15:32 - INFO - codeparrot_training - Step 37799: {'lr': 0.0004901690592345127, 'samples': 907176, 'steps': 6299, 'loss/train': 2.676635503768921}
+01/23/2022 08:15:32 - INFO - codeparrot_training - Step 37800: {'lr': 0.0004901690592345127, 'samples': 907200, 'steps': 6299, 'loss/train': 2.962184190750122}
+01/23/2022 08:15:32 - INFO - codeparrot_training - Step 37801: {'lr': 0.0004901645153394838, 'samples': 907224, 'steps': 6300, 'loss/train': 2.482883930206299}
+01/23/2022 08:15:33 - INFO - codeparrot_training - Step 37802: {'lr': 0.0004901645153394838, 'samples': 907248, 'steps': 6300, 'loss/train': 2.3962857723236084}
+01/23/2022 08:15:33 - INFO - codeparrot_training - Step 37803: {'lr': 0.0004901645153394838, 'samples': 907272, 'steps': 6300, 'loss/train': 4.041362285614014}
+01/23/2022 08:15:33 - INFO - codeparrot_training - Step 37804: {'lr': 0.0004901645153394838, 'samples': 907296, 'steps': 6300, 'loss/train': 2.1529667377471924}
+01/23/2022 08:15:34 - INFO - codeparrot_training - Step 37805: {'lr': 0.0004901645153394838, 'samples': 907320, 'steps': 6300, 'loss/train': 3.6100635528564453}
+01/23/2022 08:15:34 - INFO - codeparrot_training - Step 37806: {'lr': 0.0004901645153394838, 'samples': 907344, 'steps': 6300, 'loss/train': 3.789048433303833}
+01/23/2022 08:15:35 - INFO - codeparrot_training - Step 37807: {'lr': 0.0004901599704156664, 'samples': 907368, 'steps': 6301, 'loss/train': 3.2524428367614746}
+01/23/2022 08:15:35 - INFO - codeparrot_training - Step 37808: {'lr': 0.0004901599704156664, 'samples': 907392, 'steps': 6301, 'loss/train': 1.995202898979187}
+01/23/2022 08:15:35 - INFO - codeparrot_training - Step 37809: {'lr': 0.0004901599704156664, 'samples': 907416, 'steps': 6301, 'loss/train': 2.8518617153167725}
+01/23/2022 08:15:36 - INFO - codeparrot_training - Step 37810: {'lr': 0.0004901599704156664, 'samples': 907440, 'steps': 6301, 'loss/train': 3.1763718128204346}
+01/23/2022 08:15:36 - INFO - codeparrot_training - Step 37811: {'lr': 0.0004901599704156664, 'samples': 907464, 'steps': 6301, 'loss/train': 3.720886468887329}
+01/23/2022 08:15:37 - INFO - codeparrot_training - Step 37812: {'lr': 0.0004901599704156664, 'samples': 907488, 'steps': 6301, 'loss/train': 0.9417625665664673}
+01/23/2022 08:15:37 - INFO - codeparrot_training - Step 37813: {'lr': 0.00049015542446308, 'samples': 907512, 'steps': 6302, 'loss/train': 4.896289825439453}
+01/23/2022 08:15:37 - INFO - codeparrot_training - Step 37814: {'lr': 0.00049015542446308, 'samples': 907536, 'steps': 6302, 'loss/train': 3.2786614894866943}
+01/23/2022 08:15:38 - INFO - codeparrot_training - Step 37815: {'lr': 0.00049015542446308, 'samples': 907560, 'steps': 6302, 'loss/train': 3.213465929031372}
+01/23/2022 08:15:38 - INFO - codeparrot_training - Step 37816: {'lr': 0.00049015542446308, 'samples': 907584, 'steps': 6302, 'loss/train': 2.865536689758301}
+01/23/2022 08:15:38 - INFO - codeparrot_training - Step 37817: {'lr': 0.00049015542446308, 'samples': 907608, 'steps': 6302, 'loss/train': 3.5013296604156494}
+01/23/2022 08:15:39 - INFO - codeparrot_training - Step 37818: {'lr': 0.00049015542446308, 'samples': 907632, 'steps': 6302, 'loss/train': 3.722856044769287}
+01/23/2022 08:15:39 - INFO - codeparrot_training - Step 37819: {'lr': 0.0004901508774817438, 'samples': 907656, 'steps': 6303, 'loss/train': 3.466797113418579}
+01/23/2022 08:15:40 - INFO - codeparrot_training - Step 37820: {'lr': 0.0004901508774817438, 'samples': 907680, 'steps': 6303, 'loss/train': 3.7158279418945312}
+01/23/2022 08:15:40 - INFO - codeparrot_training - Step 37821: {'lr': 0.0004901508774817438, 'samples': 907704, 'steps': 6303, 'loss/train': 2.92877459526062}
+01/23/2022 08:15:40 - INFO - codeparrot_training - Step 37822: {'lr': 0.0004901508774817438, 'samples': 907728, 'steps': 6303, 'loss/train': 2.506671667098999}
+01/23/2022 08:15:41 - INFO - codeparrot_training - Step 37823: {'lr': 0.0004901508774817438, 'samples': 907752, 'steps': 6303, 'loss/train': 3.627605438232422}
+01/23/2022 08:15:41 - INFO - codeparrot_training - Step 37824: {'lr': 0.0004901508774817438, 'samples': 907776, 'steps': 6303, 'loss/train': 3.7398529052734375}
+01/23/2022 08:15:42 - INFO - codeparrot_training - Step 37825: {'lr': 0.0004901463294716776, 'samples': 907800, 'steps': 6304, 'loss/train': 1.8317930698394775}
+01/23/2022 08:15:42 - INFO - codeparrot_training - Step 37826: {'lr': 0.0004901463294716776, 'samples': 907824, 'steps': 6304, 'loss/train': 3.469005823135376}
+01/23/2022 08:15:42 - INFO - codeparrot_training - Step 37827: {'lr': 0.0004901463294716776, 'samples': 907848, 'steps': 6304, 'loss/train': 4.250149250030518}
+01/23/2022 08:15:43 - INFO - codeparrot_training - Step 37828: {'lr': 0.0004901463294716776, 'samples': 907872, 'steps': 6304, 'loss/train': 4.09989595413208}
+01/23/2022 08:15:43 - INFO - codeparrot_training - Step 37829: {'lr': 0.0004901463294716776, 'samples': 907896, 'steps': 6304, 'loss/train': 3.0345418453216553}
+01/23/2022 08:15:43 - INFO - codeparrot_training - Step 37830: {'lr': 0.0004901463294716776, 'samples': 907920, 'steps': 6304, 'loss/train': 2.6344492435455322}
+01/23/2022 08:15:44 - INFO - codeparrot_training - Step 37831: {'lr': 0.0004901417804329008, 'samples': 907944, 'steps': 6305, 'loss/train': 2.6744728088378906}
+01/23/2022 08:15:45 - INFO - codeparrot_training - Step 37832: {'lr': 0.0004901417804329008, 'samples': 907968, 'steps': 6305, 'loss/train': 4.304341793060303}
+01/23/2022 08:15:45 - INFO - codeparrot_training - Step 37833: {'lr': 0.0004901417804329008, 'samples': 907992, 'steps': 6305, 'loss/train': 3.427698850631714}
+01/23/2022 08:15:46 - INFO - codeparrot_training - Step 37834: {'lr': 0.0004901417804329008, 'samples': 908016, 'steps': 6305, 'loss/train': 2.675285816192627}
+01/23/2022 08:15:46 - INFO - codeparrot_training - Step 37835: {'lr': 0.0004901417804329008, 'samples': 908040, 'steps': 6305, 'loss/train': 3.654247522354126}
+01/23/2022 08:15:47 - INFO - codeparrot_training - Step 37836: {'lr': 0.0004901417804329008, 'samples': 908064, 'steps': 6305, 'loss/train': 2.359389543533325}
+01/23/2022 08:15:47 - INFO - codeparrot_training - Step 37837: {'lr': 0.0004901372303654329, 'samples': 908088, 'steps': 6306, 'loss/train': 3.560960292816162}
+01/23/2022 08:15:47 - INFO - codeparrot_training - Step 37838: {'lr': 0.0004901372303654329, 'samples': 908112, 'steps': 6306, 'loss/train': 4.056179046630859}
+01/23/2022 08:15:48 - INFO - codeparrot_training - Step 37839: {'lr': 0.0004901372303654329, 'samples': 908136, 'steps': 6306, 'loss/train': 3.2272307872772217}
+01/23/2022 08:15:48 - INFO - codeparrot_training - Step 37840: {'lr': 0.0004901372303654329, 'samples': 908160, 'steps': 6306, 'loss/train': 3.438934087753296}
+01/23/2022 08:15:49 - INFO - codeparrot_training - Step 37841: {'lr': 0.0004901372303654329, 'samples': 908184, 'steps': 6306, 'loss/train': 3.0364022254943848}
+01/23/2022 08:15:49 - INFO - codeparrot_training - Step 37842: {'lr': 0.0004901372303654329, 'samples': 908208, 'steps': 6306, 'loss/train': 3.2993569374084473}
+01/23/2022 08:15:49 - INFO - codeparrot_training - Step 37843: {'lr': 0.0004901326792692934, 'samples': 908232, 'steps': 6307, 'loss/train': 4.114577770233154}
+01/23/2022 08:15:50 - INFO - codeparrot_training - Step 37844: {'lr': 0.0004901326792692934, 'samples': 908256, 'steps': 6307, 'loss/train': 1.540542721748352}
+01/23/2022 08:15:50 - INFO - codeparrot_training - Step 37845: {'lr': 0.0004901326792692934, 'samples': 908280, 'steps': 6307, 'loss/train': 3.1390762329101562}
+01/23/2022 08:15:50 - INFO - codeparrot_training - Step 37846: {'lr': 0.0004901326792692934, 'samples': 908304, 'steps': 6307, 'loss/train': 4.246136665344238}
+01/23/2022 08:15:51 - INFO - codeparrot_training - Step 37847: {'lr': 0.0004901326792692934, 'samples': 908328, 'steps': 6307, 'loss/train': 4.867875576019287}
+01/23/2022 08:15:51 - INFO - codeparrot_training - Step 37848: {'lr': 0.0004901326792692934, 'samples': 908352, 'steps': 6307, 'loss/train': 3.0992431640625}
+01/23/2022 08:15:52 - INFO - codeparrot_training - Step 37849: {'lr': 0.0004901281271445016, 'samples': 908376, 'steps': 6308, 'loss/train': 2.3265433311462402}
+01/23/2022 08:15:52 - INFO - codeparrot_training - Step 37850: {'lr': 0.0004901281271445016, 'samples': 908400, 'steps': 6308, 'loss/train': 1.4488728046417236}
+01/23/2022 08:15:52 - INFO - codeparrot_training - Step 37851: {'lr': 0.0004901281271445016, 'samples': 908424, 'steps': 6308, 'loss/train': 1.8176743984222412}
+01/23/2022 08:15:53 - INFO - codeparrot_training - Step 37852: {'lr': 0.0004901281271445016, 'samples': 908448, 'steps': 6308, 'loss/train': 2.769704580307007}
+01/23/2022 08:15:53 - INFO - codeparrot_training - Step 37853: {'lr': 0.0004901281271445016, 'samples': 908472, 'steps': 6308, 'loss/train': 2.1375186443328857}
+01/23/2022 08:15:53 - INFO - codeparrot_training - Step 37854: {'lr': 0.0004901281271445016, 'samples': 908496, 'steps': 6308, 'loss/train': 2.1450002193450928}
+01/23/2022 08:15:54 - INFO - codeparrot_training - Step 37855: {'lr': 0.0004901235739910772, 'samples': 908520, 'steps': 6309, 'loss/train': 2.4978480339050293}
+01/23/2022 08:15:54 - INFO - codeparrot_training - Step 37856: {'lr': 0.0004901235739910772, 'samples': 908544, 'steps': 6309, 'loss/train': 2.3483476638793945}
+01/23/2022 08:15:55 - INFO - codeparrot_training - Step 37857: {'lr': 0.0004901235739910772, 'samples': 908568, 'steps': 6309, 'loss/train': 6.443174362182617}
+01/23/2022 08:15:55 - INFO - codeparrot_training - Step 37858: {'lr': 0.0004901235739910772, 'samples': 908592, 'steps': 6309, 'loss/train': 3.7954163551330566}
+01/23/2022 08:15:55 - INFO - codeparrot_training - Step 37859: {'lr': 0.0004901235739910772, 'samples': 908616, 'steps': 6309, 'loss/train': 3.1861917972564697}
+01/23/2022 08:15:56 - INFO - codeparrot_training - Step 37860: {'lr': 0.0004901235739910772, 'samples': 908640, 'steps': 6309, 'loss/train': 3.1818044185638428}
+01/23/2022 08:15:56 - INFO - codeparrot_training - Step 37861: {'lr': 0.0004901190198090397, 'samples': 908664, 'steps': 6310, 'loss/train': 4.388620853424072}
+01/23/2022 08:15:57 - INFO - codeparrot_training - Step 37862: {'lr': 0.0004901190198090397, 'samples': 908688, 'steps': 6310, 'loss/train': 2.5375514030456543}
+01/23/2022 08:15:57 - INFO - codeparrot_training - Step 37863: {'lr': 0.0004901190198090397, 'samples': 908712, 'steps': 6310, 'loss/train': 3.6910252571105957}
+01/23/2022 08:15:57 - INFO - codeparrot_training - Step 37864: {'lr': 0.0004901190198090397, 'samples': 908736, 'steps': 6310, 'loss/train': 3.6255314350128174}
+01/23/2022 08:15:58 - INFO - codeparrot_training - Step 37865: {'lr': 0.0004901190198090397, 'samples': 908760, 'steps': 6310, 'loss/train': 2.7635951042175293}
+01/23/2022 08:15:58 - INFO - codeparrot_training - Step 37866: {'lr': 0.0004901190198090397, 'samples': 908784, 'steps': 6310, 'loss/train': 3.3868486881256104}
+01/23/2022 08:15:59 - INFO - codeparrot_training - Step 37867: {'lr': 0.0004901144645984086, 'samples': 908808, 'steps': 6311, 'loss/train': 2.558995246887207}
+01/23/2022 08:15:59 - INFO - codeparrot_training - Step 37868: {'lr': 0.0004901144645984086, 'samples': 908832, 'steps': 6311, 'loss/train': 2.7170701026916504}
+01/23/2022 08:15:59 - INFO - codeparrot_training - Step 37869: {'lr': 0.0004901144645984086, 'samples': 908856, 'steps': 6311, 'loss/train': 2.0565133094787598}
+01/23/2022 08:16:00 - INFO - codeparrot_training - Step 37870: {'lr': 0.0004901144645984086, 'samples': 908880, 'steps': 6311, 'loss/train': 2.8391451835632324}
+01/23/2022 08:16:01 - INFO - codeparrot_training - Step 37871: {'lr': 0.0004901144645984086, 'samples': 908904, 'steps': 6311, 'loss/train': 3.101440668106079}
+01/23/2022 08:16:01 - INFO - codeparrot_training - Step 37872: {'lr': 0.0004901144645984086, 'samples': 908928, 'steps': 6311, 'loss/train': 1.8022412061691284}
+01/23/2022 08:16:02 - INFO - codeparrot_training - Step 37873: {'lr': 0.0004901099083592034, 'samples': 908952, 'steps': 6312, 'loss/train': 3.395784854888916}
+01/23/2022 08:16:02 - INFO - codeparrot_training - Step 37874: {'lr': 0.0004901099083592034, 'samples': 908976, 'steps': 6312, 'loss/train': 2.048429489135742}
+01/23/2022 08:16:02 - INFO - codeparrot_training - Step 37875: {'lr': 0.0004901099083592034, 'samples': 909000, 'steps': 6312, 'loss/train': 4.073262691497803}
+01/23/2022 08:16:03 - INFO - codeparrot_training - Step 37876: {'lr': 0.0004901099083592034, 'samples': 909024, 'steps': 6312, 'loss/train': 4.479583740234375}
+01/23/2022 08:16:03 - INFO - codeparrot_training - Step 37877: {'lr': 0.0004901099083592034, 'samples': 909048, 'steps': 6312, 'loss/train': 3.9642202854156494}
+01/23/2022 08:16:03 - INFO - codeparrot_training - Step 37878: {'lr': 0.0004901099083592034, 'samples': 909072, 'steps': 6312, 'loss/train': 4.094627380371094}
+01/23/2022 08:16:04 - INFO - codeparrot_training - Step 37879: {'lr': 0.0004901053510914434, 'samples': 909096, 'steps': 6313, 'loss/train': 3.68190598487854}
+01/23/2022 08:16:04 - INFO - codeparrot_training - Step 37880: {'lr': 0.0004901053510914434, 'samples': 909120, 'steps': 6313, 'loss/train': 3.3150296211242676}
+01/23/2022 08:16:05 - INFO - codeparrot_training - Step 37881: {'lr': 0.0004901053510914434, 'samples': 909144, 'steps': 6313, 'loss/train': 3.242466449737549}
+01/23/2022 08:16:05 - INFO - codeparrot_training - Step 37882: {'lr': 0.0004901053510914434, 'samples': 909168, 'steps': 6313, 'loss/train': 3.2760732173919678}
+01/23/2022 08:16:05 - INFO - codeparrot_training - Step 37883: {'lr': 0.0004901053510914434, 'samples': 909192, 'steps': 6313, 'loss/train': 3.1222784519195557}
+01/23/2022 08:16:06 - INFO - codeparrot_training - Step 37884: {'lr': 0.0004901053510914434, 'samples': 909216, 'steps': 6313, 'loss/train': 3.5196588039398193}
+01/23/2022 08:16:06 - INFO - codeparrot_training - Step 37885: {'lr': 0.0004901007927951485, 'samples': 909240, 'steps': 6314, 'loss/train': 3.0679311752319336}
+01/23/2022 08:16:07 - INFO - codeparrot_training - Step 37886: {'lr': 0.0004901007927951485, 'samples': 909264, 'steps': 6314, 'loss/train': 3.639246940612793}
+01/23/2022 08:16:07 - INFO - codeparrot_training - Step 37887: {'lr': 0.0004901007927951485, 'samples': 909288, 'steps': 6314, 'loss/train': 2.9806320667266846}
+01/23/2022 08:16:07 - INFO - codeparrot_training - Step 37888: {'lr': 0.0004901007927951485, 'samples': 909312, 'steps': 6314, 'loss/train': 2.6302311420440674}
+01/23/2022 08:16:08 - INFO - codeparrot_training - Step 37889: {'lr': 0.0004901007927951485, 'samples': 909336, 'steps': 6314, 'loss/train': 3.1323890686035156}
+01/23/2022 08:16:08 - INFO - codeparrot_training - Step 37890: {'lr': 0.0004901007927951485, 'samples': 909360, 'steps': 6314, 'loss/train': 3.2510933876037598}
+01/23/2022 08:16:09 - INFO - codeparrot_training - Step 37891: {'lr': 0.000490096233470338, 'samples': 909384, 'steps': 6315, 'loss/train': 3.1682019233703613}
+01/23/2022 08:16:09 - INFO - codeparrot_training - Step 37892: {'lr': 0.000490096233470338, 'samples': 909408, 'steps': 6315, 'loss/train': 3.2838034629821777}
+01/23/2022 08:16:09 - INFO - codeparrot_training - Step 37893: {'lr': 0.000490096233470338, 'samples': 909432, 'steps': 6315, 'loss/train': 3.5121636390686035}
+01/23/2022 08:16:10 - INFO - codeparrot_training - Step 37894: {'lr': 0.000490096233470338, 'samples': 909456, 'steps': 6315, 'loss/train': 2.4874212741851807}
+01/23/2022 08:16:10 - INFO - codeparrot_training - Step 37895: {'lr': 0.000490096233470338, 'samples': 909480, 'steps': 6315, 'loss/train': 3.6878767013549805}
+01/23/2022 08:16:10 - INFO - codeparrot_training - Step 37896: {'lr': 0.000490096233470338, 'samples': 909504, 'steps': 6315, 'loss/train': 3.30155611038208}
+01/23/2022 08:16:11 - INFO - codeparrot_training - Step 37897: {'lr': 0.0004900916731170314, 'samples': 909528, 'steps': 6316, 'loss/train': 1.8761910200119019}
+01/23/2022 08:16:11 - INFO - codeparrot_training - Step 37898: {'lr': 0.0004900916731170314, 'samples': 909552, 'steps': 6316, 'loss/train': 3.1248323917388916}
+01/23/2022 08:16:12 - INFO - codeparrot_training - Step 37899: {'lr': 0.0004900916731170314, 'samples': 909576, 'steps': 6316, 'loss/train': 2.2367300987243652}
+01/23/2022 08:16:12 - INFO - codeparrot_training - Step 37900: {'lr': 0.0004900916731170314, 'samples': 909600, 'steps': 6316, 'loss/train': 3.1830153465270996}
+01/23/2022 08:16:12 - INFO - codeparrot_training - Step 37901: {'lr': 0.0004900916731170314, 'samples': 909624, 'steps': 6316, 'loss/train': 2.971874952316284}
+01/23/2022 08:16:13 - INFO - codeparrot_training - Step 37902: {'lr': 0.0004900916731170314, 'samples': 909648, 'steps': 6316, 'loss/train': 4.968705654144287}
+01/23/2022 08:16:13 - INFO - codeparrot_training - Step 37903: {'lr': 0.0004900871117352484, 'samples': 909672, 'steps': 6317, 'loss/train': 3.2690351009368896}
+01/23/2022 08:16:13 - INFO - codeparrot_training - Step 37904: {'lr': 0.0004900871117352484, 'samples': 909696, 'steps': 6317, 'loss/train': 3.3816094398498535}
+01/23/2022 08:16:14 - INFO - codeparrot_training - Step 37905: {'lr': 0.0004900871117352484, 'samples': 909720, 'steps': 6317, 'loss/train': 0.4824896454811096}
+01/23/2022 08:16:14 - INFO - codeparrot_training - Step 37906: {'lr': 0.0004900871117352484, 'samples': 909744, 'steps': 6317, 'loss/train': 3.6622579097747803}
+01/23/2022 08:16:15 - INFO - codeparrot_training - Step 37907: {'lr': 0.0004900871117352484, 'samples': 909768, 'steps': 6317, 'loss/train': 3.1135730743408203}
+01/23/2022 08:16:15 - INFO - codeparrot_training - Step 37908: {'lr': 0.0004900871117352484, 'samples': 909792, 'steps': 6317, 'loss/train': 3.6842784881591797}
+01/23/2022 08:16:15 - INFO - codeparrot_training - Step 37909: {'lr': 0.0004900825493250084, 'samples': 909816, 'steps': 6318, 'loss/train': 2.507291316986084}
+01/23/2022 08:16:16 - INFO - codeparrot_training - Step 37910: {'lr': 0.0004900825493250084, 'samples': 909840, 'steps': 6318, 'loss/train': 3.8957862854003906}
+01/23/2022 08:16:16 - INFO - codeparrot_training - Step 37911: {'lr': 0.0004900825493250084, 'samples': 909864, 'steps': 6318, 'loss/train': 3.501027822494507}
+01/23/2022 08:16:17 - INFO - codeparrot_training - Step 37912: {'lr': 0.0004900825493250084, 'samples': 909888, 'steps': 6318, 'loss/train': 3.393113613128662}
+01/23/2022 08:16:17 - INFO - codeparrot_training - Step 37913: {'lr': 0.0004900825493250084, 'samples': 909912, 'steps': 6318, 'loss/train': 2.529006242752075}
+01/23/2022 08:16:18 - INFO - codeparrot_training - Step 37914: {'lr': 0.0004900825493250084, 'samples': 909936, 'steps': 6318, 'loss/train': 2.9598231315612793}
+01/23/2022 08:16:18 - INFO - codeparrot_training - Step 37915: {'lr': 0.000490077985886331, 'samples': 909960, 'steps': 6319, 'loss/train': 3.0205085277557373}
+01/23/2022 08:16:19 - INFO - codeparrot_training - Step 37916: {'lr': 0.000490077985886331, 'samples': 909984, 'steps': 6319, 'loss/train': 3.8287391662597656}
+01/23/2022 08:16:19 - INFO - codeparrot_training - Step 37917: {'lr': 0.000490077985886331, 'samples': 910008, 'steps': 6319, 'loss/train': 2.1230969429016113}
+01/23/2022 08:16:20 - INFO - codeparrot_training - Step 37918: {'lr': 0.000490077985886331, 'samples': 910032, 'steps': 6319, 'loss/train': 3.877682685852051}
+01/23/2022 08:16:20 - INFO - codeparrot_training - Step 37919: {'lr': 0.000490077985886331, 'samples': 910056, 'steps': 6319, 'loss/train': 2.028245449066162}
+01/23/2022 08:16:20 - INFO - codeparrot_training - Step 37920: {'lr': 0.000490077985886331, 'samples': 910080, 'steps': 6319, 'loss/train': 4.3738250732421875}
+01/23/2022 08:16:21 - INFO - codeparrot_training - Step 37921: {'lr': 0.0004900734214192358, 'samples': 910104, 'steps': 6320, 'loss/train': 2.222315549850464}
+01/23/2022 08:16:21 - INFO - codeparrot_training - Step 37922: {'lr': 0.0004900734214192358, 'samples': 910128, 'steps': 6320, 'loss/train': 3.0041308403015137}
+01/23/2022 08:16:21 - INFO - codeparrot_training - Step 37923: {'lr': 0.0004900734214192358, 'samples': 910152, 'steps': 6320, 'loss/train': 3.6765401363372803}
+01/23/2022 08:16:22 - INFO - codeparrot_training - Step 37924: {'lr': 0.0004900734214192358, 'samples': 910176, 'steps': 6320, 'loss/train': 2.2370986938476562}
+01/23/2022 08:16:22 - INFO - codeparrot_training - Step 37925: {'lr': 0.0004900734214192358, 'samples': 910200, 'steps': 6320, 'loss/train': 3.387791156768799}
+01/23/2022 08:16:23 - INFO - codeparrot_training - Step 37926: {'lr': 0.0004900734214192358, 'samples': 910224, 'steps': 6320, 'loss/train': 2.3558874130249023}
+01/23/2022 08:16:23 - INFO - codeparrot_training - Step 37927: {'lr': 0.0004900688559237422, 'samples': 910248, 'steps': 6321, 'loss/train': 3.581258535385132}
+01/23/2022 08:16:23 - INFO - codeparrot_training - Step 37928: {'lr': 0.0004900688559237422, 'samples': 910272, 'steps': 6321, 'loss/train': 4.206496715545654}
+01/23/2022 08:16:24 - INFO - codeparrot_training - Step 37929: {'lr': 0.0004900688559237422, 'samples': 910296, 'steps': 6321, 'loss/train': 2.674125909805298}
+01/23/2022 08:16:24 - INFO - codeparrot_training - Step 37930: {'lr': 0.0004900688559237422, 'samples': 910320, 'steps': 6321, 'loss/train': 3.383434295654297}
+01/23/2022 08:16:24 - INFO - codeparrot_training - Step 37931: {'lr': 0.0004900688559237422, 'samples': 910344, 'steps': 6321, 'loss/train': 3.9695990085601807}
+01/23/2022 08:16:25 - INFO - codeparrot_training - Step 37932: {'lr': 0.0004900688559237422, 'samples': 910368, 'steps': 6321, 'loss/train': 3.37210750579834}
+01/23/2022 08:16:25 - INFO - codeparrot_training - Step 37933: {'lr': 0.0004900642893998699, 'samples': 910392, 'steps': 6322, 'loss/train': 2.6676084995269775}
+01/23/2022 08:16:26 - INFO - codeparrot_training - Step 37934: {'lr': 0.0004900642893998699, 'samples': 910416, 'steps': 6322, 'loss/train': 2.8927083015441895}
+01/23/2022 08:16:26 - INFO - codeparrot_training - Step 37935: {'lr': 0.0004900642893998699, 'samples': 910440, 'steps': 6322, 'loss/train': 3.3033337593078613}
+01/23/2022 08:16:26 - INFO - codeparrot_training - Step 37936: {'lr': 0.0004900642893998699, 'samples': 910464, 'steps': 6322, 'loss/train': 2.2671427726745605}
+01/23/2022 08:16:27 - INFO - codeparrot_training - Step 37937: {'lr': 0.0004900642893998699, 'samples': 910488, 'steps': 6322, 'loss/train': 0.8688259720802307}
+01/23/2022 08:16:27 - INFO - codeparrot_training - Step 37938: {'lr': 0.0004900642893998699, 'samples': 910512, 'steps': 6322, 'loss/train': 2.5068519115448}
+01/23/2022 08:16:28 - INFO - codeparrot_training - Step 37939: {'lr': 0.0004900597218476385, 'samples': 910536, 'steps': 6323, 'loss/train': 2.9157047271728516}
+01/23/2022 08:16:28 - INFO - codeparrot_training - Step 37940: {'lr': 0.0004900597218476385, 'samples': 910560, 'steps': 6323, 'loss/train': 1.7237811088562012}
+01/23/2022 08:16:28 - INFO - codeparrot_training - Step 37941: {'lr': 0.0004900597218476385, 'samples': 910584, 'steps': 6323, 'loss/train': 1.368634581565857}
+01/23/2022 08:16:29 - INFO - codeparrot_training - Step 37942: {'lr': 0.0004900597218476385, 'samples': 910608, 'steps': 6323, 'loss/train': 4.113974094390869}
+01/23/2022 08:16:29 - INFO - codeparrot_training - Step 37943: {'lr': 0.0004900597218476385, 'samples': 910632, 'steps': 6323, 'loss/train': 3.104140043258667}
+01/23/2022 08:16:29 - INFO - codeparrot_training - Step 37944: {'lr': 0.0004900597218476385, 'samples': 910656, 'steps': 6323, 'loss/train': 3.739436388015747}
+01/23/2022 08:16:30 - INFO - codeparrot_training - Step 37945: {'lr': 0.0004900551532670673, 'samples': 910680, 'steps': 6324, 'loss/train': 2.722560405731201}
+01/23/2022 08:16:30 - INFO - codeparrot_training - Step 37946: {'lr': 0.0004900551532670673, 'samples': 910704, 'steps': 6324, 'loss/train': 2.9516565799713135}
+01/23/2022 08:16:31 - INFO - codeparrot_training - Step 37947: {'lr': 0.0004900551532670673, 'samples': 910728, 'steps': 6324, 'loss/train': 3.0801284313201904}
+01/23/2022 08:16:31 - INFO - codeparrot_training - Step 37948: {'lr': 0.0004900551532670673, 'samples': 910752, 'steps': 6324, 'loss/train': 2.7784552574157715}
+01/23/2022 08:16:31 - INFO - codeparrot_training - Step 37949: {'lr': 0.0004900551532670673, 'samples': 910776, 'steps': 6324, 'loss/train': 3.5156238079071045}
+01/23/2022 08:16:32 - INFO - codeparrot_training - Step 37950: {'lr': 0.0004900551532670673, 'samples': 910800, 'steps': 6324, 'loss/train': 2.5909926891326904}
+01/23/2022 08:16:32 - INFO - codeparrot_training - Step 37951: {'lr': 0.0004900505836581763, 'samples': 910824, 'steps': 6325, 'loss/train': 2.5858850479125977}
+01/23/2022 08:16:33 - INFO - codeparrot_training - Step 37952: {'lr': 0.0004900505836581763, 'samples': 910848, 'steps': 6325, 'loss/train': 3.5999841690063477}
+01/23/2022 08:16:33 - INFO - codeparrot_training - Step 37953: {'lr': 0.0004900505836581763, 'samples': 910872, 'steps': 6325, 'loss/train': 4.200406551361084}
+01/23/2022 08:16:33 - INFO - codeparrot_training - Step 37954: {'lr': 0.0004900505836581763, 'samples': 910896, 'steps': 6325, 'loss/train': 3.1443018913269043}
+01/23/2022 08:16:34 - INFO - codeparrot_training - Step 37955: {'lr': 0.0004900505836581763, 'samples': 910920, 'steps': 6325, 'loss/train': 3.7102556228637695}
+01/23/2022 08:16:35 - INFO - codeparrot_training - Step 37956: {'lr': 0.0004900505836581763, 'samples': 910944, 'steps': 6325, 'loss/train': 2.945425510406494}
+01/23/2022 08:16:35 - INFO - codeparrot_training - Step 37957: {'lr': 0.0004900460130209845, 'samples': 910968, 'steps': 6326, 'loss/train': 3.1119697093963623}
+01/23/2022 08:16:35 - INFO - codeparrot_training - Step 37958: {'lr': 0.0004900460130209845, 'samples': 910992, 'steps': 6326, 'loss/train': 3.8559176921844482}
+01/23/2022 08:16:36 - INFO - codeparrot_training - Step 37959: {'lr': 0.0004900460130209845, 'samples': 911016, 'steps': 6326, 'loss/train': 1.465787649154663}
+01/23/2022 08:16:36 - INFO - codeparrot_training - Step 37960: {'lr': 0.0004900460130209845, 'samples': 911040, 'steps': 6326, 'loss/train': 1.4094291925430298}
+01/23/2022 08:16:37 - INFO - codeparrot_training - Step 37961: {'lr': 0.0004900460130209845, 'samples': 911064, 'steps': 6326, 'loss/train': 3.3478808403015137}
+01/23/2022 08:16:37 - INFO - codeparrot_training - Step 37962: {'lr': 0.0004900460130209845, 'samples': 911088, 'steps': 6326, 'loss/train': 3.3907601833343506}
+01/23/2022 08:16:37 - INFO - codeparrot_training - Step 37963: {'lr': 0.000490041441355512, 'samples': 911112, 'steps': 6327, 'loss/train': 3.3859031200408936}
+01/23/2022 08:16:38 - INFO - codeparrot_training - Step 37964: {'lr': 0.000490041441355512, 'samples': 911136, 'steps': 6327, 'loss/train': 3.080981492996216}
+01/23/2022 08:16:38 - INFO - codeparrot_training - Step 37965: {'lr': 0.000490041441355512, 'samples': 911160, 'steps': 6327, 'loss/train': 2.201702356338501}
+01/23/2022 08:16:38 - INFO - codeparrot_training - Step 37966: {'lr': 0.000490041441355512, 'samples': 911184, 'steps': 6327, 'loss/train': 3.983506917953491}
+01/23/2022 08:16:39 - INFO - codeparrot_training - Step 37967: {'lr': 0.000490041441355512, 'samples': 911208, 'steps': 6327, 'loss/train': 3.0147576332092285}
+01/23/2022 08:16:39 - INFO - codeparrot_training - Step 37968: {'lr': 0.000490041441355512, 'samples': 911232, 'steps': 6327, 'loss/train': 3.5892343521118164}
+01/23/2022 08:16:40 - INFO - codeparrot_training - Step 37969: {'lr': 0.0004900368686617783, 'samples': 911256, 'steps': 6328, 'loss/train': 3.4437918663024902}
+01/23/2022 08:16:40 - INFO - codeparrot_training - Step 37970: {'lr': 0.0004900368686617783, 'samples': 911280, 'steps': 6328, 'loss/train': 2.86061954498291}
+01/23/2022 08:16:40 - INFO - codeparrot_training - Step 37971: {'lr': 0.0004900368686617783, 'samples': 911304, 'steps': 6328, 'loss/train': 1.145158290863037}
+01/23/2022 08:16:41 - INFO - codeparrot_training - Step 37972: {'lr': 0.0004900368686617783, 'samples': 911328, 'steps': 6328, 'loss/train': 3.4107208251953125}
+01/23/2022 08:16:41 - INFO - codeparrot_training - Step 37973: {'lr': 0.0004900368686617783, 'samples': 911352, 'steps': 6328, 'loss/train': 7.196609973907471}
+01/23/2022 08:16:41 - INFO - codeparrot_training - Step 37974: {'lr': 0.0004900368686617783, 'samples': 911376, 'steps': 6328, 'loss/train': 2.731614351272583}
+01/23/2022 08:16:42 - INFO - codeparrot_training - Step 37975: {'lr': 0.0004900322949398026, 'samples': 911400, 'steps': 6329, 'loss/train': 1.289086103439331}
+01/23/2022 08:16:42 - INFO - codeparrot_training - Step 37976: {'lr': 0.0004900322949398026, 'samples': 911424, 'steps': 6329, 'loss/train': 3.4267001152038574}
+01/23/2022 08:16:43 - INFO - codeparrot_training - Step 37977: {'lr': 0.0004900322949398026, 'samples': 911448, 'steps': 6329, 'loss/train': 4.500160217285156}
+01/23/2022 08:16:43 - INFO - codeparrot_training - Step 37978: {'lr': 0.0004900322949398026, 'samples': 911472, 'steps': 6329, 'loss/train': 3.4432060718536377}
+01/23/2022 08:16:43 - INFO - codeparrot_training - Step 37979: {'lr': 0.0004900322949398026, 'samples': 911496, 'steps': 6329, 'loss/train': 3.9533348083496094}
+01/23/2022 08:16:44 - INFO - codeparrot_training - Step 37980: {'lr': 0.0004900322949398026, 'samples': 911520, 'steps': 6329, 'loss/train': 2.766796588897705}
+01/23/2022 08:16:44 - INFO - codeparrot_training - Step 37981: {'lr': 0.000490027720189605, 'samples': 911544, 'steps': 6330, 'loss/train': 4.292866230010986}
+01/23/2022 08:16:45 - INFO - codeparrot_training - Step 37982: {'lr': 0.000490027720189605, 'samples': 911568, 'steps': 6330, 'loss/train': 4.108495235443115}
+01/23/2022 08:16:45 - INFO - codeparrot_training - Step 37983: {'lr': 0.000490027720189605, 'samples': 911592, 'steps': 6330, 'loss/train': 3.2395403385162354}
+01/23/2022 08:16:45 - INFO - codeparrot_training - Step 37984: {'lr': 0.000490027720189605, 'samples': 911616, 'steps': 6330, 'loss/train': 4.172255992889404}
+01/23/2022 08:16:46 - INFO - codeparrot_training - Step 37985: {'lr': 0.000490027720189605, 'samples': 911640, 'steps': 6330, 'loss/train': 3.6618399620056152}
+01/23/2022 08:16:46 - INFO - codeparrot_training - Step 37986: {'lr': 0.000490027720189605, 'samples': 911664, 'steps': 6330, 'loss/train': 3.4342422485351562}
+01/23/2022 08:16:47 - INFO - codeparrot_training - Step 37987: {'lr': 0.0004900231444112047, 'samples': 911688, 'steps': 6331, 'loss/train': 3.4557855129241943}
+01/23/2022 08:16:47 - INFO - codeparrot_training - Step 37988: {'lr': 0.0004900231444112047, 'samples': 911712, 'steps': 6331, 'loss/train': 3.9256443977355957}
+01/23/2022 08:16:47 - INFO - codeparrot_training - Step 37989: {'lr': 0.0004900231444112047, 'samples': 911736, 'steps': 6331, 'loss/train': 2.203688383102417}
+01/23/2022 08:16:48 - INFO - codeparrot_training - Step 37990: {'lr': 0.0004900231444112047, 'samples': 911760, 'steps': 6331, 'loss/train': 4.314542770385742}
+01/23/2022 08:16:48 - INFO - codeparrot_training - Step 37991: {'lr': 0.0004900231444112047, 'samples': 911784, 'steps': 6331, 'loss/train': 2.4993748664855957}
+01/23/2022 08:16:48 - INFO - codeparrot_training - Step 37992: {'lr': 0.0004900231444112047, 'samples': 911808, 'steps': 6331, 'loss/train': 3.65020489692688}
+01/23/2022 08:16:49 - INFO - codeparrot_training - Step 37993: {'lr': 0.0004900185676046214, 'samples': 911832, 'steps': 6332, 'loss/train': 3.7382631301879883}
+01/23/2022 08:16:50 - INFO - codeparrot_training - Step 37994: {'lr': 0.0004900185676046214, 'samples': 911856, 'steps': 6332, 'loss/train': 2.8767716884613037}
+01/23/2022 08:16:50 - INFO - codeparrot_training - Step 37995: {'lr': 0.0004900185676046214, 'samples': 911880, 'steps': 6332, 'loss/train': 3.575593948364258}
+01/23/2022 08:16:51 - INFO - codeparrot_training - Step 37996: {'lr': 0.0004900185676046214, 'samples': 911904, 'steps': 6332, 'loss/train': 2.567988157272339}
+01/23/2022 08:16:51 - INFO - codeparrot_training - Step 37997: {'lr': 0.0004900185676046214, 'samples': 911928, 'steps': 6332, 'loss/train': 2.341730833053589}
+01/23/2022 08:16:51 - INFO - codeparrot_training - Step 37998: {'lr': 0.0004900185676046214, 'samples': 911952, 'steps': 6332, 'loss/train': 4.445226669311523}
+01/23/2022 08:16:52 - INFO - codeparrot_training - Step 37999: {'lr': 0.0004900139897698751, 'samples': 911976, 'steps': 6333, 'loss/train': 3.0266666412353516}
+01/23/2022 08:16:52 - INFO - codeparrot_training - Step 38000: {'lr': 0.0004900139897698751, 'samples': 912000, 'steps': 6333, 'loss/train': 2.403864622116089}
+01/23/2022 08:16:52 - INFO - codeparrot_training - Step 38001: {'lr': 0.0004900139897698751, 'samples': 912024, 'steps': 6333, 'loss/train': 3.0328197479248047}
+01/23/2022 08:16:53 - INFO - codeparrot_training - Step 38002: {'lr': 0.0004900139897698751, 'samples': 912048, 'steps': 6333, 'loss/train': 3.212061643600464}
+01/23/2022 08:16:53 - INFO - codeparrot_training - Step 38003: {'lr': 0.0004900139897698751, 'samples': 912072, 'steps': 6333, 'loss/train': 2.9308104515075684}
+01/23/2022 08:16:54 - INFO - codeparrot_training - Step 38004: {'lr': 0.0004900139897698751, 'samples': 912096, 'steps': 6333, 'loss/train': 4.1667938232421875}
+01/23/2022 08:16:54 - INFO - codeparrot_training - Step 38005: {'lr': 0.0004900094109069848, 'samples': 912120, 'steps': 6334, 'loss/train': 3.5089457035064697}
+01/23/2022 08:16:54 - INFO - codeparrot_training - Step 38006: {'lr': 0.0004900094109069848, 'samples': 912144, 'steps': 6334, 'loss/train': 3.222858190536499}
+01/23/2022 08:16:55 - INFO - codeparrot_training - Step 38007: {'lr': 0.0004900094109069848, 'samples': 912168, 'steps': 6334, 'loss/train': 3.6566672325134277}
+01/23/2022 08:16:55 - INFO - codeparrot_training - Step 38008: {'lr': 0.0004900094109069848, 'samples': 912192, 'steps': 6334, 'loss/train': 1.1971683502197266}
+01/23/2022 08:16:56 - INFO - codeparrot_training - Step 38009: {'lr': 0.0004900094109069848, 'samples': 912216, 'steps': 6334, 'loss/train': 3.7623701095581055}
+01/23/2022 08:16:56 - INFO - codeparrot_training - Step 38010: {'lr': 0.0004900094109069848, 'samples': 912240, 'steps': 6334, 'loss/train': 2.0148978233337402}
+01/23/2022 08:16:56 - INFO - codeparrot_training - Step 38011: {'lr': 0.0004900048310159705, 'samples': 912264, 'steps': 6335, 'loss/train': 3.2431232929229736}
+01/23/2022 08:16:57 - INFO - codeparrot_training - Step 38012: {'lr': 0.0004900048310159705, 'samples': 912288, 'steps': 6335, 'loss/train': 1.8942197561264038}
+01/23/2022 08:16:57 - INFO - codeparrot_training - Step 38013: {'lr': 0.0004900048310159705, 'samples': 912312, 'steps': 6335, 'loss/train': 2.970069169998169}
+01/23/2022 08:16:57 - INFO - codeparrot_training - Step 38014: {'lr': 0.0004900048310159705, 'samples': 912336, 'steps': 6335, 'loss/train': 2.264730453491211}
+01/23/2022 08:16:58 - INFO - codeparrot_training - Step 38015: {'lr': 0.0004900048310159705, 'samples': 912360, 'steps': 6335, 'loss/train': 2.7072103023529053}
+01/23/2022 08:16:58 - INFO - codeparrot_training - Step 38016: {'lr': 0.0004900048310159705, 'samples': 912384, 'steps': 6335, 'loss/train': 1.7201088666915894}
+01/23/2022 08:16:59 - INFO - codeparrot_training - Step 38017: {'lr': 0.0004900002500968516, 'samples': 912408, 'steps': 6336, 'loss/train': 3.444275379180908}
+01/23/2022 08:16:59 - INFO - codeparrot_training - Step 38018: {'lr': 0.0004900002500968516, 'samples': 912432, 'steps': 6336, 'loss/train': 2.796776294708252}
+01/23/2022 08:16:59 - INFO - codeparrot_training - Step 38019: {'lr': 0.0004900002500968516, 'samples': 912456, 'steps': 6336, 'loss/train': 2.611842393875122}
+01/23/2022 08:17:00 - INFO - codeparrot_training - Step 38020: {'lr': 0.0004900002500968516, 'samples': 912480, 'steps': 6336, 'loss/train': 3.666273593902588}
+01/23/2022 08:17:00 - INFO - codeparrot_training - Step 38021: {'lr': 0.0004900002500968516, 'samples': 912504, 'steps': 6336, 'loss/train': 4.004284858703613}
+01/23/2022 08:17:01 - INFO - codeparrot_training - Step 38022: {'lr': 0.0004900002500968516, 'samples': 912528, 'steps': 6336, 'loss/train': 3.315751075744629}
+01/23/2022 08:17:01 - INFO - codeparrot_training - Step 38023: {'lr': 0.000489995668149648, 'samples': 912552, 'steps': 6337, 'loss/train': 2.80733323097229}
+01/23/2022 08:17:01 - INFO - codeparrot_training - Step 38024: {'lr': 0.000489995668149648, 'samples': 912576, 'steps': 6337, 'loss/train': 3.4687082767486572}
+01/23/2022 08:17:02 - INFO - codeparrot_training - Step 38025: {'lr': 0.000489995668149648, 'samples': 912600, 'steps': 6337, 'loss/train': 3.1189181804656982}
+01/23/2022 08:17:02 - INFO - codeparrot_training - Step 38026: {'lr': 0.000489995668149648, 'samples': 912624, 'steps': 6337, 'loss/train': 3.271352529525757}
+01/23/2022 08:17:02 - INFO - codeparrot_training - Step 38027: {'lr': 0.000489995668149648, 'samples': 912648, 'steps': 6337, 'loss/train': 4.594178199768066}
+01/23/2022 08:17:03 - INFO - codeparrot_training - Step 38028: {'lr': 0.000489995668149648, 'samples': 912672, 'steps': 6337, 'loss/train': 1.7781087160110474}
+01/23/2022 08:17:03 - INFO - codeparrot_training - Step 38029: {'lr': 0.0004899910851743791, 'samples': 912696, 'steps': 6338, 'loss/train': 2.9824485778808594}
+01/23/2022 08:17:04 - INFO - codeparrot_training - Step 38030: {'lr': 0.0004899910851743791, 'samples': 912720, 'steps': 6338, 'loss/train': 3.5676493644714355}
+01/23/2022 08:17:04 - INFO - codeparrot_training - Step 38031: {'lr': 0.0004899910851743791, 'samples': 912744, 'steps': 6338, 'loss/train': 2.2218165397644043}
+01/23/2022 08:17:04 - INFO - codeparrot_training - Step 38032: {'lr': 0.0004899910851743791, 'samples': 912768, 'steps': 6338, 'loss/train': 1.769845962524414}
+01/23/2022 08:17:05 - INFO - codeparrot_training - Step 38033: {'lr': 0.0004899910851743791, 'samples': 912792, 'steps': 6338, 'loss/train': 1.9897469282150269}
+01/23/2022 08:17:06 - INFO - codeparrot_training - Step 38034: {'lr': 0.0004899910851743791, 'samples': 912816, 'steps': 6338, 'loss/train': 3.2920007705688477}
+01/23/2022 08:17:06 - INFO - codeparrot_training - Step 38035: {'lr': 0.0004899865011710646, 'samples': 912840, 'steps': 6339, 'loss/train': 2.99180269241333}
+01/23/2022 08:17:07 - INFO - codeparrot_training - Step 38036: {'lr': 0.0004899865011710646, 'samples': 912864, 'steps': 6339, 'loss/train': 3.7646214962005615}
+01/23/2022 08:17:07 - INFO - codeparrot_training - Step 38037: {'lr': 0.0004899865011710646, 'samples': 912888, 'steps': 6339, 'loss/train': 3.370265245437622}
+01/23/2022 08:17:08 - INFO - codeparrot_training - Step 38038: {'lr': 0.0004899865011710646, 'samples': 912912, 'steps': 6339, 'loss/train': 3.6290576457977295}
+01/23/2022 08:17:08 - INFO - codeparrot_training - Step 38039: {'lr': 0.0004899865011710646, 'samples': 912936, 'steps': 6339, 'loss/train': 3.3535025119781494}
+01/23/2022 08:17:08 - INFO - codeparrot_training - Step 38040: {'lr': 0.0004899865011710646, 'samples': 912960, 'steps': 6339, 'loss/train': 3.9037883281707764}
+01/23/2022 08:17:09 - INFO - codeparrot_training - Step 38041: {'lr': 0.0004899819161397241, 'samples': 912984, 'steps': 6340, 'loss/train': 3.1012356281280518}
+01/23/2022 08:17:09 - INFO - codeparrot_training - Step 38042: {'lr': 0.0004899819161397241, 'samples': 913008, 'steps': 6340, 'loss/train': 2.2411375045776367}
+01/23/2022 08:17:09 - INFO - codeparrot_training - Step 38043: {'lr': 0.0004899819161397241, 'samples': 913032, 'steps': 6340, 'loss/train': 1.551196813583374}
+01/23/2022 08:17:10 - INFO - codeparrot_training - Step 38044: {'lr': 0.0004899819161397241, 'samples': 913056, 'steps': 6340, 'loss/train': 3.411506175994873}
+01/23/2022 08:17:10 - INFO - codeparrot_training - Step 38045: {'lr': 0.0004899819161397241, 'samples': 913080, 'steps': 6340, 'loss/train': 2.7246806621551514}
+01/23/2022 08:17:11 - INFO - codeparrot_training - Step 38046: {'lr': 0.0004899819161397241, 'samples': 913104, 'steps': 6340, 'loss/train': 3.236393928527832}
+01/23/2022 08:17:11 - INFO - codeparrot_training - Step 38047: {'lr': 0.0004899773300803774, 'samples': 913128, 'steps': 6341, 'loss/train': 3.076843500137329}
+01/23/2022 08:17:11 - INFO - codeparrot_training - Step 38048: {'lr': 0.0004899773300803774, 'samples': 913152, 'steps': 6341, 'loss/train': 3.1696553230285645}
+01/23/2022 08:17:12 - INFO - codeparrot_training - Step 38049: {'lr': 0.0004899773300803774, 'samples': 913176, 'steps': 6341, 'loss/train': 3.904862403869629}
+01/23/2022 08:17:12 - INFO - codeparrot_training - Step 38050: {'lr': 0.0004899773300803774, 'samples': 913200, 'steps': 6341, 'loss/train': 4.237708568572998}
+01/23/2022 08:17:13 - INFO - codeparrot_training - Step 38051: {'lr': 0.0004899773300803774, 'samples': 913224, 'steps': 6341, 'loss/train': 3.3015873432159424}
+01/23/2022 08:17:13 - INFO - codeparrot_training - Step 38052: {'lr': 0.0004899773300803774, 'samples': 913248, 'steps': 6341, 'loss/train': 3.5032901763916016}
+01/23/2022 08:17:13 - INFO - codeparrot_training - Step 38053: {'lr': 0.0004899727429930438, 'samples': 913272, 'steps': 6342, 'loss/train': 2.53124737739563}
+01/23/2022 08:17:14 - INFO - codeparrot_training - Step 38054: {'lr': 0.0004899727429930438, 'samples': 913296, 'steps': 6342, 'loss/train': 1.0143544673919678}
+01/23/2022 08:17:14 - INFO - codeparrot_training - Step 38055: {'lr': 0.0004899727429930438, 'samples': 913320, 'steps': 6342, 'loss/train': 3.4829835891723633}
+01/23/2022 08:17:14 - INFO - codeparrot_training - Step 38056: {'lr': 0.0004899727429930438, 'samples': 913344, 'steps': 6342, 'loss/train': 2.884268283843994}
+01/23/2022 08:17:15 - INFO - codeparrot_training - Step 38057: {'lr': 0.0004899727429930438, 'samples': 913368, 'steps': 6342, 'loss/train': 3.3271501064300537}
+01/23/2022 08:17:15 - INFO - codeparrot_training - Step 38058: {'lr': 0.0004899727429930438, 'samples': 913392, 'steps': 6342, 'loss/train': 0.8517034649848938}
+01/23/2022 08:17:16 - INFO - codeparrot_training - Step 38059: {'lr': 0.0004899681548777434, 'samples': 913416, 'steps': 6343, 'loss/train': 2.8396875858306885}
+01/23/2022 08:17:16 - INFO - codeparrot_training - Step 38060: {'lr': 0.0004899681548777434, 'samples': 913440, 'steps': 6343, 'loss/train': 2.926868438720703}
+01/23/2022 08:17:16 - INFO - codeparrot_training - Step 38061: {'lr': 0.0004899681548777434, 'samples': 913464, 'steps': 6343, 'loss/train': 4.939311504364014}
+01/23/2022 08:17:17 - INFO - codeparrot_training - Step 38062: {'lr': 0.0004899681548777434, 'samples': 913488, 'steps': 6343, 'loss/train': 3.350137233734131}
+01/23/2022 08:17:17 - INFO - codeparrot_training - Step 38063: {'lr': 0.0004899681548777434, 'samples': 913512, 'steps': 6343, 'loss/train': 2.7452340126037598}
+01/23/2022 08:17:17 - INFO - codeparrot_training - Step 38064: {'lr': 0.0004899681548777434, 'samples': 913536, 'steps': 6343, 'loss/train': 1.283879041671753}
+01/23/2022 08:17:18 - INFO - codeparrot_training - Step 38065: {'lr': 0.0004899635657344954, 'samples': 913560, 'steps': 6344, 'loss/train': 0.7082902193069458}
+01/23/2022 08:17:18 - INFO - codeparrot_training - Step 38066: {'lr': 0.0004899635657344954, 'samples': 913584, 'steps': 6344, 'loss/train': 0.9066534638404846}
+01/23/2022 08:17:19 - INFO - codeparrot_training - Step 38067: {'lr': 0.0004899635657344954, 'samples': 913608, 'steps': 6344, 'loss/train': 0.8130339980125427}
+01/23/2022 08:17:19 - INFO - codeparrot_training - Step 38068: {'lr': 0.0004899635657344954, 'samples': 913632, 'steps': 6344, 'loss/train': 0.9886643886566162}
+01/23/2022 08:17:19 - INFO - codeparrot_training - Step 38069: {'lr': 0.0004899635657344954, 'samples': 913656, 'steps': 6344, 'loss/train': 1.215029239654541}
+01/23/2022 08:17:20 - INFO - codeparrot_training - Step 38070: {'lr': 0.0004899635657344954, 'samples': 913680, 'steps': 6344, 'loss/train': 1.8282800912857056}
+01/23/2022 08:17:20 - INFO - codeparrot_training - Step 38071: {'lr': 0.0004899589755633198, 'samples': 913704, 'steps': 6345, 'loss/train': 3.363719940185547}
+01/23/2022 08:17:21 - INFO - codeparrot_training - Step 38072: {'lr': 0.0004899589755633198, 'samples': 913728, 'steps': 6345, 'loss/train': 2.481395959854126}
+01/23/2022 08:17:21 - INFO - codeparrot_training - Step 38073: {'lr': 0.0004899589755633198, 'samples': 913752, 'steps': 6345, 'loss/train': 3.317283868789673}
+01/23/2022 08:17:21 - INFO - codeparrot_training - Step 38074: {'lr': 0.0004899589755633198, 'samples': 913776, 'steps': 6345, 'loss/train': 3.3609471321105957}
+01/23/2022 08:17:22 - INFO - codeparrot_training - Step 38075: {'lr': 0.0004899589755633198, 'samples': 913800, 'steps': 6345, 'loss/train': 3.06666898727417}
+01/23/2022 08:17:23 - INFO - codeparrot_training - Step 38076: {'lr': 0.0004899589755633198, 'samples': 913824, 'steps': 6345, 'loss/train': 2.731696367263794}
+01/23/2022 08:17:23 - INFO - codeparrot_training - Step 38077: {'lr': 0.0004899543843642362, 'samples': 913848, 'steps': 6346, 'loss/train': 2.0676395893096924}
+01/23/2022 08:17:24 - INFO - codeparrot_training - Step 38078: {'lr': 0.0004899543843642362, 'samples': 913872, 'steps': 6346, 'loss/train': 3.38061785697937}
+01/23/2022 08:17:24 - INFO - codeparrot_training - Step 38079: {'lr': 0.0004899543843642362, 'samples': 913896, 'steps': 6346, 'loss/train': 1.328669786453247}
+01/23/2022 08:17:24 - INFO - codeparrot_training - Step 38080: {'lr': 0.0004899543843642362, 'samples': 913920, 'steps': 6346, 'loss/train': 2.979881763458252}
+01/23/2022 08:17:25 - INFO - codeparrot_training - Step 38081: {'lr': 0.0004899543843642362, 'samples': 913944, 'steps': 6346, 'loss/train': 3.2528834342956543}
+01/23/2022 08:17:25 - INFO - codeparrot_training - Step 38082: {'lr': 0.0004899543843642362, 'samples': 913968, 'steps': 6346, 'loss/train': 3.238300323486328}
+01/23/2022 08:17:25 - INFO - codeparrot_training - Step 38083: {'lr': 0.0004899497921372641, 'samples': 913992, 'steps': 6347, 'loss/train': 3.581804037094116}
+01/23/2022 08:17:26 - INFO - codeparrot_training - Step 38084: {'lr': 0.0004899497921372641, 'samples': 914016, 'steps': 6347, 'loss/train': 2.6571044921875}
+01/23/2022 08:17:26 - INFO - codeparrot_training - Step 38085: {'lr': 0.0004899497921372641, 'samples': 914040, 'steps': 6347, 'loss/train': 2.808481454849243}
+01/23/2022 08:17:27 - INFO - codeparrot_training - Step 38086: {'lr': 0.0004899497921372641, 'samples': 914064, 'steps': 6347, 'loss/train': 3.0118350982666016}
+01/23/2022 08:17:27 - INFO - codeparrot_training - Step 38087: {'lr': 0.0004899497921372641, 'samples': 914088, 'steps': 6347, 'loss/train': 1.9484857320785522}
+01/23/2022 08:17:27 - INFO - codeparrot_training - Step 38088: {'lr': 0.0004899497921372641, 'samples': 914112, 'steps': 6347, 'loss/train': 2.891709804534912}
+01/23/2022 08:17:28 - INFO - codeparrot_training - Step 38089: {'lr': 0.0004899451988824233, 'samples': 914136, 'steps': 6348, 'loss/train': 2.922081232070923}
+01/23/2022 08:17:28 - INFO - codeparrot_training - Step 38090: {'lr': 0.0004899451988824233, 'samples': 914160, 'steps': 6348, 'loss/train': 3.130289077758789}
+01/23/2022 08:17:29 - INFO - codeparrot_training - Step 38091: {'lr': 0.0004899451988824233, 'samples': 914184, 'steps': 6348, 'loss/train': 3.147634744644165}
+01/23/2022 08:17:29 - INFO - codeparrot_training - Step 38092: {'lr': 0.0004899451988824233, 'samples': 914208, 'steps': 6348, 'loss/train': 4.49931001663208}
+01/23/2022 08:17:29 - INFO - codeparrot_training - Step 38093: {'lr': 0.0004899451988824233, 'samples': 914232, 'steps': 6348, 'loss/train': 0.728672444820404}
+01/23/2022 08:17:30 - INFO - codeparrot_training - Step 38094: {'lr': 0.0004899451988824233, 'samples': 914256, 'steps': 6348, 'loss/train': 2.9165680408477783}
+01/23/2022 08:17:30 - INFO - codeparrot_training - Step 38095: {'lr': 0.0004899406045997336, 'samples': 914280, 'steps': 6349, 'loss/train': 3.2809574604034424}
+01/23/2022 08:17:31 - INFO - codeparrot_training - Step 38096: {'lr': 0.0004899406045997336, 'samples': 914304, 'steps': 6349, 'loss/train': 2.4929707050323486}
+01/23/2022 08:17:31 - INFO - codeparrot_training - Step 38097: {'lr': 0.0004899406045997336, 'samples': 914328, 'steps': 6349, 'loss/train': 3.902559757232666}
+01/23/2022 08:17:31 - INFO - codeparrot_training - Step 38098: {'lr': 0.0004899406045997336, 'samples': 914352, 'steps': 6349, 'loss/train': 3.2509851455688477}
+01/23/2022 08:17:32 - INFO - codeparrot_training - Step 38099: {'lr': 0.0004899406045997336, 'samples': 914376, 'steps': 6349, 'loss/train': 3.6423888206481934}
+01/23/2022 08:17:32 - INFO - codeparrot_training - Step 38100: {'lr': 0.0004899406045997336, 'samples': 914400, 'steps': 6349, 'loss/train': 4.303433418273926}
+01/23/2022 08:17:33 - INFO - codeparrot_training - Step 38101: {'lr': 0.0004899360092892143, 'samples': 914424, 'steps': 6350, 'loss/train': 4.248858451843262}
+01/23/2022 08:17:33 - INFO - codeparrot_training - Step 38102: {'lr': 0.0004899360092892143, 'samples': 914448, 'steps': 6350, 'loss/train': 2.9227585792541504}
+01/23/2022 08:17:33 - INFO - codeparrot_training - Step 38103: {'lr': 0.0004899360092892143, 'samples': 914472, 'steps': 6350, 'loss/train': 2.532728672027588}
+01/23/2022 08:17:34 - INFO - codeparrot_training - Step 38104: {'lr': 0.0004899360092892143, 'samples': 914496, 'steps': 6350, 'loss/train': 3.908188819885254}
+01/23/2022 08:17:34 - INFO - codeparrot_training - Step 38105: {'lr': 0.0004899360092892143, 'samples': 914520, 'steps': 6350, 'loss/train': 3.154721736907959}
+01/23/2022 08:17:34 - INFO - codeparrot_training - Step 38106: {'lr': 0.0004899360092892143, 'samples': 914544, 'steps': 6350, 'loss/train': 3.1965229511260986}
+01/23/2022 08:17:35 - INFO - codeparrot_training - Step 38107: {'lr': 0.0004899314129508855, 'samples': 914568, 'steps': 6351, 'loss/train': 2.9464023113250732}
+01/23/2022 08:17:35 - INFO - codeparrot_training - Step 38108: {'lr': 0.0004899314129508855, 'samples': 914592, 'steps': 6351, 'loss/train': 2.919407606124878}
+01/23/2022 08:17:36 - INFO - codeparrot_training - Step 38109: {'lr': 0.0004899314129508855, 'samples': 914616, 'steps': 6351, 'loss/train': 2.4262962341308594}
+01/23/2022 08:17:36 - INFO - codeparrot_training - Step 38110: {'lr': 0.0004899314129508855, 'samples': 914640, 'steps': 6351, 'loss/train': 1.0408339500427246}
+01/23/2022 08:17:36 - INFO - codeparrot_training - Step 38111: {'lr': 0.0004899314129508855, 'samples': 914664, 'steps': 6351, 'loss/train': 3.8066234588623047}
+01/23/2022 08:17:37 - INFO - codeparrot_training - Step 38112: {'lr': 0.0004899314129508855, 'samples': 914688, 'steps': 6351, 'loss/train': 3.708969831466675}
+01/23/2022 08:17:37 - INFO - codeparrot_training - Step 38113: {'lr': 0.0004899268155847667, 'samples': 914712, 'steps': 6352, 'loss/train': 3.4499166011810303}
+01/23/2022 08:17:38 - INFO - codeparrot_training - Step 38114: {'lr': 0.0004899268155847667, 'samples': 914736, 'steps': 6352, 'loss/train': 2.020582437515259}
+01/23/2022 08:17:38 - INFO - codeparrot_training - Step 38115: {'lr': 0.0004899268155847667, 'samples': 914760, 'steps': 6352, 'loss/train': 3.103517532348633}
+01/23/2022 08:17:38 - INFO - codeparrot_training - Step 38116: {'lr': 0.0004899268155847667, 'samples': 914784, 'steps': 6352, 'loss/train': 3.5258491039276123}
+01/23/2022 08:17:39 - INFO - codeparrot_training - Step 38117: {'lr': 0.0004899268155847667, 'samples': 914808, 'steps': 6352, 'loss/train': 3.8134870529174805}
+01/23/2022 08:17:41 - INFO - codeparrot_training - Step 38118: {'lr': 0.0004899268155847667, 'samples': 914832, 'steps': 6352, 'loss/train': 3.958624839782715}
+01/23/2022 08:17:42 - INFO - codeparrot_training - Step 38119: {'lr': 0.0004899222171908776, 'samples': 914856, 'steps': 6353, 'loss/train': 2.830050468444824}
+01/23/2022 08:17:42 - INFO - codeparrot_training - Step 38120: {'lr': 0.0004899222171908776, 'samples': 914880, 'steps': 6353, 'loss/train': 2.822413921356201}
+01/23/2022 08:17:43 - INFO - codeparrot_training - Step 38121: {'lr': 0.0004899222171908776, 'samples': 914904, 'steps': 6353, 'loss/train': 2.75319242477417}
+01/23/2022 08:17:43 - INFO - codeparrot_training - Step 38122: {'lr': 0.0004899222171908776, 'samples': 914928, 'steps': 6353, 'loss/train': 1.929611325263977}
+01/23/2022 08:17:43 - INFO - codeparrot_training - Step 38123: {'lr': 0.0004899222171908776, 'samples': 914952, 'steps': 6353, 'loss/train': 3.5451245307922363}
+01/23/2022 08:17:44 - INFO - codeparrot_training - Step 38124: {'lr': 0.0004899222171908776, 'samples': 914976, 'steps': 6353, 'loss/train': 3.4105775356292725}
+01/23/2022 08:17:44 - INFO - codeparrot_training - Step 38125: {'lr': 0.0004899176177692379, 'samples': 915000, 'steps': 6354, 'loss/train': 3.2422220706939697}
+01/23/2022 08:17:45 - INFO - codeparrot_training - Step 38126: {'lr': 0.0004899176177692379, 'samples': 915024, 'steps': 6354, 'loss/train': 2.1545181274414062}
+01/23/2022 08:17:45 - INFO - codeparrot_training - Step 38127: {'lr': 0.0004899176177692379, 'samples': 915048, 'steps': 6354, 'loss/train': 3.5006208419799805}
+01/23/2022 08:17:45 - INFO - codeparrot_training - Step 38128: {'lr': 0.0004899176177692379, 'samples': 915072, 'steps': 6354, 'loss/train': 2.405958414077759}
+01/23/2022 08:17:46 - INFO - codeparrot_training - Step 38129: {'lr': 0.0004899176177692379, 'samples': 915096, 'steps': 6354, 'loss/train': 3.762758255004883}
+01/23/2022 08:17:46 - INFO - codeparrot_training - Step 38130: {'lr': 0.0004899176177692379, 'samples': 915120, 'steps': 6354, 'loss/train': 3.865910768508911}
+01/23/2022 08:17:46 - INFO - codeparrot_training - Step 38131: {'lr': 0.0004899130173198672, 'samples': 915144, 'steps': 6355, 'loss/train': 3.773228406906128}
+01/23/2022 08:17:47 - INFO - codeparrot_training - Step 38132: {'lr': 0.0004899130173198672, 'samples': 915168, 'steps': 6355, 'loss/train': 2.1782941818237305}
+01/23/2022 08:17:47 - INFO - codeparrot_training - Step 38133: {'lr': 0.0004899130173198672, 'samples': 915192, 'steps': 6355, 'loss/train': 4.1765055656433105}
+01/23/2022 08:17:48 - INFO - codeparrot_training - Step 38134: {'lr': 0.0004899130173198672, 'samples': 915216, 'steps': 6355, 'loss/train': 3.401130437850952}
+01/23/2022 08:17:48 - INFO - codeparrot_training - Step 38135: {'lr': 0.0004899130173198672, 'samples': 915240, 'steps': 6355, 'loss/train': 3.5965940952301025}
+01/23/2022 08:17:48 - INFO - codeparrot_training - Step 38136: {'lr': 0.0004899130173198672, 'samples': 915264, 'steps': 6355, 'loss/train': 3.7707760334014893}
+01/23/2022 08:17:49 - INFO - codeparrot_training - Step 38137: {'lr': 0.0004899084158427855, 'samples': 915288, 'steps': 6356, 'loss/train': 2.263072967529297}
+01/23/2022 08:17:49 - INFO - codeparrot_training - Step 38138: {'lr': 0.0004899084158427855, 'samples': 915312, 'steps': 6356, 'loss/train': 1.3801406621932983}
+01/23/2022 08:17:49 - INFO - codeparrot_training - Step 38139: {'lr': 0.0004899084158427855, 'samples': 915336, 'steps': 6356, 'loss/train': 2.48618745803833}
+01/23/2022 08:17:50 - INFO - codeparrot_training - Step 38140: {'lr': 0.0004899084158427855, 'samples': 915360, 'steps': 6356, 'loss/train': 3.022531509399414}
+01/23/2022 08:17:50 - INFO - codeparrot_training - Step 38141: {'lr': 0.0004899084158427855, 'samples': 915384, 'steps': 6356, 'loss/train': 1.4696546792984009}
+01/23/2022 08:17:51 - INFO - codeparrot_training - Step 38142: {'lr': 0.0004899084158427855, 'samples': 915408, 'steps': 6356, 'loss/train': 2.8884263038635254}
+01/23/2022 08:17:51 - INFO - codeparrot_training - Step 38143: {'lr': 0.0004899038133380121, 'samples': 915432, 'steps': 6357, 'loss/train': 2.60729718208313}
+01/23/2022 08:17:51 - INFO - codeparrot_training - Step 38144: {'lr': 0.0004899038133380121, 'samples': 915456, 'steps': 6357, 'loss/train': 3.8576347827911377}
+01/23/2022 08:17:52 - INFO - codeparrot_training - Step 38145: {'lr': 0.0004899038133380121, 'samples': 915480, 'steps': 6357, 'loss/train': 2.502068281173706}
+01/23/2022 08:17:52 - INFO - codeparrot_training - Step 38146: {'lr': 0.0004899038133380121, 'samples': 915504, 'steps': 6357, 'loss/train': 2.059119939804077}
+01/23/2022 08:17:52 - INFO - codeparrot_training - Step 38147: {'lr': 0.0004899038133380121, 'samples': 915528, 'steps': 6357, 'loss/train': 0.3350234627723694}
+01/23/2022 08:17:53 - INFO - codeparrot_training - Step 38148: {'lr': 0.0004899038133380121, 'samples': 915552, 'steps': 6357, 'loss/train': 2.639911651611328}
+01/23/2022 08:17:53 - INFO - codeparrot_training - Step 38149: {'lr': 0.0004898992098055671, 'samples': 915576, 'steps': 6358, 'loss/train': 3.7138493061065674}
+01/23/2022 08:17:54 - INFO - codeparrot_training - Step 38150: {'lr': 0.0004898992098055671, 'samples': 915600, 'steps': 6358, 'loss/train': 2.5239033699035645}
+01/23/2022 08:17:54 - INFO - codeparrot_training - Step 38151: {'lr': 0.0004898992098055671, 'samples': 915624, 'steps': 6358, 'loss/train': 3.581073522567749}
+01/23/2022 08:17:54 - INFO - codeparrot_training - Step 38152: {'lr': 0.0004898992098055671, 'samples': 915648, 'steps': 6358, 'loss/train': 4.798467636108398}
+01/23/2022 08:17:55 - INFO - codeparrot_training - Step 38153: {'lr': 0.0004898992098055671, 'samples': 915672, 'steps': 6358, 'loss/train': 3.647848606109619}
+01/23/2022 08:17:55 - INFO - codeparrot_training - Step 38154: {'lr': 0.0004898992098055671, 'samples': 915696, 'steps': 6358, 'loss/train': 4.256852626800537}
+01/23/2022 08:17:56 - INFO - codeparrot_training - Step 38155: {'lr': 0.00048989460524547, 'samples': 915720, 'steps': 6359, 'loss/train': 2.179739475250244}
+01/23/2022 08:17:56 - INFO - codeparrot_training - Step 38156: {'lr': 0.00048989460524547, 'samples': 915744, 'steps': 6359, 'loss/train': 1.6021089553833008}
+01/23/2022 08:17:57 - INFO - codeparrot_training - Step 38157: {'lr': 0.00048989460524547, 'samples': 915768, 'steps': 6359, 'loss/train': 0.8729740977287292}
+01/23/2022 08:17:57 - INFO - codeparrot_training - Step 38158: {'lr': 0.00048989460524547, 'samples': 915792, 'steps': 6359, 'loss/train': 4.099199295043945}
+01/23/2022 08:17:58 - INFO - codeparrot_training - Step 38159: {'lr': 0.00048989460524547, 'samples': 915816, 'steps': 6359, 'loss/train': 2.846693992614746}
+01/23/2022 08:17:58 - INFO - codeparrot_training - Step 38160: {'lr': 0.00048989460524547, 'samples': 915840, 'steps': 6359, 'loss/train': 2.365767478942871}
+01/23/2022 08:17:58 - INFO - codeparrot_training - Step 38161: {'lr': 0.0004898899996577407, 'samples': 915864, 'steps': 6360, 'loss/train': 3.287539005279541}
+01/23/2022 08:17:59 - INFO - codeparrot_training - Step 38162: {'lr': 0.0004898899996577407, 'samples': 915888, 'steps': 6360, 'loss/train': 4.323760032653809}
+01/23/2022 08:17:59 - INFO - codeparrot_training - Step 38163: {'lr': 0.0004898899996577407, 'samples': 915912, 'steps': 6360, 'loss/train': 2.913491725921631}
+01/23/2022 08:18:00 - INFO - codeparrot_training - Step 38164: {'lr': 0.0004898899996577407, 'samples': 915936, 'steps': 6360, 'loss/train': 3.5177295207977295}
+01/23/2022 08:18:00 - INFO - codeparrot_training - Step 38165: {'lr': 0.0004898899996577407, 'samples': 915960, 'steps': 6360, 'loss/train': 3.220371961593628}
+01/23/2022 08:18:00 - INFO - codeparrot_training - Step 38166: {'lr': 0.0004898899996577407, 'samples': 915984, 'steps': 6360, 'loss/train': 3.9639060497283936}
+01/23/2022 08:18:01 - INFO - codeparrot_training - Step 38167: {'lr': 0.0004898853930423987, 'samples': 916008, 'steps': 6361, 'loss/train': 3.4330966472625732}
+01/23/2022 08:18:01 - INFO - codeparrot_training - Step 38168: {'lr': 0.0004898853930423987, 'samples': 916032, 'steps': 6361, 'loss/train': 5.5290937423706055}
+01/23/2022 08:18:02 - INFO - codeparrot_training - Step 38169: {'lr': 0.0004898853930423987, 'samples': 916056, 'steps': 6361, 'loss/train': 3.673762798309326}
+01/23/2022 08:18:02 - INFO - codeparrot_training - Step 38170: {'lr': 0.0004898853930423987, 'samples': 916080, 'steps': 6361, 'loss/train': 3.883852005004883}
+01/23/2022 08:18:02 - INFO - codeparrot_training - Step 38171: {'lr': 0.0004898853930423987, 'samples': 916104, 'steps': 6361, 'loss/train': 3.036440849304199}
+01/23/2022 08:18:03 - INFO - codeparrot_training - Step 38172: {'lr': 0.0004898853930423987, 'samples': 916128, 'steps': 6361, 'loss/train': 4.120728015899658}
+01/23/2022 08:18:03 - INFO - codeparrot_training - Step 38173: {'lr': 0.0004898807853994639, 'samples': 916152, 'steps': 6362, 'loss/train': 3.3594822883605957}
+01/23/2022 08:18:03 - INFO - codeparrot_training - Step 38174: {'lr': 0.0004898807853994639, 'samples': 916176, 'steps': 6362, 'loss/train': 3.1588242053985596}
+01/23/2022 08:18:04 - INFO - codeparrot_training - Step 38175: {'lr': 0.0004898807853994639, 'samples': 916200, 'steps': 6362, 'loss/train': 3.4814698696136475}
+01/23/2022 08:18:04 - INFO - codeparrot_training - Step 38176: {'lr': 0.0004898807853994639, 'samples': 916224, 'steps': 6362, 'loss/train': 3.6570332050323486}
+01/23/2022 08:18:05 - INFO - codeparrot_training - Step 38177: {'lr': 0.0004898807853994639, 'samples': 916248, 'steps': 6362, 'loss/train': 2.1221845149993896}
+01/23/2022 08:18:05 - INFO - codeparrot_training - Step 38178: {'lr': 0.0004898807853994639, 'samples': 916272, 'steps': 6362, 'loss/train': 2.207866668701172}
+01/23/2022 08:18:05 - INFO - codeparrot_training - Step 38179: {'lr': 0.000489876176728956, 'samples': 916296, 'steps': 6363, 'loss/train': 2.3899848461151123}
+01/23/2022 08:18:06 - INFO - codeparrot_training - Step 38180: {'lr': 0.000489876176728956, 'samples': 916320, 'steps': 6363, 'loss/train': 2.321554183959961}
+01/23/2022 08:18:06 - INFO - codeparrot_training - Step 38181: {'lr': 0.000489876176728956, 'samples': 916344, 'steps': 6363, 'loss/train': 4.870482444763184}
+01/23/2022 08:18:06 - INFO - codeparrot_training - Step 38182: {'lr': 0.000489876176728956, 'samples': 916368, 'steps': 6363, 'loss/train': 1.3872848749160767}
+01/23/2022 08:18:07 - INFO - codeparrot_training - Step 38183: {'lr': 0.000489876176728956, 'samples': 916392, 'steps': 6363, 'loss/train': 3.124969244003296}
+01/23/2022 08:18:07 - INFO - codeparrot_training - Step 38184: {'lr': 0.000489876176728956, 'samples': 916416, 'steps': 6363, 'loss/train': 2.3442795276641846}
+01/23/2022 08:18:08 - INFO - codeparrot_training - Step 38185: {'lr': 0.0004898715670308947, 'samples': 916440, 'steps': 6364, 'loss/train': 3.130053758621216}
+01/23/2022 08:18:08 - INFO - codeparrot_training - Step 38186: {'lr': 0.0004898715670308947, 'samples': 916464, 'steps': 6364, 'loss/train': 4.003077983856201}
+01/23/2022 08:18:08 - INFO - codeparrot_training - Step 38187: {'lr': 0.0004898715670308947, 'samples': 916488, 'steps': 6364, 'loss/train': 3.591078281402588}
+01/23/2022 08:18:09 - INFO - codeparrot_training - Step 38188: {'lr': 0.0004898715670308947, 'samples': 916512, 'steps': 6364, 'loss/train': 3.0211620330810547}
+01/23/2022 08:18:09 - INFO - codeparrot_training - Step 38189: {'lr': 0.0004898715670308947, 'samples': 916536, 'steps': 6364, 'loss/train': 3.4296085834503174}
+01/23/2022 08:18:09 - INFO - codeparrot_training - Step 38190: {'lr': 0.0004898715670308947, 'samples': 916560, 'steps': 6364, 'loss/train': 2.967550039291382}
+01/23/2022 08:18:10 - INFO - codeparrot_training - Step 38191: {'lr': 0.0004898669563052997, 'samples': 916584, 'steps': 6365, 'loss/train': 2.760723352432251}
+01/23/2022 08:18:10 - INFO - codeparrot_training - Step 38192: {'lr': 0.0004898669563052997, 'samples': 916608, 'steps': 6365, 'loss/train': 3.1331050395965576}
+01/23/2022 08:18:11 - INFO - codeparrot_training - Step 38193: {'lr': 0.0004898669563052997, 'samples': 916632, 'steps': 6365, 'loss/train': 3.3369693756103516}
+01/23/2022 08:18:11 - INFO - codeparrot_training - Step 38194: {'lr': 0.0004898669563052997, 'samples': 916656, 'steps': 6365, 'loss/train': 1.2115774154663086}
+01/23/2022 08:18:12 - INFO - codeparrot_training - Step 38195: {'lr': 0.0004898669563052997, 'samples': 916680, 'steps': 6365, 'loss/train': 3.2525370121002197}
+01/23/2022 08:18:12 - INFO - codeparrot_training - Step 38196: {'lr': 0.0004898669563052997, 'samples': 916704, 'steps': 6365, 'loss/train': 1.7797733545303345}
+01/23/2022 08:18:13 - INFO - codeparrot_training - Step 38197: {'lr': 0.0004898623445521909, 'samples': 916728, 'steps': 6366, 'loss/train': 1.9004286527633667}
+01/23/2022 08:18:13 - INFO - codeparrot_training - Step 38198: {'lr': 0.0004898623445521909, 'samples': 916752, 'steps': 6366, 'loss/train': 3.381396770477295}
+01/23/2022 08:18:14 - INFO - codeparrot_training - Step 38199: {'lr': 0.0004898623445521909, 'samples': 916776, 'steps': 6366, 'loss/train': 3.425806999206543}
+01/23/2022 08:18:14 - INFO - codeparrot_training - Step 38200: {'lr': 0.0004898623445521909, 'samples': 916800, 'steps': 6366, 'loss/train': 3.7784905433654785}
+01/23/2022 08:18:14 - INFO - codeparrot_training - Step 38201: {'lr': 0.0004898623445521909, 'samples': 916824, 'steps': 6366, 'loss/train': 0.6941621899604797}
+01/23/2022 08:18:15 - INFO - codeparrot_training - Step 38202: {'lr': 0.0004898623445521909, 'samples': 916848, 'steps': 6366, 'loss/train': 2.0884029865264893}
+01/23/2022 08:18:15 - INFO - codeparrot_training - Step 38203: {'lr': 0.000489857731771588, 'samples': 916872, 'steps': 6367, 'loss/train': 3.495872974395752}
+01/23/2022 08:18:16 - INFO - codeparrot_training - Step 38204: {'lr': 0.000489857731771588, 'samples': 916896, 'steps': 6367, 'loss/train': 4.300698280334473}
+01/23/2022 08:18:16 - INFO - codeparrot_training - Step 38205: {'lr': 0.000489857731771588, 'samples': 916920, 'steps': 6367, 'loss/train': 1.5953290462493896}
+01/23/2022 08:18:16 - INFO - codeparrot_training - Step 38206: {'lr': 0.000489857731771588, 'samples': 916944, 'steps': 6367, 'loss/train': 2.2198429107666016}
+01/23/2022 08:18:17 - INFO - codeparrot_training - Step 38207: {'lr': 0.000489857731771588, 'samples': 916968, 'steps': 6367, 'loss/train': 1.9437302350997925}
+01/23/2022 08:18:17 - INFO - codeparrot_training - Step 38208: {'lr': 0.000489857731771588, 'samples': 916992, 'steps': 6367, 'loss/train': 4.0869975090026855}
+01/23/2022 08:18:18 - INFO - codeparrot_training - Step 38209: {'lr': 0.0004898531179635108, 'samples': 917016, 'steps': 6368, 'loss/train': 6.228350639343262}
+01/23/2022 08:18:18 - INFO - codeparrot_training - Step 38210: {'lr': 0.0004898531179635108, 'samples': 917040, 'steps': 6368, 'loss/train': 3.7240591049194336}
+01/23/2022 08:18:18 - INFO - codeparrot_training - Step 38211: {'lr': 0.0004898531179635108, 'samples': 917064, 'steps': 6368, 'loss/train': 4.015963077545166}
+01/23/2022 08:18:19 - INFO - codeparrot_training - Step 38212: {'lr': 0.0004898531179635108, 'samples': 917088, 'steps': 6368, 'loss/train': 3.267871141433716}
+01/23/2022 08:18:19 - INFO - codeparrot_training - Step 38213: {'lr': 0.0004898531179635108, 'samples': 917112, 'steps': 6368, 'loss/train': 3.7386980056762695}
+01/23/2022 08:18:19 - INFO - codeparrot_training - Step 38214: {'lr': 0.0004898531179635108, 'samples': 917136, 'steps': 6368, 'loss/train': 2.302375078201294}
+01/23/2022 08:18:20 - INFO - codeparrot_training - Step 38215: {'lr': 0.0004898485031279788, 'samples': 917160, 'steps': 6369, 'loss/train': 2.4817845821380615}
+01/23/2022 08:18:20 - INFO - codeparrot_training - Step 38216: {'lr': 0.0004898485031279788, 'samples': 917184, 'steps': 6369, 'loss/train': 3.312445640563965}
+01/23/2022 08:18:21 - INFO - codeparrot_training - Step 38217: {'lr': 0.0004898485031279788, 'samples': 917208, 'steps': 6369, 'loss/train': 4.484522342681885}
+01/23/2022 08:18:21 - INFO - codeparrot_training - Step 38218: {'lr': 0.0004898485031279788, 'samples': 917232, 'steps': 6369, 'loss/train': 3.3678274154663086}
+01/23/2022 08:18:21 - INFO - codeparrot_training - Step 38219: {'lr': 0.0004898485031279788, 'samples': 917256, 'steps': 6369, 'loss/train': 2.6260104179382324}
+01/23/2022 08:18:22 - INFO - codeparrot_training - Step 38220: {'lr': 0.0004898485031279788, 'samples': 917280, 'steps': 6369, 'loss/train': 4.2833662033081055}
+01/23/2022 08:18:22 - INFO - codeparrot_training - Step 38221: {'lr': 0.0004898438872650121, 'samples': 917304, 'steps': 6370, 'loss/train': 2.435486316680908}
+01/23/2022 08:18:23 - INFO - codeparrot_training - Step 38222: {'lr': 0.0004898438872650121, 'samples': 917328, 'steps': 6370, 'loss/train': 2.747389078140259}
+01/23/2022 08:18:23 - INFO - codeparrot_training - Step 38223: {'lr': 0.0004898438872650121, 'samples': 917352, 'steps': 6370, 'loss/train': 4.36800479888916}
+01/23/2022 08:18:23 - INFO - codeparrot_training - Step 38224: {'lr': 0.0004898438872650121, 'samples': 917376, 'steps': 6370, 'loss/train': 4.023513317108154}
+01/23/2022 08:18:24 - INFO - codeparrot_training - Step 38225: {'lr': 0.0004898438872650121, 'samples': 917400, 'steps': 6370, 'loss/train': 5.549417018890381}
+01/23/2022 08:18:24 - INFO - codeparrot_training - Step 38226: {'lr': 0.0004898438872650121, 'samples': 917424, 'steps': 6370, 'loss/train': 2.602726459503174}
+01/23/2022 08:18:24 - INFO - codeparrot_training - Step 38227: {'lr': 0.0004898392703746304, 'samples': 917448, 'steps': 6371, 'loss/train': 3.141883373260498}
+01/23/2022 08:18:25 - INFO - codeparrot_training - Step 38228: {'lr': 0.0004898392703746304, 'samples': 917472, 'steps': 6371, 'loss/train': 1.8205872774124146}
+01/23/2022 08:18:25 - INFO - codeparrot_training - Step 38229: {'lr': 0.0004898392703746304, 'samples': 917496, 'steps': 6371, 'loss/train': 3.1520376205444336}
+01/23/2022 08:18:26 - INFO - codeparrot_training - Step 38230: {'lr': 0.0004898392703746304, 'samples': 917520, 'steps': 6371, 'loss/train': 4.444386959075928}
+01/23/2022 08:18:26 - INFO - codeparrot_training - Step 38231: {'lr': 0.0004898392703746304, 'samples': 917544, 'steps': 6371, 'loss/train': 3.85422682762146}
+01/23/2022 08:18:26 - INFO - codeparrot_training - Step 38232: {'lr': 0.0004898392703746304, 'samples': 917568, 'steps': 6371, 'loss/train': 3.759194850921631}
+01/23/2022 08:18:27 - INFO - codeparrot_training - Step 38233: {'lr': 0.0004898346524568533, 'samples': 917592, 'steps': 6372, 'loss/train': 3.1423327922821045}
+01/23/2022 08:18:27 - INFO - codeparrot_training - Step 38234: {'lr': 0.0004898346524568533, 'samples': 917616, 'steps': 6372, 'loss/train': 2.7885050773620605}
+01/23/2022 08:18:28 - INFO - codeparrot_training - Step 38235: {'lr': 0.0004898346524568533, 'samples': 917640, 'steps': 6372, 'loss/train': 3.435724973678589}
+01/23/2022 08:18:29 - INFO - codeparrot_training - Step 38236: {'lr': 0.0004898346524568533, 'samples': 917664, 'steps': 6372, 'loss/train': 3.223828077316284}
+01/23/2022 08:18:29 - INFO - codeparrot_training - Step 38237: {'lr': 0.0004898346524568533, 'samples': 917688, 'steps': 6372, 'loss/train': 2.0006043910980225}
+01/23/2022 08:18:29 - INFO - codeparrot_training - Step 38238: {'lr': 0.0004898346524568533, 'samples': 917712, 'steps': 6372, 'loss/train': 4.405363082885742}
+01/23/2022 08:18:30 - INFO - codeparrot_training - Step 38239: {'lr': 0.0004898300335117008, 'samples': 917736, 'steps': 6373, 'loss/train': 2.9164648056030273}
+01/23/2022 08:18:30 - INFO - codeparrot_training - Step 38240: {'lr': 0.0004898300335117008, 'samples': 917760, 'steps': 6373, 'loss/train': 4.158557415008545}
+01/23/2022 08:18:31 - INFO - codeparrot_training - Step 38241: {'lr': 0.0004898300335117008, 'samples': 917784, 'steps': 6373, 'loss/train': 2.6195151805877686}
+01/23/2022 08:18:31 - INFO - codeparrot_training - Step 38242: {'lr': 0.0004898300335117008, 'samples': 917808, 'steps': 6373, 'loss/train': 3.0255515575408936}
+01/23/2022 08:18:31 - INFO - codeparrot_training - Step 38243: {'lr': 0.0004898300335117008, 'samples': 917832, 'steps': 6373, 'loss/train': 4.704525947570801}
+01/23/2022 08:18:32 - INFO - codeparrot_training - Step 38244: {'lr': 0.0004898300335117008, 'samples': 917856, 'steps': 6373, 'loss/train': 3.544569969177246}
+01/23/2022 08:18:32 - INFO - codeparrot_training - Step 38245: {'lr': 0.0004898254135391926, 'samples': 917880, 'steps': 6374, 'loss/train': 3.520625114440918}
+01/23/2022 08:18:32 - INFO - codeparrot_training - Step 38246: {'lr': 0.0004898254135391926, 'samples': 917904, 'steps': 6374, 'loss/train': 3.1598100662231445}
+01/23/2022 08:18:33 - INFO - codeparrot_training - Step 38247: {'lr': 0.0004898254135391926, 'samples': 917928, 'steps': 6374, 'loss/train': 3.4902069568634033}
+01/23/2022 08:18:33 - INFO - codeparrot_training - Step 38248: {'lr': 0.0004898254135391926, 'samples': 917952, 'steps': 6374, 'loss/train': 2.9529941082000732}
+01/23/2022 08:18:34 - INFO - codeparrot_training - Step 38249: {'lr': 0.0004898254135391926, 'samples': 917976, 'steps': 6374, 'loss/train': 3.3891217708587646}
+01/23/2022 08:18:34 - INFO - codeparrot_training - Step 38250: {'lr': 0.0004898254135391926, 'samples': 918000, 'steps': 6374, 'loss/train': 3.3574085235595703}
+01/23/2022 08:18:34 - INFO - codeparrot_training - Step 38251: {'lr': 0.0004898207925393485, 'samples': 918024, 'steps': 6375, 'loss/train': 3.0928428173065186}
+01/23/2022 08:18:35 - INFO - codeparrot_training - Step 38252: {'lr': 0.0004898207925393485, 'samples': 918048, 'steps': 6375, 'loss/train': 2.3558835983276367}
+01/23/2022 08:18:35 - INFO - codeparrot_training - Step 38253: {'lr': 0.0004898207925393485, 'samples': 918072, 'steps': 6375, 'loss/train': 3.1004385948181152}
+01/23/2022 08:18:35 - INFO - codeparrot_training - Step 38254: {'lr': 0.0004898207925393485, 'samples': 918096, 'steps': 6375, 'loss/train': 2.7639570236206055}
+01/23/2022 08:18:36 - INFO - codeparrot_training - Step 38255: {'lr': 0.0004898207925393485, 'samples': 918120, 'steps': 6375, 'loss/train': 2.988399028778076}
+01/23/2022 08:18:36 - INFO - codeparrot_training - Step 38256: {'lr': 0.0004898207925393485, 'samples': 918144, 'steps': 6375, 'loss/train': 3.6183886528015137}
+01/23/2022 08:18:37 - INFO - codeparrot_training - Step 38257: {'lr': 0.0004898161705121882, 'samples': 918168, 'steps': 6376, 'loss/train': 2.740432024002075}
+01/23/2022 08:18:37 - INFO - codeparrot_training - Step 38258: {'lr': 0.0004898161705121882, 'samples': 918192, 'steps': 6376, 'loss/train': 4.439938068389893}
+01/23/2022 08:18:37 - INFO - codeparrot_training - Step 38259: {'lr': 0.0004898161705121882, 'samples': 918216, 'steps': 6376, 'loss/train': 3.1763720512390137}
+01/23/2022 08:18:38 - INFO - codeparrot_training - Step 38260: {'lr': 0.0004898161705121882, 'samples': 918240, 'steps': 6376, 'loss/train': 4.427206039428711}
+01/23/2022 08:18:38 - INFO - codeparrot_training - Step 38261: {'lr': 0.0004898161705121882, 'samples': 918264, 'steps': 6376, 'loss/train': 2.5697762966156006}
+01/23/2022 08:18:38 - INFO - codeparrot_training - Step 38262: {'lr': 0.0004898161705121882, 'samples': 918288, 'steps': 6376, 'loss/train': 3.963862419128418}
+01/23/2022 08:18:39 - INFO - codeparrot_training - Step 38263: {'lr': 0.0004898115474577315, 'samples': 918312, 'steps': 6377, 'loss/train': 3.3820888996124268}
+01/23/2022 08:18:39 - INFO - codeparrot_training - Step 38264: {'lr': 0.0004898115474577315, 'samples': 918336, 'steps': 6377, 'loss/train': 1.2828997373580933}
+01/23/2022 08:18:40 - INFO - codeparrot_training - Step 38265: {'lr': 0.0004898115474577315, 'samples': 918360, 'steps': 6377, 'loss/train': 2.102559804916382}
+01/23/2022 08:18:40 - INFO - codeparrot_training - Step 38266: {'lr': 0.0004898115474577315, 'samples': 918384, 'steps': 6377, 'loss/train': 2.4134914875030518}
+01/23/2022 08:18:40 - INFO - codeparrot_training - Step 38267: {'lr': 0.0004898115474577315, 'samples': 918408, 'steps': 6377, 'loss/train': 3.750861167907715}
+01/23/2022 08:18:41 - INFO - codeparrot_training - Step 38268: {'lr': 0.0004898115474577315, 'samples': 918432, 'steps': 6377, 'loss/train': 2.537480592727661}
+01/23/2022 08:18:41 - INFO - codeparrot_training - Step 38269: {'lr': 0.0004898069233759985, 'samples': 918456, 'steps': 6378, 'loss/train': 4.217484474182129}
+01/23/2022 08:18:42 - INFO - codeparrot_training - Step 38270: {'lr': 0.0004898069233759985, 'samples': 918480, 'steps': 6378, 'loss/train': 3.380801200866699}
+01/23/2022 08:18:42 - INFO - codeparrot_training - Step 38271: {'lr': 0.0004898069233759985, 'samples': 918504, 'steps': 6378, 'loss/train': 2.8406808376312256}
+01/23/2022 08:18:42 - INFO - codeparrot_training - Step 38272: {'lr': 0.0004898069233759985, 'samples': 918528, 'steps': 6378, 'loss/train': 2.128937005996704}
+01/23/2022 08:18:43 - INFO - codeparrot_training - Step 38273: {'lr': 0.0004898069233759985, 'samples': 918552, 'steps': 6378, 'loss/train': 2.4887189865112305}
+01/23/2022 08:18:43 - INFO - codeparrot_training - Step 38274: {'lr': 0.0004898069233759985, 'samples': 918576, 'steps': 6378, 'loss/train': 3.587932825088501}
+01/23/2022 08:18:45 - INFO - codeparrot_training - Step 38275: {'lr': 0.0004898022982670085, 'samples': 918600, 'steps': 6379, 'loss/train': 3.266282796859741}
+01/23/2022 08:18:45 - INFO - codeparrot_training - Step 38276: {'lr': 0.0004898022982670085, 'samples': 918624, 'steps': 6379, 'loss/train': 2.0515546798706055}
+01/23/2022 08:18:45 - INFO - codeparrot_training - Step 38277: {'lr': 0.0004898022982670085, 'samples': 918648, 'steps': 6379, 'loss/train': 3.6085147857666016}
+01/23/2022 08:18:46 - INFO - codeparrot_training - Step 38278: {'lr': 0.0004898022982670085, 'samples': 918672, 'steps': 6379, 'loss/train': 2.1753923892974854}
+01/23/2022 08:18:46 - INFO - codeparrot_training - Step 38279: {'lr': 0.0004898022982670085, 'samples': 918696, 'steps': 6379, 'loss/train': 3.5904111862182617}
+01/23/2022 08:18:46 - INFO - codeparrot_training - Step 38280: {'lr': 0.0004898022982670085, 'samples': 918720, 'steps': 6379, 'loss/train': 3.299781084060669}
+01/23/2022 08:18:47 - INFO - codeparrot_training - Step 38281: {'lr': 0.0004897976721307818, 'samples': 918744, 'steps': 6380, 'loss/train': 3.4533417224884033}
+01/23/2022 08:18:47 - INFO - codeparrot_training - Step 38282: {'lr': 0.0004897976721307818, 'samples': 918768, 'steps': 6380, 'loss/train': 3.357304096221924}
+01/23/2022 08:18:48 - INFO - codeparrot_training - Step 38283: {'lr': 0.0004897976721307818, 'samples': 918792, 'steps': 6380, 'loss/train': 3.8616607189178467}
+01/23/2022 08:18:48 - INFO - codeparrot_training - Step 38284: {'lr': 0.0004897976721307818, 'samples': 918816, 'steps': 6380, 'loss/train': 2.365204095840454}
+01/23/2022 08:18:48 - INFO - codeparrot_training - Step 38285: {'lr': 0.0004897976721307818, 'samples': 918840, 'steps': 6380, 'loss/train': 2.207291603088379}
+01/23/2022 08:18:49 - INFO - codeparrot_training - Step 38286: {'lr': 0.0004897976721307818, 'samples': 918864, 'steps': 6380, 'loss/train': 2.6722540855407715}
+01/23/2022 08:18:49 - INFO - codeparrot_training - Step 38287: {'lr': 0.000489793044967338, 'samples': 918888, 'steps': 6381, 'loss/train': 2.076080560684204}
+01/23/2022 08:18:49 - INFO - codeparrot_training - Step 38288: {'lr': 0.000489793044967338, 'samples': 918912, 'steps': 6381, 'loss/train': 3.5774059295654297}
+01/23/2022 08:18:50 - INFO - codeparrot_training - Step 38289: {'lr': 0.000489793044967338, 'samples': 918936, 'steps': 6381, 'loss/train': 2.7509045600891113}
+01/23/2022 08:18:50 - INFO - codeparrot_training - Step 38290: {'lr': 0.000489793044967338, 'samples': 918960, 'steps': 6381, 'loss/train': 2.621507167816162}
+01/23/2022 08:18:51 - INFO - codeparrot_training - Step 38291: {'lr': 0.000489793044967338, 'samples': 918984, 'steps': 6381, 'loss/train': 5.174093723297119}
+01/23/2022 08:18:51 - INFO - codeparrot_training - Step 38292: {'lr': 0.000489793044967338, 'samples': 919008, 'steps': 6381, 'loss/train': 3.9807186126708984}
+01/23/2022 08:18:51 - INFO - codeparrot_training - Step 38293: {'lr': 0.0004897884167766969, 'samples': 919032, 'steps': 6382, 'loss/train': 3.916809558868408}
+01/23/2022 08:18:52 - INFO - codeparrot_training - Step 38294: {'lr': 0.0004897884167766969, 'samples': 919056, 'steps': 6382, 'loss/train': 3.848700523376465}
+01/23/2022 08:18:52 - INFO - codeparrot_training - Step 38295: {'lr': 0.0004897884167766969, 'samples': 919080, 'steps': 6382, 'loss/train': 4.010155200958252}
+01/23/2022 08:18:53 - INFO - codeparrot_training - Step 38296: {'lr': 0.0004897884167766969, 'samples': 919104, 'steps': 6382, 'loss/train': 3.8086509704589844}
+01/23/2022 08:18:53 - INFO - codeparrot_training - Step 38297: {'lr': 0.0004897884167766969, 'samples': 919128, 'steps': 6382, 'loss/train': 1.995051622390747}
+01/23/2022 08:18:53 - INFO - codeparrot_training - Step 38298: {'lr': 0.0004897884167766969, 'samples': 919152, 'steps': 6382, 'loss/train': 2.3024845123291016}
+01/23/2022 08:18:54 - INFO - codeparrot_training - Step 38299: {'lr': 0.0004897837875588784, 'samples': 919176, 'steps': 6383, 'loss/train': 1.9139002561569214}
+01/23/2022 08:18:54 - INFO - codeparrot_training - Step 38300: {'lr': 0.0004897837875588784, 'samples': 919200, 'steps': 6383, 'loss/train': 1.8531967401504517}
+01/23/2022 08:18:54 - INFO - codeparrot_training - Step 38301: {'lr': 0.0004897837875588784, 'samples': 919224, 'steps': 6383, 'loss/train': 1.897322416305542}
+01/23/2022 08:18:55 - INFO - codeparrot_training - Step 38302: {'lr': 0.0004897837875588784, 'samples': 919248, 'steps': 6383, 'loss/train': 1.8646262884140015}
+01/23/2022 08:18:55 - INFO - codeparrot_training - Step 38303: {'lr': 0.0004897837875588784, 'samples': 919272, 'steps': 6383, 'loss/train': 1.928025245666504}
+01/23/2022 08:18:56 - INFO - codeparrot_training - Step 38304: {'lr': 0.0004897837875588784, 'samples': 919296, 'steps': 6383, 'loss/train': 3.687389612197876}
+01/23/2022 08:18:56 - INFO - codeparrot_training - Step 38305: {'lr': 0.0004897791573139022, 'samples': 919320, 'steps': 6384, 'loss/train': 3.245541572570801}
+01/23/2022 08:18:56 - INFO - codeparrot_training - Step 38306: {'lr': 0.0004897791573139022, 'samples': 919344, 'steps': 6384, 'loss/train': 3.288637399673462}
+01/23/2022 08:18:57 - INFO - codeparrot_training - Step 38307: {'lr': 0.0004897791573139022, 'samples': 919368, 'steps': 6384, 'loss/train': 1.6613634824752808}
+01/23/2022 08:18:57 - INFO - codeparrot_training - Step 38308: {'lr': 0.0004897791573139022, 'samples': 919392, 'steps': 6384, 'loss/train': 3.0984208583831787}
+01/23/2022 08:18:57 - INFO - codeparrot_training - Step 38309: {'lr': 0.0004897791573139022, 'samples': 919416, 'steps': 6384, 'loss/train': 3.8946759700775146}
+01/23/2022 08:18:58 - INFO - codeparrot_training - Step 38310: {'lr': 0.0004897791573139022, 'samples': 919440, 'steps': 6384, 'loss/train': 1.345027208328247}
+01/23/2022 08:18:58 - INFO - codeparrot_training - Step 38311: {'lr': 0.0004897745260417884, 'samples': 919464, 'steps': 6385, 'loss/train': 3.350377082824707}
+01/23/2022 08:18:59 - INFO - codeparrot_training - Step 38312: {'lr': 0.0004897745260417884, 'samples': 919488, 'steps': 6385, 'loss/train': 2.0760388374328613}
+01/23/2022 08:18:59 - INFO - codeparrot_training - Step 38313: {'lr': 0.0004897745260417884, 'samples': 919512, 'steps': 6385, 'loss/train': 1.884885549545288}
+01/23/2022 08:18:59 - INFO - codeparrot_training - Step 38314: {'lr': 0.0004897745260417884, 'samples': 919536, 'steps': 6385, 'loss/train': 2.743055820465088}
+01/23/2022 08:19:00 - INFO - codeparrot_training - Step 38315: {'lr': 0.0004897745260417884, 'samples': 919560, 'steps': 6385, 'loss/train': 3.1738791465759277}
+01/23/2022 08:19:00 - INFO - codeparrot_training - Step 38316: {'lr': 0.0004897745260417884, 'samples': 919584, 'steps': 6385, 'loss/train': 3.198718786239624}
+01/23/2022 08:19:01 - INFO - codeparrot_training - Step 38317: {'lr': 0.0004897698937425566, 'samples': 919608, 'steps': 6386, 'loss/train': 4.239553451538086}
+01/23/2022 08:19:01 - INFO - codeparrot_training - Step 38318: {'lr': 0.0004897698937425566, 'samples': 919632, 'steps': 6386, 'loss/train': 5.582184791564941}
+01/23/2022 08:19:01 - INFO - codeparrot_training - Step 38319: {'lr': 0.0004897698937425566, 'samples': 919656, 'steps': 6386, 'loss/train': 1.9430623054504395}
+01/23/2022 08:19:02 - INFO - codeparrot_training - Step 38320: {'lr': 0.0004897698937425566, 'samples': 919680, 'steps': 6386, 'loss/train': 4.893961429595947}
+01/23/2022 08:19:02 - INFO - codeparrot_training - Step 38321: {'lr': 0.0004897698937425566, 'samples': 919704, 'steps': 6386, 'loss/train': 3.647883176803589}
+01/23/2022 08:19:02 - INFO - codeparrot_training - Step 38322: {'lr': 0.0004897698937425566, 'samples': 919728, 'steps': 6386, 'loss/train': 2.812589406967163}
+01/23/2022 08:19:05 - INFO - codeparrot_training - Step 38323: {'lr': 0.0004897652604162266, 'samples': 919752, 'steps': 6387, 'loss/train': 2.115694522857666}
+01/23/2022 08:19:06 - INFO - codeparrot_training - Step 38324: {'lr': 0.0004897652604162266, 'samples': 919776, 'steps': 6387, 'loss/train': 3.901871919631958}
+01/23/2022 08:19:06 - INFO - codeparrot_training - Step 38325: {'lr': 0.0004897652604162266, 'samples': 919800, 'steps': 6387, 'loss/train': 3.367323398590088}
+01/23/2022 08:19:06 - INFO - codeparrot_training - Step 38326: {'lr': 0.0004897652604162266, 'samples': 919824, 'steps': 6387, 'loss/train': 2.90391206741333}
+01/23/2022 08:19:07 - INFO - codeparrot_training - Step 38327: {'lr': 0.0004897652604162266, 'samples': 919848, 'steps': 6387, 'loss/train': 3.7901718616485596}
+01/23/2022 08:19:07 - INFO - codeparrot_training - Step 38328: {'lr': 0.0004897652604162266, 'samples': 919872, 'steps': 6387, 'loss/train': 0.7065055966377258}
+01/23/2022 08:19:08 - INFO - codeparrot_training - Step 38329: {'lr': 0.0004897606260628184, 'samples': 919896, 'steps': 6388, 'loss/train': 3.773815631866455}
+01/23/2022 08:19:08 - INFO - codeparrot_training - Step 38330: {'lr': 0.0004897606260628184, 'samples': 919920, 'steps': 6388, 'loss/train': 1.895822525024414}
+01/23/2022 08:19:08 - INFO - codeparrot_training - Step 38331: {'lr': 0.0004897606260628184, 'samples': 919944, 'steps': 6388, 'loss/train': 3.0650455951690674}
+01/23/2022 08:19:09 - INFO - codeparrot_training - Step 38332: {'lr': 0.0004897606260628184, 'samples': 919968, 'steps': 6388, 'loss/train': 3.735569477081299}
+01/23/2022 08:19:09 - INFO - codeparrot_training - Step 38333: {'lr': 0.0004897606260628184, 'samples': 919992, 'steps': 6388, 'loss/train': 3.3240294456481934}
+01/23/2022 08:19:09 - INFO - codeparrot_training - Step 38334: {'lr': 0.0004897606260628184, 'samples': 920016, 'steps': 6388, 'loss/train': 3.309536933898926}
+01/23/2022 08:19:10 - INFO - codeparrot_training - Step 38335: {'lr': 0.0004897559906823518, 'samples': 920040, 'steps': 6389, 'loss/train': 2.9099175930023193}
+01/23/2022 08:19:10 - INFO - codeparrot_training - Step 38336: {'lr': 0.0004897559906823518, 'samples': 920064, 'steps': 6389, 'loss/train': 3.2360429763793945}
+01/23/2022 08:19:11 - INFO - codeparrot_training - Step 38337: {'lr': 0.0004897559906823518, 'samples': 920088, 'steps': 6389, 'loss/train': 2.7861416339874268}
+01/23/2022 08:19:11 - INFO - codeparrot_training - Step 38338: {'lr': 0.0004897559906823518, 'samples': 920112, 'steps': 6389, 'loss/train': 2.9347474575042725}
+01/23/2022 08:19:11 - INFO - codeparrot_training - Step 38339: {'lr': 0.0004897559906823518, 'samples': 920136, 'steps': 6389, 'loss/train': 0.7073172330856323}
+01/23/2022 08:19:12 - INFO - codeparrot_training - Step 38340: {'lr': 0.0004897559906823518, 'samples': 920160, 'steps': 6389, 'loss/train': 3.12640380859375}
+01/23/2022 08:19:12 - INFO - codeparrot_training - Step 38341: {'lr': 0.0004897513542748468, 'samples': 920184, 'steps': 6390, 'loss/train': 2.9760921001434326}
+01/23/2022 08:19:13 - INFO - codeparrot_training - Step 38342: {'lr': 0.0004897513542748468, 'samples': 920208, 'steps': 6390, 'loss/train': 2.9951353073120117}
+01/23/2022 08:19:13 - INFO - codeparrot_training - Step 38343: {'lr': 0.0004897513542748468, 'samples': 920232, 'steps': 6390, 'loss/train': 2.866187810897827}
+01/23/2022 08:19:13 - INFO - codeparrot_training - Step 38344: {'lr': 0.0004897513542748468, 'samples': 920256, 'steps': 6390, 'loss/train': 3.9797213077545166}
+01/23/2022 08:19:14 - INFO - codeparrot_training - Step 38345: {'lr': 0.0004897513542748468, 'samples': 920280, 'steps': 6390, 'loss/train': 1.828201174736023}
+01/23/2022 08:19:14 - INFO - codeparrot_training - Step 38346: {'lr': 0.0004897513542748468, 'samples': 920304, 'steps': 6390, 'loss/train': 2.424567461013794}
+01/23/2022 08:19:14 - INFO - codeparrot_training - Step 38347: {'lr': 0.0004897467168403231, 'samples': 920328, 'steps': 6391, 'loss/train': 2.7737667560577393}
+01/23/2022 08:19:15 - INFO - codeparrot_training - Step 38348: {'lr': 0.0004897467168403231, 'samples': 920352, 'steps': 6391, 'loss/train': 2.096569538116455}
+01/23/2022 08:19:15 - INFO - codeparrot_training - Step 38349: {'lr': 0.0004897467168403231, 'samples': 920376, 'steps': 6391, 'loss/train': 3.0174479484558105}
+01/23/2022 08:19:16 - INFO - codeparrot_training - Step 38350: {'lr': 0.0004897467168403231, 'samples': 920400, 'steps': 6391, 'loss/train': 3.7903034687042236}
+01/23/2022 08:19:16 - INFO - codeparrot_training - Step 38351: {'lr': 0.0004897467168403231, 'samples': 920424, 'steps': 6391, 'loss/train': 4.867925643920898}
+01/23/2022 08:19:16 - INFO - codeparrot_training - Step 38352: {'lr': 0.0004897467168403231, 'samples': 920448, 'steps': 6391, 'loss/train': 4.714549541473389}
+01/23/2022 08:19:17 - INFO - codeparrot_training - Step 38353: {'lr': 0.0004897420783788006, 'samples': 920472, 'steps': 6392, 'loss/train': 3.3681986331939697}
+01/23/2022 08:19:17 - INFO - codeparrot_training - Step 38354: {'lr': 0.0004897420783788006, 'samples': 920496, 'steps': 6392, 'loss/train': 4.766732215881348}
+01/23/2022 08:19:18 - INFO - codeparrot_training - Step 38355: {'lr': 0.0004897420783788006, 'samples': 920520, 'steps': 6392, 'loss/train': 2.7419822216033936}
+01/23/2022 08:19:18 - INFO - codeparrot_training - Step 38356: {'lr': 0.0004897420783788006, 'samples': 920544, 'steps': 6392, 'loss/train': 3.0971298217773438}
+01/23/2022 08:19:18 - INFO - codeparrot_training - Step 38357: {'lr': 0.0004897420783788006, 'samples': 920568, 'steps': 6392, 'loss/train': 2.9217193126678467}
+01/23/2022 08:19:19 - INFO - codeparrot_training - Step 38358: {'lr': 0.0004897420783788006, 'samples': 920592, 'steps': 6392, 'loss/train': 3.5454812049865723}
+01/23/2022 08:19:19 - INFO - codeparrot_training - Step 38359: {'lr': 0.0004897374388902991, 'samples': 920616, 'steps': 6393, 'loss/train': 3.4571266174316406}
+01/23/2022 08:19:19 - INFO - codeparrot_training - Step 38360: {'lr': 0.0004897374388902991, 'samples': 920640, 'steps': 6393, 'loss/train': 3.8262248039245605}
+01/23/2022 08:19:20 - INFO - codeparrot_training - Step 38361: {'lr': 0.0004897374388902991, 'samples': 920664, 'steps': 6393, 'loss/train': 2.710247039794922}
+01/23/2022 08:19:21 - INFO - codeparrot_training - Step 38362: {'lr': 0.0004897374388902991, 'samples': 920688, 'steps': 6393, 'loss/train': 3.402773380279541}
+01/23/2022 08:19:21 - INFO - codeparrot_training - Step 38363: {'lr': 0.0004897374388902991, 'samples': 920712, 'steps': 6393, 'loss/train': 3.657935857772827}
+01/23/2022 08:19:22 - INFO - codeparrot_training - Step 38364: {'lr': 0.0004897374388902991, 'samples': 920736, 'steps': 6393, 'loss/train': 3.7434728145599365}
+01/23/2022 08:19:22 - INFO - codeparrot_training - Step 38365: {'lr': 0.0004897327983748385, 'samples': 920760, 'steps': 6394, 'loss/train': 2.821279287338257}
+01/23/2022 08:19:22 - INFO - codeparrot_training - Step 38366: {'lr': 0.0004897327983748385, 'samples': 920784, 'steps': 6394, 'loss/train': 3.0058021545410156}
+01/23/2022 08:19:23 - INFO - codeparrot_training - Step 38367: {'lr': 0.0004897327983748385, 'samples': 920808, 'steps': 6394, 'loss/train': 4.460212230682373}
+01/23/2022 08:19:23 - INFO - codeparrot_training - Step 38368: {'lr': 0.0004897327983748385, 'samples': 920832, 'steps': 6394, 'loss/train': 2.9284350872039795}
+01/23/2022 08:19:23 - INFO - codeparrot_training - Step 38369: {'lr': 0.0004897327983748385, 'samples': 920856, 'steps': 6394, 'loss/train': 2.5279195308685303}
+01/23/2022 08:19:24 - INFO - codeparrot_training - Step 38370: {'lr': 0.0004897327983748385, 'samples': 920880, 'steps': 6394, 'loss/train': 2.188695192337036}
+01/23/2022 08:19:24 - INFO - codeparrot_training - Step 38371: {'lr': 0.0004897281568324387, 'samples': 920904, 'steps': 6395, 'loss/train': 4.589432716369629}
+01/23/2022 08:19:25 - INFO - codeparrot_training - Step 38372: {'lr': 0.0004897281568324387, 'samples': 920928, 'steps': 6395, 'loss/train': 3.624817132949829}
+01/23/2022 08:19:25 - INFO - codeparrot_training - Step 38373: {'lr': 0.0004897281568324387, 'samples': 920952, 'steps': 6395, 'loss/train': 2.4223146438598633}
+01/23/2022 08:19:25 - INFO - codeparrot_training - Step 38374: {'lr': 0.0004897281568324387, 'samples': 920976, 'steps': 6395, 'loss/train': 4.218161106109619}
+01/23/2022 08:19:26 - INFO - codeparrot_training - Step 38375: {'lr': 0.0004897281568324387, 'samples': 921000, 'steps': 6395, 'loss/train': 3.4335479736328125}
+01/23/2022 08:19:26 - INFO - codeparrot_training - Step 38376: {'lr': 0.0004897281568324387, 'samples': 921024, 'steps': 6395, 'loss/train': 3.7632296085357666}
+01/23/2022 08:19:27 - INFO - codeparrot_training - Step 38377: {'lr': 0.0004897235142631197, 'samples': 921048, 'steps': 6396, 'loss/train': 2.9170641899108887}
+01/23/2022 08:19:27 - INFO - codeparrot_training - Step 38378: {'lr': 0.0004897235142631197, 'samples': 921072, 'steps': 6396, 'loss/train': 2.7290849685668945}
+01/23/2022 08:19:27 - INFO - codeparrot_training - Step 38379: {'lr': 0.0004897235142631197, 'samples': 921096, 'steps': 6396, 'loss/train': 3.5781357288360596}
+01/23/2022 08:19:28 - INFO - codeparrot_training - Step 38380: {'lr': 0.0004897235142631197, 'samples': 921120, 'steps': 6396, 'loss/train': 3.7504210472106934}
+01/23/2022 08:19:28 - INFO - codeparrot_training - Step 38381: {'lr': 0.0004897235142631197, 'samples': 921144, 'steps': 6396, 'loss/train': 0.9954725503921509}
+01/23/2022 08:19:28 - INFO - codeparrot_training - Step 38382: {'lr': 0.0004897235142631197, 'samples': 921168, 'steps': 6396, 'loss/train': 2.6910829544067383}
+01/23/2022 08:19:29 - INFO - codeparrot_training - Step 38383: {'lr': 0.0004897188706669012, 'samples': 921192, 'steps': 6397, 'loss/train': 4.204830169677734}
+01/23/2022 08:19:29 - INFO - codeparrot_training - Step 38384: {'lr': 0.0004897188706669012, 'samples': 921216, 'steps': 6397, 'loss/train': 3.790212869644165}
+01/23/2022 08:19:30 - INFO - codeparrot_training - Step 38385: {'lr': 0.0004897188706669012, 'samples': 921240, 'steps': 6397, 'loss/train': 3.903050184249878}
+01/23/2022 08:19:30 - INFO - codeparrot_training - Step 38386: {'lr': 0.0004897188706669012, 'samples': 921264, 'steps': 6397, 'loss/train': 4.1228790283203125}
+01/23/2022 08:19:30 - INFO - codeparrot_training - Step 38387: {'lr': 0.0004897188706669012, 'samples': 921288, 'steps': 6397, 'loss/train': 3.3859288692474365}
+01/23/2022 08:19:31 - INFO - codeparrot_training - Step 38388: {'lr': 0.0004897188706669012, 'samples': 921312, 'steps': 6397, 'loss/train': 2.5242364406585693}
+01/23/2022 08:19:31 - INFO - codeparrot_training - Step 38389: {'lr': 0.0004897142260438032, 'samples': 921336, 'steps': 6398, 'loss/train': 3.329318046569824}
+01/23/2022 08:19:32 - INFO - codeparrot_training - Step 38390: {'lr': 0.0004897142260438032, 'samples': 921360, 'steps': 6398, 'loss/train': 3.468360185623169}
+01/23/2022 08:19:32 - INFO - codeparrot_training - Step 38391: {'lr': 0.0004897142260438032, 'samples': 921384, 'steps': 6398, 'loss/train': 2.0514976978302}
+01/23/2022 08:19:32 - INFO - codeparrot_training - Step 38392: {'lr': 0.0004897142260438032, 'samples': 921408, 'steps': 6398, 'loss/train': 4.041840553283691}
+01/23/2022 08:19:33 - INFO - codeparrot_training - Step 38393: {'lr': 0.0004897142260438032, 'samples': 921432, 'steps': 6398, 'loss/train': 1.5091480016708374}
+01/23/2022 08:19:33 - INFO - codeparrot_training - Step 38394: {'lr': 0.0004897142260438032, 'samples': 921456, 'steps': 6398, 'loss/train': 3.4002459049224854}
+01/23/2022 08:19:33 - INFO - codeparrot_training - Step 38395: {'lr': 0.0004897095803938456, 'samples': 921480, 'steps': 6399, 'loss/train': 3.8550713062286377}
+01/23/2022 08:19:34 - INFO - codeparrot_training - Step 38396: {'lr': 0.0004897095803938456, 'samples': 921504, 'steps': 6399, 'loss/train': 1.8075330257415771}
+01/23/2022 08:19:34 - INFO - codeparrot_training - Step 38397: {'lr': 0.0004897095803938456, 'samples': 921528, 'steps': 6399, 'loss/train': 3.37608003616333}
+01/23/2022 08:19:35 - INFO - codeparrot_training - Step 38398: {'lr': 0.0004897095803938456, 'samples': 921552, 'steps': 6399, 'loss/train': 4.054450988769531}
+01/23/2022 08:19:35 - INFO - codeparrot_training - Step 38399: {'lr': 0.0004897095803938456, 'samples': 921576, 'steps': 6399, 'loss/train': 3.0495097637176514}
+01/23/2022 08:19:35 - INFO - codeparrot_training - Step 38400: {'lr': 0.0004897095803938456, 'samples': 921600, 'steps': 6399, 'loss/train': 3.109722375869751}
+01/23/2022 08:19:36 - INFO - codeparrot_training - Step 38401: {'lr': 0.0004897049337170483, 'samples': 921624, 'steps': 6400, 'loss/train': 3.437558650970459}
+01/23/2022 08:19:36 - INFO - codeparrot_training - Step 38402: {'lr': 0.0004897049337170483, 'samples': 921648, 'steps': 6400, 'loss/train': 1.9653011560440063}
+01/23/2022 08:19:37 - INFO - codeparrot_training - Step 38403: {'lr': 0.0004897049337170483, 'samples': 921672, 'steps': 6400, 'loss/train': 3.120359420776367}
+01/23/2022 08:19:38 - INFO - codeparrot_training - Step 38404: {'lr': 0.0004897049337170483, 'samples': 921696, 'steps': 6400, 'loss/train': 1.7035903930664062}
+01/23/2022 08:19:38 - INFO - codeparrot_training - Step 38405: {'lr': 0.0004897049337170483, 'samples': 921720, 'steps': 6400, 'loss/train': 3.94996976852417}
+01/23/2022 08:19:38 - INFO - codeparrot_training - Step 38406: {'lr': 0.0004897049337170483, 'samples': 921744, 'steps': 6400, 'loss/train': 2.6755385398864746}
+01/23/2022 08:19:39 - INFO - codeparrot_training - Step 38407: {'lr': 0.0004897002860134311, 'samples': 921768, 'steps': 6401, 'loss/train': 1.738555669784546}
+01/23/2022 08:19:39 - INFO - codeparrot_training - Step 38408: {'lr': 0.0004897002860134311, 'samples': 921792, 'steps': 6401, 'loss/train': 3.7592666149139404}
+01/23/2022 08:19:39 - INFO - codeparrot_training - Step 38409: {'lr': 0.0004897002860134311, 'samples': 921816, 'steps': 6401, 'loss/train': 2.126542568206787}
+01/23/2022 08:19:40 - INFO - codeparrot_training - Step 38410: {'lr': 0.0004897002860134311, 'samples': 921840, 'steps': 6401, 'loss/train': 2.54447340965271}
+01/23/2022 08:19:40 - INFO - codeparrot_training - Step 38411: {'lr': 0.0004897002860134311, 'samples': 921864, 'steps': 6401, 'loss/train': 3.299975872039795}
+01/23/2022 08:19:41 - INFO - codeparrot_training - Step 38412: {'lr': 0.0004897002860134311, 'samples': 921888, 'steps': 6401, 'loss/train': 2.648731231689453}
+01/23/2022 08:19:41 - INFO - codeparrot_training - Step 38413: {'lr': 0.0004896956372830141, 'samples': 921912, 'steps': 6402, 'loss/train': 3.3456318378448486}
+01/23/2022 08:19:41 - INFO - codeparrot_training - Step 38414: {'lr': 0.0004896956372830141, 'samples': 921936, 'steps': 6402, 'loss/train': 5.407870769500732}
+01/23/2022 08:19:42 - INFO - codeparrot_training - Step 38415: {'lr': 0.0004896956372830141, 'samples': 921960, 'steps': 6402, 'loss/train': 2.3927364349365234}
+01/23/2022 08:19:42 - INFO - codeparrot_training - Step 38416: {'lr': 0.0004896956372830141, 'samples': 921984, 'steps': 6402, 'loss/train': 3.847862482070923}
+01/23/2022 08:19:42 - INFO - codeparrot_training - Step 38417: {'lr': 0.0004896956372830141, 'samples': 922008, 'steps': 6402, 'loss/train': 2.7571890354156494}
+01/23/2022 08:19:43 - INFO - codeparrot_training - Step 38418: {'lr': 0.0004896956372830141, 'samples': 922032, 'steps': 6402, 'loss/train': 3.442195177078247}
+01/23/2022 08:19:43 - INFO - codeparrot_training - Step 38419: {'lr': 0.000489690987525817, 'samples': 922056, 'steps': 6403, 'loss/train': 3.2620439529418945}
+01/23/2022 08:19:44 - INFO - codeparrot_training - Step 38420: {'lr': 0.000489690987525817, 'samples': 922080, 'steps': 6403, 'loss/train': 4.353015899658203}
+01/23/2022 08:19:44 - INFO - codeparrot_training - Step 38421: {'lr': 0.000489690987525817, 'samples': 922104, 'steps': 6403, 'loss/train': 2.735203266143799}
+01/23/2022 08:19:44 - INFO - codeparrot_training - Step 38422: {'lr': 0.000489690987525817, 'samples': 922128, 'steps': 6403, 'loss/train': 3.422060489654541}
+01/23/2022 08:19:45 - INFO - codeparrot_training - Step 38423: {'lr': 0.000489690987525817, 'samples': 922152, 'steps': 6403, 'loss/train': 3.638707399368286}
+01/23/2022 08:19:45 - INFO - codeparrot_training - Step 38424: {'lr': 0.000489690987525817, 'samples': 922176, 'steps': 6403, 'loss/train': 2.691129446029663}
+01/23/2022 08:19:46 - INFO - codeparrot_training - Step 38425: {'lr': 0.0004896863367418598, 'samples': 922200, 'steps': 6404, 'loss/train': 3.4518954753875732}
+01/23/2022 08:19:46 - INFO - codeparrot_training - Step 38426: {'lr': 0.0004896863367418598, 'samples': 922224, 'steps': 6404, 'loss/train': 3.4640414714813232}
+01/23/2022 08:19:46 - INFO - codeparrot_training - Step 38427: {'lr': 0.0004896863367418598, 'samples': 922248, 'steps': 6404, 'loss/train': 2.6739392280578613}
+01/23/2022 08:19:47 - INFO - codeparrot_training - Step 38428: {'lr': 0.0004896863367418598, 'samples': 922272, 'steps': 6404, 'loss/train': 2.650406837463379}
+01/23/2022 08:19:47 - INFO - codeparrot_training - Step 38429: {'lr': 0.0004896863367418598, 'samples': 922296, 'steps': 6404, 'loss/train': 2.770581007003784}
+01/23/2022 08:19:47 - INFO - codeparrot_training - Step 38430: {'lr': 0.0004896863367418598, 'samples': 922320, 'steps': 6404, 'loss/train': 3.720250368118286}
+01/23/2022 08:19:48 - INFO - codeparrot_training - Step 38431: {'lr': 0.0004896816849311625, 'samples': 922344, 'steps': 6405, 'loss/train': 3.3763067722320557}
+01/23/2022 08:19:48 - INFO - codeparrot_training - Step 38432: {'lr': 0.0004896816849311625, 'samples': 922368, 'steps': 6405, 'loss/train': 3.8325371742248535}
+01/23/2022 08:19:49 - INFO - codeparrot_training - Step 38433: {'lr': 0.0004896816849311625, 'samples': 922392, 'steps': 6405, 'loss/train': 1.8593730926513672}
+01/23/2022 08:19:49 - INFO - codeparrot_training - Step 38434: {'lr': 0.0004896816849311625, 'samples': 922416, 'steps': 6405, 'loss/train': 2.8974721431732178}
+01/23/2022 08:19:49 - INFO - codeparrot_training - Step 38435: {'lr': 0.0004896816849311625, 'samples': 922440, 'steps': 6405, 'loss/train': 3.4560301303863525}
+01/23/2022 08:19:50 - INFO - codeparrot_training - Step 38436: {'lr': 0.0004896816849311625, 'samples': 922464, 'steps': 6405, 'loss/train': 3.6362621784210205}
+01/23/2022 08:19:50 - INFO - codeparrot_training - Step 38437: {'lr': 0.000489677032093745, 'samples': 922488, 'steps': 6406, 'loss/train': 3.828505516052246}
+01/23/2022 08:19:51 - INFO - codeparrot_training - Step 38438: {'lr': 0.000489677032093745, 'samples': 922512, 'steps': 6406, 'loss/train': 3.600602865219116}
+01/23/2022 08:19:51 - INFO - codeparrot_training - Step 38439: {'lr': 0.000489677032093745, 'samples': 922536, 'steps': 6406, 'loss/train': 3.015441656112671}
+01/23/2022 08:19:51 - INFO - codeparrot_training - Step 38440: {'lr': 0.000489677032093745, 'samples': 922560, 'steps': 6406, 'loss/train': 2.250263214111328}
+01/23/2022 08:19:52 - INFO - codeparrot_training - Step 38441: {'lr': 0.000489677032093745, 'samples': 922584, 'steps': 6406, 'loss/train': 3.4914894104003906}
+01/23/2022 08:19:53 - INFO - codeparrot_training - Step 38442: {'lr': 0.000489677032093745, 'samples': 922608, 'steps': 6406, 'loss/train': 2.448352336883545}
+01/23/2022 08:19:53 - INFO - codeparrot_training - Step 38443: {'lr': 0.0004896723782296272, 'samples': 922632, 'steps': 6407, 'loss/train': 3.158888816833496}
+01/23/2022 08:19:53 - INFO - codeparrot_training - Step 38444: {'lr': 0.0004896723782296272, 'samples': 922656, 'steps': 6407, 'loss/train': 3.0646402835845947}
+01/23/2022 08:19:54 - INFO - codeparrot_training - Step 38445: {'lr': 0.0004896723782296272, 'samples': 922680, 'steps': 6407, 'loss/train': 3.146850347518921}
+01/23/2022 08:19:54 - INFO - codeparrot_training - Step 38446: {'lr': 0.0004896723782296272, 'samples': 922704, 'steps': 6407, 'loss/train': 3.4048757553100586}
+01/23/2022 08:19:55 - INFO - codeparrot_training - Step 38447: {'lr': 0.0004896723782296272, 'samples': 922728, 'steps': 6407, 'loss/train': 3.1598033905029297}
+01/23/2022 08:19:55 - INFO - codeparrot_training - Step 38448: {'lr': 0.0004896723782296272, 'samples': 922752, 'steps': 6407, 'loss/train': 4.509325981140137}
+01/23/2022 08:19:55 - INFO - codeparrot_training - Step 38449: {'lr': 0.0004896677233388289, 'samples': 922776, 'steps': 6408, 'loss/train': 0.8184714317321777}
+01/23/2022 08:19:56 - INFO - codeparrot_training - Step 38450: {'lr': 0.0004896677233388289, 'samples': 922800, 'steps': 6408, 'loss/train': 0.7016302347183228}
+01/23/2022 08:19:56 - INFO - codeparrot_training - Step 38451: {'lr': 0.0004896677233388289, 'samples': 922824, 'steps': 6408, 'loss/train': 3.621129274368286}
+01/23/2022 08:19:57 - INFO - codeparrot_training - Step 38452: {'lr': 0.0004896677233388289, 'samples': 922848, 'steps': 6408, 'loss/train': 3.0853524208068848}
+01/23/2022 08:19:57 - INFO - codeparrot_training - Step 38453: {'lr': 0.0004896677233388289, 'samples': 922872, 'steps': 6408, 'loss/train': 2.0322792530059814}
+01/23/2022 08:19:57 - INFO - codeparrot_training - Step 38454: {'lr': 0.0004896677233388289, 'samples': 922896, 'steps': 6408, 'loss/train': 3.7944869995117188}
+01/23/2022 08:19:58 - INFO - codeparrot_training - Step 38455: {'lr': 0.0004896630674213703, 'samples': 922920, 'steps': 6409, 'loss/train': 3.290149211883545}
+01/23/2022 08:19:58 - INFO - codeparrot_training - Step 38456: {'lr': 0.0004896630674213703, 'samples': 922944, 'steps': 6409, 'loss/train': 2.8184170722961426}
+01/23/2022 08:19:58 - INFO - codeparrot_training - Step 38457: {'lr': 0.0004896630674213703, 'samples': 922968, 'steps': 6409, 'loss/train': 2.3613405227661133}
+01/23/2022 08:19:59 - INFO - codeparrot_training - Step 38458: {'lr': 0.0004896630674213703, 'samples': 922992, 'steps': 6409, 'loss/train': 2.7129580974578857}
+01/23/2022 08:19:59 - INFO - codeparrot_training - Step 38459: {'lr': 0.0004896630674213703, 'samples': 923016, 'steps': 6409, 'loss/train': 3.9967260360717773}
+01/23/2022 08:20:00 - INFO - codeparrot_training - Step 38460: {'lr': 0.0004896630674213703, 'samples': 923040, 'steps': 6409, 'loss/train': 2.645642042160034}
+01/23/2022 08:20:00 - INFO - codeparrot_training - Step 38461: {'lr': 0.0004896584104772712, 'samples': 923064, 'steps': 6410, 'loss/train': 2.5288162231445312}
+01/23/2022 08:20:00 - INFO - codeparrot_training - Step 38462: {'lr': 0.0004896584104772712, 'samples': 923088, 'steps': 6410, 'loss/train': 2.265665054321289}
+01/23/2022 08:20:01 - INFO - codeparrot_training - Step 38463: {'lr': 0.0004896584104772712, 'samples': 923112, 'steps': 6410, 'loss/train': 1.9029203653335571}
+01/23/2022 08:20:01 - INFO - codeparrot_training - Step 38464: {'lr': 0.0004896584104772712, 'samples': 923136, 'steps': 6410, 'loss/train': 2.34804630279541}
+01/23/2022 08:20:01 - INFO - codeparrot_training - Step 38465: {'lr': 0.0004896584104772712, 'samples': 923160, 'steps': 6410, 'loss/train': 2.641005516052246}
+01/23/2022 08:20:02 - INFO - codeparrot_training - Step 38466: {'lr': 0.0004896584104772712, 'samples': 923184, 'steps': 6410, 'loss/train': 4.033544540405273}
+01/23/2022 08:20:02 - INFO - codeparrot_training - Step 38467: {'lr': 0.0004896537525065516, 'samples': 923208, 'steps': 6411, 'loss/train': 2.8758199214935303}
+01/23/2022 08:20:03 - INFO - codeparrot_training - Step 38468: {'lr': 0.0004896537525065516, 'samples': 923232, 'steps': 6411, 'loss/train': 5.497586250305176}
+01/23/2022 08:20:03 - INFO - codeparrot_training - Step 38469: {'lr': 0.0004896537525065516, 'samples': 923256, 'steps': 6411, 'loss/train': 3.5338175296783447}
+01/23/2022 08:20:03 - INFO - codeparrot_training - Step 38470: {'lr': 0.0004896537525065516, 'samples': 923280, 'steps': 6411, 'loss/train': 3.2176976203918457}
+01/23/2022 08:20:04 - INFO - codeparrot_training - Step 38471: {'lr': 0.0004896537525065516, 'samples': 923304, 'steps': 6411, 'loss/train': 3.199712038040161}
+01/23/2022 08:20:04 - INFO - codeparrot_training - Step 38472: {'lr': 0.0004896537525065516, 'samples': 923328, 'steps': 6411, 'loss/train': 3.9885520935058594}
+01/23/2022 08:20:05 - INFO - codeparrot_training - Step 38473: {'lr': 0.0004896490935092314, 'samples': 923352, 'steps': 6412, 'loss/train': 3.560443162918091}
+01/23/2022 08:20:05 - INFO - codeparrot_training - Step 38474: {'lr': 0.0004896490935092314, 'samples': 923376, 'steps': 6412, 'loss/train': 2.850834846496582}
+01/23/2022 08:20:05 - INFO - codeparrot_training - Step 38475: {'lr': 0.0004896490935092314, 'samples': 923400, 'steps': 6412, 'loss/train': 3.622962236404419}
+01/23/2022 08:20:06 - INFO - codeparrot_training - Step 38476: {'lr': 0.0004896490935092314, 'samples': 923424, 'steps': 6412, 'loss/train': 3.512899160385132}
+01/23/2022 08:20:06 - INFO - codeparrot_training - Step 38477: {'lr': 0.0004896490935092314, 'samples': 923448, 'steps': 6412, 'loss/train': 3.5082924365997314}
+01/23/2022 08:20:06 - INFO - codeparrot_training - Step 38478: {'lr': 0.0004896490935092314, 'samples': 923472, 'steps': 6412, 'loss/train': 4.004765510559082}
+01/23/2022 08:20:07 - INFO - codeparrot_training - Step 38479: {'lr': 0.0004896444334853305, 'samples': 923496, 'steps': 6413, 'loss/train': 3.6021358966827393}
+01/23/2022 08:20:07 - INFO - codeparrot_training - Step 38480: {'lr': 0.0004896444334853305, 'samples': 923520, 'steps': 6413, 'loss/train': 3.8557119369506836}
+01/23/2022 08:20:08 - INFO - codeparrot_training - Step 38481: {'lr': 0.0004896444334853305, 'samples': 923544, 'steps': 6413, 'loss/train': 1.9954676628112793}
+01/23/2022 08:20:09 - INFO - codeparrot_training - Step 38482: {'lr': 0.0004896444334853305, 'samples': 923568, 'steps': 6413, 'loss/train': 3.110870122909546}
+01/23/2022 08:20:09 - INFO - codeparrot_training - Step 38483: {'lr': 0.0004896444334853305, 'samples': 923592, 'steps': 6413, 'loss/train': 2.348238229751587}
+01/23/2022 08:20:09 - INFO - codeparrot_training - Step 38484: {'lr': 0.0004896444334853305, 'samples': 923616, 'steps': 6413, 'loss/train': 3.0866992473602295}
+01/23/2022 08:20:10 - INFO - codeparrot_training - Step 38485: {'lr': 0.000489639772434869, 'samples': 923640, 'steps': 6414, 'loss/train': 2.9812467098236084}
+01/23/2022 08:20:10 - INFO - codeparrot_training - Step 38486: {'lr': 0.000489639772434869, 'samples': 923664, 'steps': 6414, 'loss/train': 2.8924295902252197}
+01/23/2022 08:20:11 - INFO - codeparrot_training - Step 38487: {'lr': 0.000489639772434869, 'samples': 923688, 'steps': 6414, 'loss/train': 2.717664957046509}
+01/23/2022 08:20:11 - INFO - codeparrot_training - Step 38488: {'lr': 0.000489639772434869, 'samples': 923712, 'steps': 6414, 'loss/train': 2.4982779026031494}
+01/23/2022 08:20:11 - INFO - codeparrot_training - Step 38489: {'lr': 0.000489639772434869, 'samples': 923736, 'steps': 6414, 'loss/train': 4.158077239990234}
+01/23/2022 08:20:12 - INFO - codeparrot_training - Step 38490: {'lr': 0.000489639772434869, 'samples': 923760, 'steps': 6414, 'loss/train': 2.3332924842834473}
+01/23/2022 08:20:12 - INFO - codeparrot_training - Step 38491: {'lr': 0.0004896351103578669, 'samples': 923784, 'steps': 6415, 'loss/train': 2.8611538410186768}
+01/23/2022 08:20:13 - INFO - codeparrot_training - Step 38492: {'lr': 0.0004896351103578669, 'samples': 923808, 'steps': 6415, 'loss/train': 2.7953031063079834}
+01/23/2022 08:20:13 - INFO - codeparrot_training - Step 38493: {'lr': 0.0004896351103578669, 'samples': 923832, 'steps': 6415, 'loss/train': 2.560634136199951}
+01/23/2022 08:20:13 - INFO - codeparrot_training - Step 38494: {'lr': 0.0004896351103578669, 'samples': 923856, 'steps': 6415, 'loss/train': 1.785752534866333}
+01/23/2022 08:20:14 - INFO - codeparrot_training - Step 38495: {'lr': 0.0004896351103578669, 'samples': 923880, 'steps': 6415, 'loss/train': 2.830796957015991}
+01/23/2022 08:20:14 - INFO - codeparrot_training - Step 38496: {'lr': 0.0004896351103578669, 'samples': 923904, 'steps': 6415, 'loss/train': 2.9894838333129883}
+01/23/2022 08:20:14 - INFO - codeparrot_training - Step 38497: {'lr': 0.0004896304472543439, 'samples': 923928, 'steps': 6416, 'loss/train': 3.6107749938964844}
+01/23/2022 08:20:15 - INFO - codeparrot_training - Step 38498: {'lr': 0.0004896304472543439, 'samples': 923952, 'steps': 6416, 'loss/train': 3.672435998916626}
+01/23/2022 08:20:15 - INFO - codeparrot_training - Step 38499: {'lr': 0.0004896304472543439, 'samples': 923976, 'steps': 6416, 'loss/train': 3.703963041305542}
+01/23/2022 08:20:16 - INFO - codeparrot_training - Step 38500: {'lr': 0.0004896304472543439, 'samples': 924000, 'steps': 6416, 'loss/train': 4.177247047424316}
+01/23/2022 08:20:16 - INFO - codeparrot_training - Step 38501: {'lr': 0.0004896304472543439, 'samples': 924024, 'steps': 6416, 'loss/train': 3.934448003768921}
+01/23/2022 08:20:16 - INFO - codeparrot_training - Step 38502: {'lr': 0.0004896304472543439, 'samples': 924048, 'steps': 6416, 'loss/train': 2.049797534942627}
+01/23/2022 08:20:17 - INFO - codeparrot_training - Step 38503: {'lr': 0.0004896257831243204, 'samples': 924072, 'steps': 6417, 'loss/train': 1.2120749950408936}
+01/23/2022 08:20:17 - INFO - codeparrot_training - Step 38504: {'lr': 0.0004896257831243204, 'samples': 924096, 'steps': 6417, 'loss/train': 3.468775510787964}
+01/23/2022 08:20:17 - INFO - codeparrot_training - Step 38505: {'lr': 0.0004896257831243204, 'samples': 924120, 'steps': 6417, 'loss/train': 2.940701723098755}
+01/23/2022 08:20:18 - INFO - codeparrot_training - Step 38506: {'lr': 0.0004896257831243204, 'samples': 924144, 'steps': 6417, 'loss/train': 3.3503055572509766}
+01/23/2022 08:20:18 - INFO - codeparrot_training - Step 38507: {'lr': 0.0004896257831243204, 'samples': 924168, 'steps': 6417, 'loss/train': 3.0831961631774902}
+01/23/2022 08:20:19 - INFO - codeparrot_training - Step 38508: {'lr': 0.0004896257831243204, 'samples': 924192, 'steps': 6417, 'loss/train': 3.5401618480682373}
+01/23/2022 08:20:19 - INFO - codeparrot_training - Step 38509: {'lr': 0.0004896211179678159, 'samples': 924216, 'steps': 6418, 'loss/train': 3.349616289138794}
+01/23/2022 08:20:19 - INFO - codeparrot_training - Step 38510: {'lr': 0.0004896211179678159, 'samples': 924240, 'steps': 6418, 'loss/train': 2.7678651809692383}
+01/23/2022 08:20:20 - INFO - codeparrot_training - Step 38511: {'lr': 0.0004896211179678159, 'samples': 924264, 'steps': 6418, 'loss/train': 3.395984172821045}
+01/23/2022 08:20:20 - INFO - codeparrot_training - Step 38512: {'lr': 0.0004896211179678159, 'samples': 924288, 'steps': 6418, 'loss/train': 3.203035354614258}
+01/23/2022 08:20:21 - INFO - codeparrot_training - Step 38513: {'lr': 0.0004896211179678159, 'samples': 924312, 'steps': 6418, 'loss/train': 2.2347936630249023}
+01/23/2022 08:20:21 - INFO - codeparrot_training - Step 38514: {'lr': 0.0004896211179678159, 'samples': 924336, 'steps': 6418, 'loss/train': 3.4052281379699707}
+01/23/2022 08:20:21 - INFO - codeparrot_training - Step 38515: {'lr': 0.0004896164517848508, 'samples': 924360, 'steps': 6419, 'loss/train': 3.0344576835632324}
+01/23/2022 08:20:22 - INFO - codeparrot_training - Step 38516: {'lr': 0.0004896164517848508, 'samples': 924384, 'steps': 6419, 'loss/train': 3.3292481899261475}
+01/23/2022 08:20:22 - INFO - codeparrot_training - Step 38517: {'lr': 0.0004896164517848508, 'samples': 924408, 'steps': 6419, 'loss/train': 2.2974610328674316}
+01/23/2022 08:20:22 - INFO - codeparrot_training - Step 38518: {'lr': 0.0004896164517848508, 'samples': 924432, 'steps': 6419, 'loss/train': 2.9930477142333984}
+01/23/2022 08:20:23 - INFO - codeparrot_training - Step 38519: {'lr': 0.0004896164517848508, 'samples': 924456, 'steps': 6419, 'loss/train': 0.8060670495033264}
+01/23/2022 08:20:26 - INFO - codeparrot_training - Step 38520: {'lr': 0.0004896164517848508, 'samples': 924480, 'steps': 6419, 'loss/train': 3.8644213676452637}
+01/23/2022 08:20:26 - INFO - codeparrot_training - Step 38521: {'lr': 0.0004896117845754448, 'samples': 924504, 'steps': 6420, 'loss/train': 3.49403977394104}
+01/23/2022 08:20:26 - INFO - codeparrot_training - Step 38522: {'lr': 0.0004896117845754448, 'samples': 924528, 'steps': 6420, 'loss/train': 2.257190465927124}
+01/23/2022 08:20:27 - INFO - codeparrot_training - Step 38523: {'lr': 0.0004896117845754448, 'samples': 924552, 'steps': 6420, 'loss/train': 3.8117308616638184}
+01/23/2022 08:20:27 - INFO - codeparrot_training - Step 38524: {'lr': 0.0004896117845754448, 'samples': 924576, 'steps': 6420, 'loss/train': 3.6020941734313965}
+01/23/2022 08:20:28 - INFO - codeparrot_training - Step 38525: {'lr': 0.0004896117845754448, 'samples': 924600, 'steps': 6420, 'loss/train': 2.9462890625}
+01/23/2022 08:20:28 - INFO - codeparrot_training - Step 38526: {'lr': 0.0004896117845754448, 'samples': 924624, 'steps': 6420, 'loss/train': 4.199103355407715}
+01/23/2022 08:20:28 - INFO - codeparrot_training - Step 38527: {'lr': 0.0004896071163396179, 'samples': 924648, 'steps': 6421, 'loss/train': 3.1073575019836426}
+01/23/2022 08:20:29 - INFO - codeparrot_training - Step 38528: {'lr': 0.0004896071163396179, 'samples': 924672, 'steps': 6421, 'loss/train': 2.4626340866088867}
+01/23/2022 08:20:29 - INFO - codeparrot_training - Step 38529: {'lr': 0.0004896071163396179, 'samples': 924696, 'steps': 6421, 'loss/train': 2.2542309761047363}
+01/23/2022 08:20:29 - INFO - codeparrot_training - Step 38530: {'lr': 0.0004896071163396179, 'samples': 924720, 'steps': 6421, 'loss/train': 3.2180063724517822}
+01/23/2022 08:20:30 - INFO - codeparrot_training - Step 38531: {'lr': 0.0004896071163396179, 'samples': 924744, 'steps': 6421, 'loss/train': 3.293732166290283}
+01/23/2022 08:20:30 - INFO - codeparrot_training - Step 38532: {'lr': 0.0004896071163396179, 'samples': 924768, 'steps': 6421, 'loss/train': 1.9640439748764038}
+01/23/2022 08:20:31 - INFO - codeparrot_training - Step 38533: {'lr': 0.0004896024470773904, 'samples': 924792, 'steps': 6422, 'loss/train': 3.292327880859375}
+01/23/2022 08:20:31 - INFO - codeparrot_training - Step 38534: {'lr': 0.0004896024470773904, 'samples': 924816, 'steps': 6422, 'loss/train': 2.1230931282043457}
+01/23/2022 08:20:31 - INFO - codeparrot_training - Step 38535: {'lr': 0.0004896024470773904, 'samples': 924840, 'steps': 6422, 'loss/train': 3.578763246536255}
+01/23/2022 08:20:32 - INFO - codeparrot_training - Step 38536: {'lr': 0.0004896024470773904, 'samples': 924864, 'steps': 6422, 'loss/train': 2.6155927181243896}
+01/23/2022 08:20:32 - INFO - codeparrot_training - Step 38537: {'lr': 0.0004896024470773904, 'samples': 924888, 'steps': 6422, 'loss/train': 2.9970855712890625}
+01/23/2022 08:20:32 - INFO - codeparrot_training - Step 38538: {'lr': 0.0004896024470773904, 'samples': 924912, 'steps': 6422, 'loss/train': 2.795574188232422}
+01/23/2022 08:20:33 - INFO - codeparrot_training - Step 38539: {'lr': 0.000489597776788782, 'samples': 924936, 'steps': 6423, 'loss/train': 3.1235642433166504}
+01/23/2022 08:20:33 - INFO - codeparrot_training - Step 38540: {'lr': 0.000489597776788782, 'samples': 924960, 'steps': 6423, 'loss/train': 2.2072269916534424}
+01/23/2022 08:20:34 - INFO - codeparrot_training - Step 38541: {'lr': 0.000489597776788782, 'samples': 924984, 'steps': 6423, 'loss/train': 3.453449249267578}
+01/23/2022 08:20:34 - INFO - codeparrot_training - Step 38542: {'lr': 0.000489597776788782, 'samples': 925008, 'steps': 6423, 'loss/train': 2.424166440963745}
+01/23/2022 08:20:34 - INFO - codeparrot_training - Step 38543: {'lr': 0.000489597776788782, 'samples': 925032, 'steps': 6423, 'loss/train': 3.473278045654297}
+01/23/2022 08:20:35 - INFO - codeparrot_training - Step 38544: {'lr': 0.000489597776788782, 'samples': 925056, 'steps': 6423, 'loss/train': 3.490731954574585}
+01/23/2022 08:20:35 - INFO - codeparrot_training - Step 38545: {'lr': 0.0004895931054738128, 'samples': 925080, 'steps': 6424, 'loss/train': 2.4195144176483154}
+01/23/2022 08:20:36 - INFO - codeparrot_training - Step 38546: {'lr': 0.0004895931054738128, 'samples': 925104, 'steps': 6424, 'loss/train': 2.926129102706909}
+01/23/2022 08:20:36 - INFO - codeparrot_training - Step 38547: {'lr': 0.0004895931054738128, 'samples': 925128, 'steps': 6424, 'loss/train': 3.7635037899017334}
+01/23/2022 08:20:36 - INFO - codeparrot_training - Step 38548: {'lr': 0.0004895931054738128, 'samples': 925152, 'steps': 6424, 'loss/train': 2.5586254596710205}
+01/23/2022 08:20:37 - INFO - codeparrot_training - Step 38549: {'lr': 0.0004895931054738128, 'samples': 925176, 'steps': 6424, 'loss/train': 2.809349775314331}
+01/23/2022 08:20:37 - INFO - codeparrot_training - Step 38550: {'lr': 0.0004895931054738128, 'samples': 925200, 'steps': 6424, 'loss/train': 2.7859790325164795}
+01/23/2022 08:20:37 - INFO - codeparrot_training - Step 38551: {'lr': 0.0004895884331325028, 'samples': 925224, 'steps': 6425, 'loss/train': 3.200096368789673}
+01/23/2022 08:20:38 - INFO - codeparrot_training - Step 38552: {'lr': 0.0004895884331325028, 'samples': 925248, 'steps': 6425, 'loss/train': 3.0678534507751465}
+01/23/2022 08:20:38 - INFO - codeparrot_training - Step 38553: {'lr': 0.0004895884331325028, 'samples': 925272, 'steps': 6425, 'loss/train': 2.9271562099456787}
+01/23/2022 08:20:39 - INFO - codeparrot_training - Step 38554: {'lr': 0.0004895884331325028, 'samples': 925296, 'steps': 6425, 'loss/train': 1.9573811292648315}
+01/23/2022 08:20:39 - INFO - codeparrot_training - Step 38555: {'lr': 0.0004895884331325028, 'samples': 925320, 'steps': 6425, 'loss/train': 3.424410104751587}
+01/23/2022 08:20:39 - INFO - codeparrot_training - Step 38556: {'lr': 0.0004895884331325028, 'samples': 925344, 'steps': 6425, 'loss/train': 3.1510493755340576}
+01/23/2022 08:20:40 - INFO - codeparrot_training - Step 38557: {'lr': 0.0004895837597648721, 'samples': 925368, 'steps': 6426, 'loss/train': 3.1267783641815186}
+01/23/2022 08:20:40 - INFO - codeparrot_training - Step 38558: {'lr': 0.0004895837597648721, 'samples': 925392, 'steps': 6426, 'loss/train': 3.7113561630249023}
+01/23/2022 08:20:41 - INFO - codeparrot_training - Step 38559: {'lr': 0.0004895837597648721, 'samples': 925416, 'steps': 6426, 'loss/train': 2.365629196166992}
+01/23/2022 08:20:42 - INFO - codeparrot_training - Step 38560: {'lr': 0.0004895837597648721, 'samples': 925440, 'steps': 6426, 'loss/train': 3.3740346431732178}
+01/23/2022 08:20:42 - INFO - codeparrot_training - Step 38561: {'lr': 0.0004895837597648721, 'samples': 925464, 'steps': 6426, 'loss/train': 3.880136013031006}
+01/23/2022 08:20:42 - INFO - codeparrot_training - Step 38562: {'lr': 0.0004895837597648721, 'samples': 925488, 'steps': 6426, 'loss/train': 3.612410306930542}
+01/23/2022 08:20:43 - INFO - codeparrot_training - Step 38563: {'lr': 0.0004895790853709406, 'samples': 925512, 'steps': 6427, 'loss/train': 4.080677509307861}
+01/23/2022 08:20:43 - INFO - codeparrot_training - Step 38564: {'lr': 0.0004895790853709406, 'samples': 925536, 'steps': 6427, 'loss/train': 3.1478230953216553}
+01/23/2022 08:20:44 - INFO - codeparrot_training - Step 38565: {'lr': 0.0004895790853709406, 'samples': 925560, 'steps': 6427, 'loss/train': 3.1517674922943115}
+01/23/2022 08:20:44 - INFO - codeparrot_training - Step 38566: {'lr': 0.0004895790853709406, 'samples': 925584, 'steps': 6427, 'loss/train': 3.0223915576934814}
+01/23/2022 08:20:44 - INFO - codeparrot_training - Step 38567: {'lr': 0.0004895790853709406, 'samples': 925608, 'steps': 6427, 'loss/train': 3.9839301109313965}
+01/23/2022 08:20:45 - INFO - codeparrot_training - Step 38568: {'lr': 0.0004895790853709406, 'samples': 925632, 'steps': 6427, 'loss/train': 4.302687644958496}
+01/23/2022 08:20:45 - INFO - codeparrot_training - Step 38569: {'lr': 0.0004895744099507284, 'samples': 925656, 'steps': 6428, 'loss/train': 3.4419965744018555}
+01/23/2022 08:20:45 - INFO - codeparrot_training - Step 38570: {'lr': 0.0004895744099507284, 'samples': 925680, 'steps': 6428, 'loss/train': 2.9579648971557617}
+01/23/2022 08:20:46 - INFO - codeparrot_training - Step 38571: {'lr': 0.0004895744099507284, 'samples': 925704, 'steps': 6428, 'loss/train': 2.6866073608398438}
+01/23/2022 08:20:46 - INFO - codeparrot_training - Step 38572: {'lr': 0.0004895744099507284, 'samples': 925728, 'steps': 6428, 'loss/train': 3.1704931259155273}
+01/23/2022 08:20:47 - INFO - codeparrot_training - Step 38573: {'lr': 0.0004895744099507284, 'samples': 925752, 'steps': 6428, 'loss/train': 3.405298948287964}
+01/23/2022 08:20:47 - INFO - codeparrot_training - Step 38574: {'lr': 0.0004895744099507284, 'samples': 925776, 'steps': 6428, 'loss/train': 3.490849733352661}
+01/23/2022 08:20:47 - INFO - codeparrot_training - Step 38575: {'lr': 0.0004895697335042555, 'samples': 925800, 'steps': 6429, 'loss/train': 4.220956802368164}
+01/23/2022 08:20:48 - INFO - codeparrot_training - Step 38576: {'lr': 0.0004895697335042555, 'samples': 925824, 'steps': 6429, 'loss/train': 2.70302414894104}
+01/23/2022 08:20:48 - INFO - codeparrot_training - Step 38577: {'lr': 0.0004895697335042555, 'samples': 925848, 'steps': 6429, 'loss/train': 3.03184175491333}
+01/23/2022 08:20:48 - INFO - codeparrot_training - Step 38578: {'lr': 0.0004895697335042555, 'samples': 925872, 'steps': 6429, 'loss/train': 4.234567165374756}
+01/23/2022 08:20:49 - INFO - codeparrot_training - Step 38579: {'lr': 0.0004895697335042555, 'samples': 925896, 'steps': 6429, 'loss/train': 3.9803802967071533}
+01/23/2022 08:20:49 - INFO - codeparrot_training - Step 38580: {'lr': 0.0004895697335042555, 'samples': 925920, 'steps': 6429, 'loss/train': 3.409986972808838}
+01/23/2022 08:20:50 - INFO - codeparrot_training - Step 38581: {'lr': 0.0004895650560315419, 'samples': 925944, 'steps': 6430, 'loss/train': 3.393418788909912}
+01/23/2022 08:20:50 - INFO - codeparrot_training - Step 38582: {'lr': 0.0004895650560315419, 'samples': 925968, 'steps': 6430, 'loss/train': 2.7597835063934326}
+01/23/2022 08:20:50 - INFO - codeparrot_training - Step 38583: {'lr': 0.0004895650560315419, 'samples': 925992, 'steps': 6430, 'loss/train': 3.362752914428711}
+01/23/2022 08:20:51 - INFO - codeparrot_training - Step 38584: {'lr': 0.0004895650560315419, 'samples': 926016, 'steps': 6430, 'loss/train': 0.7582119107246399}
+01/23/2022 08:20:51 - INFO - codeparrot_training - Step 38585: {'lr': 0.0004895650560315419, 'samples': 926040, 'steps': 6430, 'loss/train': 3.1978487968444824}
+01/23/2022 08:20:52 - INFO - codeparrot_training - Step 38586: {'lr': 0.0004895650560315419, 'samples': 926064, 'steps': 6430, 'loss/train': 3.7584080696105957}
+01/23/2022 08:20:52 - INFO - codeparrot_training - Step 38587: {'lr': 0.0004895603775326077, 'samples': 926088, 'steps': 6431, 'loss/train': 3.758002758026123}
+01/23/2022 08:20:52 - INFO - codeparrot_training - Step 38588: {'lr': 0.0004895603775326077, 'samples': 926112, 'steps': 6431, 'loss/train': 2.6015372276306152}
+01/23/2022 08:20:53 - INFO - codeparrot_training - Step 38589: {'lr': 0.0004895603775326077, 'samples': 926136, 'steps': 6431, 'loss/train': 4.30220890045166}
+01/23/2022 08:20:53 - INFO - codeparrot_training - Step 38590: {'lr': 0.0004895603775326077, 'samples': 926160, 'steps': 6431, 'loss/train': 3.584336519241333}
+01/23/2022 08:20:53 - INFO - codeparrot_training - Step 38591: {'lr': 0.0004895603775326077, 'samples': 926184, 'steps': 6431, 'loss/train': 2.842710494995117}
+01/23/2022 08:20:54 - INFO - codeparrot_training - Step 38592: {'lr': 0.0004895603775326077, 'samples': 926208, 'steps': 6431, 'loss/train': 2.9876625537872314}
+01/23/2022 08:20:54 - INFO - codeparrot_training - Step 38593: {'lr': 0.0004895556980074729, 'samples': 926232, 'steps': 6432, 'loss/train': 3.6657047271728516}
+01/23/2022 08:20:55 - INFO - codeparrot_training - Step 38594: {'lr': 0.0004895556980074729, 'samples': 926256, 'steps': 6432, 'loss/train': 3.232124090194702}
+01/23/2022 08:20:55 - INFO - codeparrot_training - Step 38595: {'lr': 0.0004895556980074729, 'samples': 926280, 'steps': 6432, 'loss/train': 3.256593704223633}
+01/23/2022 08:20:55 - INFO - codeparrot_training - Step 38596: {'lr': 0.0004895556980074729, 'samples': 926304, 'steps': 6432, 'loss/train': 2.4894967079162598}
+01/23/2022 08:20:56 - INFO - codeparrot_training - Step 38597: {'lr': 0.0004895556980074729, 'samples': 926328, 'steps': 6432, 'loss/train': 3.7192599773406982}
+01/23/2022 08:20:57 - INFO - codeparrot_training - Step 38598: {'lr': 0.0004895556980074729, 'samples': 926352, 'steps': 6432, 'loss/train': 3.6583845615386963}
+01/23/2022 08:20:57 - INFO - codeparrot_training - Step 38599: {'lr': 0.0004895510174561576, 'samples': 926376, 'steps': 6433, 'loss/train': 3.421215772628784}
+01/23/2022 08:20:58 - INFO - codeparrot_training - Step 38600: {'lr': 0.0004895510174561576, 'samples': 926400, 'steps': 6433, 'loss/train': 3.3653078079223633}
+01/23/2022 08:20:58 - INFO - codeparrot_training - Step 38601: {'lr': 0.0004895510174561576, 'samples': 926424, 'steps': 6433, 'loss/train': 2.1523234844207764}
+01/23/2022 08:20:58 - INFO - codeparrot_training - Step 38602: {'lr': 0.0004895510174561576, 'samples': 926448, 'steps': 6433, 'loss/train': 3.7201128005981445}
+01/23/2022 08:20:59 - INFO - codeparrot_training - Step 38603: {'lr': 0.0004895510174561576, 'samples': 926472, 'steps': 6433, 'loss/train': 3.7799437046051025}
+01/23/2022 08:20:59 - INFO - codeparrot_training - Step 38604: {'lr': 0.0004895510174561576, 'samples': 926496, 'steps': 6433, 'loss/train': 4.25877046585083}
+01/23/2022 08:21:00 - INFO - codeparrot_training - Step 38605: {'lr': 0.0004895463358786818, 'samples': 926520, 'steps': 6434, 'loss/train': 3.2448890209198}
+01/23/2022 08:21:00 - INFO - codeparrot_training - Step 38606: {'lr': 0.0004895463358786818, 'samples': 926544, 'steps': 6434, 'loss/train': 2.1093509197235107}
+01/23/2022 08:21:00 - INFO - codeparrot_training - Step 38607: {'lr': 0.0004895463358786818, 'samples': 926568, 'steps': 6434, 'loss/train': 2.898498296737671}
+01/23/2022 08:21:01 - INFO - codeparrot_training - Step 38608: {'lr': 0.0004895463358786818, 'samples': 926592, 'steps': 6434, 'loss/train': 3.7478489875793457}
+01/23/2022 08:21:01 - INFO - codeparrot_training - Step 38609: {'lr': 0.0004895463358786818, 'samples': 926616, 'steps': 6434, 'loss/train': 2.1732358932495117}
+01/23/2022 08:21:01 - INFO - codeparrot_training - Step 38610: {'lr': 0.0004895463358786818, 'samples': 926640, 'steps': 6434, 'loss/train': 3.6119744777679443}
+01/23/2022 08:21:02 - INFO - codeparrot_training - Step 38611: {'lr': 0.0004895416532750655, 'samples': 926664, 'steps': 6435, 'loss/train': 3.4363930225372314}
+01/23/2022 08:21:02 - INFO - codeparrot_training - Step 38612: {'lr': 0.0004895416532750655, 'samples': 926688, 'steps': 6435, 'loss/train': 4.165652751922607}
+01/23/2022 08:21:03 - INFO - codeparrot_training - Step 38613: {'lr': 0.0004895416532750655, 'samples': 926712, 'steps': 6435, 'loss/train': 3.046131134033203}
+01/23/2022 08:21:03 - INFO - codeparrot_training - Step 38614: {'lr': 0.0004895416532750655, 'samples': 926736, 'steps': 6435, 'loss/train': 2.8011281490325928}
+01/23/2022 08:21:03 - INFO - codeparrot_training - Step 38615: {'lr': 0.0004895416532750655, 'samples': 926760, 'steps': 6435, 'loss/train': 3.104243040084839}
+01/23/2022 08:21:04 - INFO - codeparrot_training - Step 38616: {'lr': 0.0004895416532750655, 'samples': 926784, 'steps': 6435, 'loss/train': 3.338986873626709}
+01/23/2022 08:21:04 - INFO - codeparrot_training - Step 38617: {'lr': 0.0004895369696453289, 'samples': 926808, 'steps': 6436, 'loss/train': 2.971121311187744}
+01/23/2022 08:21:05 - INFO - codeparrot_training - Step 38618: {'lr': 0.0004895369696453289, 'samples': 926832, 'steps': 6436, 'loss/train': 2.7995452880859375}
+01/23/2022 08:21:05 - INFO - codeparrot_training - Step 38619: {'lr': 0.0004895369696453289, 'samples': 926856, 'steps': 6436, 'loss/train': 4.239180088043213}
+01/23/2022 08:21:05 - INFO - codeparrot_training - Step 38620: {'lr': 0.0004895369696453289, 'samples': 926880, 'steps': 6436, 'loss/train': 2.5431389808654785}
+01/23/2022 08:21:06 - INFO - codeparrot_training - Step 38621: {'lr': 0.0004895369696453289, 'samples': 926904, 'steps': 6436, 'loss/train': 2.4338581562042236}
+01/23/2022 08:21:06 - INFO - codeparrot_training - Step 38622: {'lr': 0.0004895369696453289, 'samples': 926928, 'steps': 6436, 'loss/train': 3.3858792781829834}
+01/23/2022 08:21:06 - INFO - codeparrot_training - Step 38623: {'lr': 0.0004895322849894918, 'samples': 926952, 'steps': 6437, 'loss/train': 2.42268705368042}
+01/23/2022 08:21:07 - INFO - codeparrot_training - Step 38624: {'lr': 0.0004895322849894918, 'samples': 926976, 'steps': 6437, 'loss/train': 1.8808778524398804}
+01/23/2022 08:21:07 - INFO - codeparrot_training - Step 38625: {'lr': 0.0004895322849894918, 'samples': 927000, 'steps': 6437, 'loss/train': 2.9845550060272217}
+01/23/2022 08:21:08 - INFO - codeparrot_training - Step 38626: {'lr': 0.0004895322849894918, 'samples': 927024, 'steps': 6437, 'loss/train': 3.61991286277771}
+01/23/2022 08:21:08 - INFO - codeparrot_training - Step 38627: {'lr': 0.0004895322849894918, 'samples': 927048, 'steps': 6437, 'loss/train': 3.6893298625946045}
+01/23/2022 08:21:08 - INFO - codeparrot_training - Step 38628: {'lr': 0.0004895322849894918, 'samples': 927072, 'steps': 6437, 'loss/train': 3.4081358909606934}
+01/23/2022 08:21:09 - INFO - codeparrot_training - Step 38629: {'lr': 0.0004895275993075747, 'samples': 927096, 'steps': 6438, 'loss/train': 3.3033933639526367}
+01/23/2022 08:21:09 - INFO - codeparrot_training - Step 38630: {'lr': 0.0004895275993075747, 'samples': 927120, 'steps': 6438, 'loss/train': 4.63369607925415}
+01/23/2022 08:21:10 - INFO - codeparrot_training - Step 38631: {'lr': 0.0004895275993075747, 'samples': 927144, 'steps': 6438, 'loss/train': 3.2487456798553467}
+01/23/2022 08:21:10 - INFO - codeparrot_training - Step 38632: {'lr': 0.0004895275993075747, 'samples': 927168, 'steps': 6438, 'loss/train': 4.9401116371154785}
+01/23/2022 08:21:10 - INFO - codeparrot_training - Step 38633: {'lr': 0.0004895275993075747, 'samples': 927192, 'steps': 6438, 'loss/train': 3.879523277282715}
+01/23/2022 08:21:11 - INFO - codeparrot_training - Step 38634: {'lr': 0.0004895275993075747, 'samples': 927216, 'steps': 6438, 'loss/train': 2.3761141300201416}
+01/23/2022 08:21:11 - INFO - codeparrot_training - Step 38635: {'lr': 0.0004895229125995973, 'samples': 927240, 'steps': 6439, 'loss/train': 2.904277801513672}
+01/23/2022 08:21:11 - INFO - codeparrot_training - Step 38636: {'lr': 0.0004895229125995973, 'samples': 927264, 'steps': 6439, 'loss/train': 3.4088070392608643}
+01/23/2022 08:21:12 - INFO - codeparrot_training - Step 38637: {'lr': 0.0004895229125995973, 'samples': 927288, 'steps': 6439, 'loss/train': 2.9543440341949463}
+01/23/2022 08:21:13 - INFO - codeparrot_training - Step 38638: {'lr': 0.0004895229125995973, 'samples': 927312, 'steps': 6439, 'loss/train': 3.366260528564453}
+01/23/2022 08:21:13 - INFO - codeparrot_training - Step 38639: {'lr': 0.0004895229125995973, 'samples': 927336, 'steps': 6439, 'loss/train': 3.288713216781616}
+01/23/2022 08:21:14 - INFO - codeparrot_training - Step 38640: {'lr': 0.0004895229125995973, 'samples': 927360, 'steps': 6439, 'loss/train': 3.1284618377685547}
+01/23/2022 08:21:14 - INFO - codeparrot_training - Step 38641: {'lr': 0.0004895182248655798, 'samples': 927384, 'steps': 6440, 'loss/train': 2.260723829269409}
+01/23/2022 08:21:14 - INFO - codeparrot_training - Step 38642: {'lr': 0.0004895182248655798, 'samples': 927408, 'steps': 6440, 'loss/train': 2.9992129802703857}
+01/23/2022 08:21:15 - INFO - codeparrot_training - Step 38643: {'lr': 0.0004895182248655798, 'samples': 927432, 'steps': 6440, 'loss/train': 3.40120005607605}
+01/23/2022 08:21:15 - INFO - codeparrot_training - Step 38644: {'lr': 0.0004895182248655798, 'samples': 927456, 'steps': 6440, 'loss/train': 2.573195695877075}
+01/23/2022 08:21:16 - INFO - codeparrot_training - Step 38645: {'lr': 0.0004895182248655798, 'samples': 927480, 'steps': 6440, 'loss/train': 3.4622621536254883}
+01/23/2022 08:21:16 - INFO - codeparrot_training - Step 38646: {'lr': 0.0004895182248655798, 'samples': 927504, 'steps': 6440, 'loss/train': 2.585956335067749}
+01/23/2022 08:21:16 - INFO - codeparrot_training - Step 38647: {'lr': 0.0004895135361055422, 'samples': 927528, 'steps': 6441, 'loss/train': 3.2081198692321777}
+01/23/2022 08:21:17 - INFO - codeparrot_training - Step 38648: {'lr': 0.0004895135361055422, 'samples': 927552, 'steps': 6441, 'loss/train': 2.4956743717193604}
+01/23/2022 08:21:17 - INFO - codeparrot_training - Step 38649: {'lr': 0.0004895135361055422, 'samples': 927576, 'steps': 6441, 'loss/train': 3.2778337001800537}
+01/23/2022 08:21:17 - INFO - codeparrot_training - Step 38650: {'lr': 0.0004895135361055422, 'samples': 927600, 'steps': 6441, 'loss/train': 3.030963897705078}
+01/23/2022 08:21:18 - INFO - codeparrot_training - Step 38651: {'lr': 0.0004895135361055422, 'samples': 927624, 'steps': 6441, 'loss/train': 2.4501302242279053}
+01/23/2022 08:21:18 - INFO - codeparrot_training - Step 38652: {'lr': 0.0004895135361055422, 'samples': 927648, 'steps': 6441, 'loss/train': 3.030735492706299}
+01/23/2022 08:21:19 - INFO - codeparrot_training - Step 38653: {'lr': 0.0004895088463195049, 'samples': 927672, 'steps': 6442, 'loss/train': 4.021695137023926}
+01/23/2022 08:21:19 - INFO - codeparrot_training - Step 38654: {'lr': 0.0004895088463195049, 'samples': 927696, 'steps': 6442, 'loss/train': 3.222456693649292}
+01/23/2022 08:21:19 - INFO - codeparrot_training - Step 38655: {'lr': 0.0004895088463195049, 'samples': 927720, 'steps': 6442, 'loss/train': 3.353102922439575}
+01/23/2022 08:21:20 - INFO - codeparrot_training - Step 38656: {'lr': 0.0004895088463195049, 'samples': 927744, 'steps': 6442, 'loss/train': 3.7340989112854004}
+01/23/2022 08:21:20 - INFO - codeparrot_training - Step 38657: {'lr': 0.0004895088463195049, 'samples': 927768, 'steps': 6442, 'loss/train': 2.489584445953369}
+01/23/2022 08:21:20 - INFO - codeparrot_training - Step 38658: {'lr': 0.0004895088463195049, 'samples': 927792, 'steps': 6442, 'loss/train': 4.9511590003967285}
+01/23/2022 08:21:21 - INFO - codeparrot_training - Step 38659: {'lr': 0.0004895041555074875, 'samples': 927816, 'steps': 6443, 'loss/train': 3.5631635189056396}
+01/23/2022 08:21:21 - INFO - codeparrot_training - Step 38660: {'lr': 0.0004895041555074875, 'samples': 927840, 'steps': 6443, 'loss/train': 2.3088326454162598}
+01/23/2022 08:21:22 - INFO - codeparrot_training - Step 38661: {'lr': 0.0004895041555074875, 'samples': 927864, 'steps': 6443, 'loss/train': 4.202453136444092}
+01/23/2022 08:21:22 - INFO - codeparrot_training - Step 38662: {'lr': 0.0004895041555074875, 'samples': 927888, 'steps': 6443, 'loss/train': 3.4680111408233643}
+01/23/2022 08:21:22 - INFO - codeparrot_training - Step 38663: {'lr': 0.0004895041555074875, 'samples': 927912, 'steps': 6443, 'loss/train': 2.6426072120666504}
+01/23/2022 08:21:23 - INFO - codeparrot_training - Step 38664: {'lr': 0.0004895041555074875, 'samples': 927936, 'steps': 6443, 'loss/train': 3.908531427383423}
+01/23/2022 08:21:23 - INFO - codeparrot_training - Step 38665: {'lr': 0.0004894994636695105, 'samples': 927960, 'steps': 6444, 'loss/train': 3.633702278137207}
+01/23/2022 08:21:24 - INFO - codeparrot_training - Step 38666: {'lr': 0.0004894994636695105, 'samples': 927984, 'steps': 6444, 'loss/train': 1.6818146705627441}
+01/23/2022 08:21:24 - INFO - codeparrot_training - Step 38667: {'lr': 0.0004894994636695105, 'samples': 928008, 'steps': 6444, 'loss/train': 2.845937728881836}
+01/23/2022 08:21:24 - INFO - codeparrot_training - Step 38668: {'lr': 0.0004894994636695105, 'samples': 928032, 'steps': 6444, 'loss/train': 3.2154648303985596}
+01/23/2022 08:21:25 - INFO - codeparrot_training - Step 38669: {'lr': 0.0004894994636695105, 'samples': 928056, 'steps': 6444, 'loss/train': 2.974132776260376}
+01/23/2022 08:21:25 - INFO - codeparrot_training - Step 38670: {'lr': 0.0004894994636695105, 'samples': 928080, 'steps': 6444, 'loss/train': 1.859479546546936}
+01/23/2022 08:21:26 - INFO - codeparrot_training - Step 38671: {'lr': 0.0004894947708055938, 'samples': 928104, 'steps': 6445, 'loss/train': 2.6400537490844727}
+01/23/2022 08:21:26 - INFO - codeparrot_training - Step 38672: {'lr': 0.0004894947708055938, 'samples': 928128, 'steps': 6445, 'loss/train': 2.618072032928467}
+01/23/2022 08:21:26 - INFO - codeparrot_training - Step 38673: {'lr': 0.0004894947708055938, 'samples': 928152, 'steps': 6445, 'loss/train': 3.060577392578125}
+01/23/2022 08:21:27 - INFO - codeparrot_training - Step 38674: {'lr': 0.0004894947708055938, 'samples': 928176, 'steps': 6445, 'loss/train': 3.207160711288452}
+01/23/2022 08:21:27 - INFO - codeparrot_training - Step 38675: {'lr': 0.0004894947708055938, 'samples': 928200, 'steps': 6445, 'loss/train': 2.945213794708252}
+01/23/2022 08:21:27 - INFO - codeparrot_training - Step 38676: {'lr': 0.0004894947708055938, 'samples': 928224, 'steps': 6445, 'loss/train': 2.8500306606292725}
+01/23/2022 08:21:28 - INFO - codeparrot_training - Step 38677: {'lr': 0.0004894900769157576, 'samples': 928248, 'steps': 6446, 'loss/train': 2.946276903152466}
+01/23/2022 08:21:28 - INFO - codeparrot_training - Step 38678: {'lr': 0.0004894900769157576, 'samples': 928272, 'steps': 6446, 'loss/train': 3.6711316108703613}
+01/23/2022 08:21:29 - INFO - codeparrot_training - Step 38679: {'lr': 0.0004894900769157576, 'samples': 928296, 'steps': 6446, 'loss/train': 3.2632744312286377}
+01/23/2022 08:21:30 - INFO - codeparrot_training - Step 38680: {'lr': 0.0004894900769157576, 'samples': 928320, 'steps': 6446, 'loss/train': 3.036179542541504}
+01/23/2022 08:21:30 - INFO - codeparrot_training - Step 38681: {'lr': 0.0004894900769157576, 'samples': 928344, 'steps': 6446, 'loss/train': 1.4203373193740845}
+01/23/2022 08:21:30 - INFO - codeparrot_training - Step 38682: {'lr': 0.0004894900769157576, 'samples': 928368, 'steps': 6446, 'loss/train': 3.568352460861206}
+01/23/2022 08:21:31 - INFO - codeparrot_training - Step 38683: {'lr': 0.0004894853820000219, 'samples': 928392, 'steps': 6447, 'loss/train': 1.9830576181411743}
+01/23/2022 08:21:31 - INFO - codeparrot_training - Step 38684: {'lr': 0.0004894853820000219, 'samples': 928416, 'steps': 6447, 'loss/train': 2.6040167808532715}
+01/23/2022 08:21:31 - INFO - codeparrot_training - Step 38685: {'lr': 0.0004894853820000219, 'samples': 928440, 'steps': 6447, 'loss/train': 1.9001253843307495}
+01/23/2022 08:21:32 - INFO - codeparrot_training - Step 38686: {'lr': 0.0004894853820000219, 'samples': 928464, 'steps': 6447, 'loss/train': 4.56664514541626}
+01/23/2022 08:21:32 - INFO - codeparrot_training - Step 38687: {'lr': 0.0004894853820000219, 'samples': 928488, 'steps': 6447, 'loss/train': 2.9695072174072266}
+01/23/2022 08:21:33 - INFO - codeparrot_training - Step 38688: {'lr': 0.0004894853820000219, 'samples': 928512, 'steps': 6447, 'loss/train': 3.432199478149414}
+01/23/2022 08:21:33 - INFO - codeparrot_training - Step 38689: {'lr': 0.000489480686058407, 'samples': 928536, 'steps': 6448, 'loss/train': 2.876699924468994}
+01/23/2022 08:21:33 - INFO - codeparrot_training - Step 38690: {'lr': 0.000489480686058407, 'samples': 928560, 'steps': 6448, 'loss/train': 2.5087904930114746}
+01/23/2022 08:21:34 - INFO - codeparrot_training - Step 38691: {'lr': 0.000489480686058407, 'samples': 928584, 'steps': 6448, 'loss/train': 3.2632784843444824}
+01/23/2022 08:21:34 - INFO - codeparrot_training - Step 38692: {'lr': 0.000489480686058407, 'samples': 928608, 'steps': 6448, 'loss/train': 3.5358057022094727}
+01/23/2022 08:21:35 - INFO - codeparrot_training - Step 38693: {'lr': 0.000489480686058407, 'samples': 928632, 'steps': 6448, 'loss/train': 4.442020893096924}
+01/23/2022 08:21:35 - INFO - codeparrot_training - Step 38694: {'lr': 0.000489480686058407, 'samples': 928656, 'steps': 6448, 'loss/train': 2.527543306350708}
+01/23/2022 08:21:35 - INFO - codeparrot_training - Step 38695: {'lr': 0.0004894759890909326, 'samples': 928680, 'steps': 6449, 'loss/train': 3.222386360168457}
+01/23/2022 08:21:36 - INFO - codeparrot_training - Step 38696: {'lr': 0.0004894759890909326, 'samples': 928704, 'steps': 6449, 'loss/train': 3.0297696590423584}
+01/23/2022 08:21:36 - INFO - codeparrot_training - Step 38697: {'lr': 0.0004894759890909326, 'samples': 928728, 'steps': 6449, 'loss/train': 2.2519493103027344}
+01/23/2022 08:21:36 - INFO - codeparrot_training - Step 38698: {'lr': 0.0004894759890909326, 'samples': 928752, 'steps': 6449, 'loss/train': 3.0304200649261475}
+01/23/2022 08:21:37 - INFO - codeparrot_training - Step 38699: {'lr': 0.0004894759890909326, 'samples': 928776, 'steps': 6449, 'loss/train': 4.497818470001221}
+01/23/2022 08:21:37 - INFO - codeparrot_training - Step 38700: {'lr': 0.0004894759890909326, 'samples': 928800, 'steps': 6449, 'loss/train': 3.114743709564209}
+01/23/2022 08:21:38 - INFO - codeparrot_training - Step 38701: {'lr': 0.0004894712910976193, 'samples': 928824, 'steps': 6450, 'loss/train': 2.871375560760498}
+01/23/2022 08:21:38 - INFO - codeparrot_training - Step 38702: {'lr': 0.0004894712910976193, 'samples': 928848, 'steps': 6450, 'loss/train': 3.49603271484375}
+01/23/2022 08:21:38 - INFO - codeparrot_training - Step 38703: {'lr': 0.0004894712910976193, 'samples': 928872, 'steps': 6450, 'loss/train': 3.664781093597412}
+01/23/2022 08:21:39 - INFO - codeparrot_training - Step 38704: {'lr': 0.0004894712910976193, 'samples': 928896, 'steps': 6450, 'loss/train': 3.522747278213501}
+01/23/2022 08:21:39 - INFO - codeparrot_training - Step 38705: {'lr': 0.0004894712910976193, 'samples': 928920, 'steps': 6450, 'loss/train': 3.421035051345825}
+01/23/2022 08:21:40 - INFO - codeparrot_training - Step 38706: {'lr': 0.0004894712910976193, 'samples': 928944, 'steps': 6450, 'loss/train': 2.2627625465393066}
+01/23/2022 08:21:40 - INFO - codeparrot_training - Step 38707: {'lr': 0.000489466592078487, 'samples': 928968, 'steps': 6451, 'loss/train': 3.1178436279296875}
+01/23/2022 08:21:40 - INFO - codeparrot_training - Step 38708: {'lr': 0.000489466592078487, 'samples': 928992, 'steps': 6451, 'loss/train': 2.8943874835968018}
+01/23/2022 08:21:41 - INFO - codeparrot_training - Step 38709: {'lr': 0.000489466592078487, 'samples': 929016, 'steps': 6451, 'loss/train': 1.2158780097961426}
+01/23/2022 08:21:41 - INFO - codeparrot_training - Step 38710: {'lr': 0.000489466592078487, 'samples': 929040, 'steps': 6451, 'loss/train': 3.2443318367004395}
+01/23/2022 08:21:41 - INFO - codeparrot_training - Step 38711: {'lr': 0.000489466592078487, 'samples': 929064, 'steps': 6451, 'loss/train': 3.7234346866607666}
+01/23/2022 08:21:42 - INFO - codeparrot_training - Step 38712: {'lr': 0.000489466592078487, 'samples': 929088, 'steps': 6451, 'loss/train': 3.49601411819458}
+01/23/2022 08:21:42 - INFO - codeparrot_training - Step 38713: {'lr': 0.0004894618920335558, 'samples': 929112, 'steps': 6452, 'loss/train': 2.524228572845459}
+01/23/2022 08:21:43 - INFO - codeparrot_training - Step 38714: {'lr': 0.0004894618920335558, 'samples': 929136, 'steps': 6452, 'loss/train': 2.3118021488189697}
+01/23/2022 08:21:43 - INFO - codeparrot_training - Step 38715: {'lr': 0.0004894618920335558, 'samples': 929160, 'steps': 6452, 'loss/train': 3.080047845840454}
+01/23/2022 08:21:43 - INFO - codeparrot_training - Step 38716: {'lr': 0.0004894618920335558, 'samples': 929184, 'steps': 6452, 'loss/train': 3.631383180618286}
+01/23/2022 08:21:44 - INFO - codeparrot_training - Step 38717: {'lr': 0.0004894618920335558, 'samples': 929208, 'steps': 6452, 'loss/train': 1.6138545274734497}
+01/23/2022 08:21:44 - INFO - codeparrot_training - Step 38718: {'lr': 0.0004894618920335558, 'samples': 929232, 'steps': 6452, 'loss/train': 7.053964138031006}
+01/23/2022 08:21:47 - INFO - codeparrot_training - Step 38719: {'lr': 0.000489457190962846, 'samples': 929256, 'steps': 6453, 'loss/train': 2.253422260284424}
+01/23/2022 08:21:47 - INFO - codeparrot_training - Step 38720: {'lr': 0.000489457190962846, 'samples': 929280, 'steps': 6453, 'loss/train': 1.8404394388198853}
+01/23/2022 08:21:48 - INFO - codeparrot_training - Step 38721: {'lr': 0.000489457190962846, 'samples': 929304, 'steps': 6453, 'loss/train': 3.641253709793091}
+01/23/2022 08:21:48 - INFO - codeparrot_training - Step 38722: {'lr': 0.000489457190962846, 'samples': 929328, 'steps': 6453, 'loss/train': 3.620940923690796}
+01/23/2022 08:21:48 - INFO - codeparrot_training - Step 38723: {'lr': 0.000489457190962846, 'samples': 929352, 'steps': 6453, 'loss/train': 5.104555130004883}
+01/23/2022 08:21:49 - INFO - codeparrot_training - Step 38724: {'lr': 0.000489457190962846, 'samples': 929376, 'steps': 6453, 'loss/train': 2.5646190643310547}
+01/23/2022 08:21:49 - INFO - codeparrot_training - Step 38725: {'lr': 0.0004894524888663776, 'samples': 929400, 'steps': 6454, 'loss/train': 2.7683680057525635}
+01/23/2022 08:21:50 - INFO - codeparrot_training - Step 38726: {'lr': 0.0004894524888663776, 'samples': 929424, 'steps': 6454, 'loss/train': 4.013469219207764}
+01/23/2022 08:21:50 - INFO - codeparrot_training - Step 38727: {'lr': 0.0004894524888663776, 'samples': 929448, 'steps': 6454, 'loss/train': 2.62135648727417}
+01/23/2022 08:21:50 - INFO - codeparrot_training - Step 38728: {'lr': 0.0004894524888663776, 'samples': 929472, 'steps': 6454, 'loss/train': 3.5233776569366455}
+01/23/2022 08:21:51 - INFO - codeparrot_training - Step 38729: {'lr': 0.0004894524888663776, 'samples': 929496, 'steps': 6454, 'loss/train': 2.8874356746673584}
+01/23/2022 08:21:51 - INFO - codeparrot_training - Step 38730: {'lr': 0.0004894524888663776, 'samples': 929520, 'steps': 6454, 'loss/train': 3.8089911937713623}
+01/23/2022 08:21:52 - INFO - codeparrot_training - Step 38731: {'lr': 0.0004894477857441707, 'samples': 929544, 'steps': 6455, 'loss/train': 2.94277286529541}
+01/23/2022 08:21:52 - INFO - codeparrot_training - Step 38732: {'lr': 0.0004894477857441707, 'samples': 929568, 'steps': 6455, 'loss/train': 2.7593722343444824}
+01/23/2022 08:21:52 - INFO - codeparrot_training - Step 38733: {'lr': 0.0004894477857441707, 'samples': 929592, 'steps': 6455, 'loss/train': 2.7769253253936768}
+01/23/2022 08:21:53 - INFO - codeparrot_training - Step 38734: {'lr': 0.0004894477857441707, 'samples': 929616, 'steps': 6455, 'loss/train': 1.8843140602111816}
+01/23/2022 08:21:53 - INFO - codeparrot_training - Step 38735: {'lr': 0.0004894477857441707, 'samples': 929640, 'steps': 6455, 'loss/train': 3.697472095489502}
+01/23/2022 08:21:53 - INFO - codeparrot_training - Step 38736: {'lr': 0.0004894477857441707, 'samples': 929664, 'steps': 6455, 'loss/train': 2.9846062660217285}
+01/23/2022 08:21:54 - INFO - codeparrot_training - Step 38737: {'lr': 0.0004894430815962456, 'samples': 929688, 'steps': 6456, 'loss/train': 3.162978172302246}
+01/23/2022 08:21:54 - INFO - codeparrot_training - Step 38738: {'lr': 0.0004894430815962456, 'samples': 929712, 'steps': 6456, 'loss/train': 2.875458240509033}
+01/23/2022 08:21:55 - INFO - codeparrot_training - Step 38739: {'lr': 0.0004894430815962456, 'samples': 929736, 'steps': 6456, 'loss/train': 3.0631210803985596}
+01/23/2022 08:21:55 - INFO - codeparrot_training - Step 38740: {'lr': 0.0004894430815962456, 'samples': 929760, 'steps': 6456, 'loss/train': 4.025586128234863}
+01/23/2022 08:21:55 - INFO - codeparrot_training - Step 38741: {'lr': 0.0004894430815962456, 'samples': 929784, 'steps': 6456, 'loss/train': 3.231593370437622}
+01/23/2022 08:21:56 - INFO - codeparrot_training - Step 38742: {'lr': 0.0004894430815962456, 'samples': 929808, 'steps': 6456, 'loss/train': 3.284590721130371}
+01/23/2022 08:21:56 - INFO - codeparrot_training - Step 38743: {'lr': 0.0004894383764226224, 'samples': 929832, 'steps': 6457, 'loss/train': 2.6115360260009766}
+01/23/2022 08:21:56 - INFO - codeparrot_training - Step 38744: {'lr': 0.0004894383764226224, 'samples': 929856, 'steps': 6457, 'loss/train': 3.269838333129883}
+01/23/2022 08:21:57 - INFO - codeparrot_training - Step 38745: {'lr': 0.0004894383764226224, 'samples': 929880, 'steps': 6457, 'loss/train': 2.5544137954711914}
+01/23/2022 08:21:57 - INFO - codeparrot_training - Step 38746: {'lr': 0.0004894383764226224, 'samples': 929904, 'steps': 6457, 'loss/train': 1.5495275259017944}
+01/23/2022 08:21:58 - INFO - codeparrot_training - Step 38747: {'lr': 0.0004894383764226224, 'samples': 929928, 'steps': 6457, 'loss/train': 4.4853434562683105}
+01/23/2022 08:21:58 - INFO - codeparrot_training - Step 38748: {'lr': 0.0004894383764226224, 'samples': 929952, 'steps': 6457, 'loss/train': 3.6249921321868896}
+01/23/2022 08:21:58 - INFO - codeparrot_training - Step 38749: {'lr': 0.0004894336702233212, 'samples': 929976, 'steps': 6458, 'loss/train': 2.5046372413635254}
+01/23/2022 08:21:59 - INFO - codeparrot_training - Step 38750: {'lr': 0.0004894336702233212, 'samples': 930000, 'steps': 6458, 'loss/train': 3.271158218383789}
+01/23/2022 08:21:59 - INFO - codeparrot_training - Step 38751: {'lr': 0.0004894336702233212, 'samples': 930024, 'steps': 6458, 'loss/train': 3.9243521690368652}
+01/23/2022 08:22:00 - INFO - codeparrot_training - Step 38752: {'lr': 0.0004894336702233212, 'samples': 930048, 'steps': 6458, 'loss/train': 3.7152371406555176}
+01/23/2022 08:22:00 - INFO - codeparrot_training - Step 38753: {'lr': 0.0004894336702233212, 'samples': 930072, 'steps': 6458, 'loss/train': 3.092691659927368}
+01/23/2022 08:22:00 - INFO - codeparrot_training - Step 38754: {'lr': 0.0004894336702233212, 'samples': 930096, 'steps': 6458, 'loss/train': 2.274843692779541}
+01/23/2022 08:22:01 - INFO - codeparrot_training - Step 38755: {'lr': 0.0004894289629983621, 'samples': 930120, 'steps': 6459, 'loss/train': 1.93436861038208}
+01/23/2022 08:22:01 - INFO - codeparrot_training - Step 38756: {'lr': 0.0004894289629983621, 'samples': 930144, 'steps': 6459, 'loss/train': 3.2652578353881836}
+01/23/2022 08:22:01 - INFO - codeparrot_training - Step 38757: {'lr': 0.0004894289629983621, 'samples': 930168, 'steps': 6459, 'loss/train': 2.7741079330444336}
+01/23/2022 08:22:02 - INFO - codeparrot_training - Step 38758: {'lr': 0.0004894289629983621, 'samples': 930192, 'steps': 6459, 'loss/train': 2.1075713634490967}
+01/23/2022 08:22:03 - INFO - codeparrot_training - Step 38759: {'lr': 0.0004894289629983621, 'samples': 930216, 'steps': 6459, 'loss/train': 5.013484001159668}
+01/23/2022 08:22:03 - INFO - codeparrot_training - Step 38760: {'lr': 0.0004894289629983621, 'samples': 930240, 'steps': 6459, 'loss/train': 2.4943110942840576}
+01/23/2022 08:22:04 - INFO - codeparrot_training - Step 38761: {'lr': 0.0004894242547477654, 'samples': 930264, 'steps': 6460, 'loss/train': 2.892388105392456}
+01/23/2022 08:22:04 - INFO - codeparrot_training - Step 38762: {'lr': 0.0004894242547477654, 'samples': 930288, 'steps': 6460, 'loss/train': 2.3516387939453125}
+01/23/2022 08:22:04 - INFO - codeparrot_training - Step 38763: {'lr': 0.0004894242547477654, 'samples': 930312, 'steps': 6460, 'loss/train': 1.3875378370285034}
+01/23/2022 08:22:05 - INFO - codeparrot_training - Step 38764: {'lr': 0.0004894242547477654, 'samples': 930336, 'steps': 6460, 'loss/train': 3.3765697479248047}
+01/23/2022 08:22:05 - INFO - codeparrot_training - Step 38765: {'lr': 0.0004894242547477654, 'samples': 930360, 'steps': 6460, 'loss/train': 4.305394649505615}
+01/23/2022 08:22:06 - INFO - codeparrot_training - Step 38766: {'lr': 0.0004894242547477654, 'samples': 930384, 'steps': 6460, 'loss/train': 3.340549945831299}
+01/23/2022 08:22:06 - INFO - codeparrot_training - Step 38767: {'lr': 0.0004894195454715512, 'samples': 930408, 'steps': 6461, 'loss/train': 3.0701870918273926}
+01/23/2022 08:22:06 - INFO - codeparrot_training - Step 38768: {'lr': 0.0004894195454715512, 'samples': 930432, 'steps': 6461, 'loss/train': 3.914731740951538}
+01/23/2022 08:22:07 - INFO - codeparrot_training - Step 38769: {'lr': 0.0004894195454715512, 'samples': 930456, 'steps': 6461, 'loss/train': 2.7940421104431152}
+01/23/2022 08:22:07 - INFO - codeparrot_training - Step 38770: {'lr': 0.0004894195454715512, 'samples': 930480, 'steps': 6461, 'loss/train': 3.8338637351989746}
+01/23/2022 08:22:07 - INFO - codeparrot_training - Step 38771: {'lr': 0.0004894195454715512, 'samples': 930504, 'steps': 6461, 'loss/train': 2.946382522583008}
+01/23/2022 08:22:08 - INFO - codeparrot_training - Step 38772: {'lr': 0.0004894195454715512, 'samples': 930528, 'steps': 6461, 'loss/train': 3.4350905418395996}
+01/23/2022 08:22:08 - INFO - codeparrot_training - Step 38773: {'lr': 0.0004894148351697398, 'samples': 930552, 'steps': 6462, 'loss/train': 3.000047445297241}
+01/23/2022 08:22:09 - INFO - codeparrot_training - Step 38774: {'lr': 0.0004894148351697398, 'samples': 930576, 'steps': 6462, 'loss/train': 2.512599468231201}
+01/23/2022 08:22:09 - INFO - codeparrot_training - Step 38775: {'lr': 0.0004894148351697398, 'samples': 930600, 'steps': 6462, 'loss/train': 2.1572980880737305}
+01/23/2022 08:22:09 - INFO - codeparrot_training - Step 38776: {'lr': 0.0004894148351697398, 'samples': 930624, 'steps': 6462, 'loss/train': 1.6169795989990234}
+01/23/2022 08:22:10 - INFO - codeparrot_training - Step 38777: {'lr': 0.0004894148351697398, 'samples': 930648, 'steps': 6462, 'loss/train': 1.7593575716018677}
+01/23/2022 08:22:10 - INFO - codeparrot_training - Step 38778: {'lr': 0.0004894148351697398, 'samples': 930672, 'steps': 6462, 'loss/train': 4.0119500160217285}
+01/23/2022 08:22:11 - INFO - codeparrot_training - Step 38779: {'lr': 0.0004894101238423512, 'samples': 930696, 'steps': 6463, 'loss/train': 3.3336851596832275}
+01/23/2022 08:22:11 - INFO - codeparrot_training - Step 38780: {'lr': 0.0004894101238423512, 'samples': 930720, 'steps': 6463, 'loss/train': 3.068653106689453}
+01/23/2022 08:22:11 - INFO - codeparrot_training - Step 38781: {'lr': 0.0004894101238423512, 'samples': 930744, 'steps': 6463, 'loss/train': 3.2185418605804443}
+01/23/2022 08:22:12 - INFO - codeparrot_training - Step 38782: {'lr': 0.0004894101238423512, 'samples': 930768, 'steps': 6463, 'loss/train': 3.457617998123169}
+01/23/2022 08:22:12 - INFO - codeparrot_training - Step 38783: {'lr': 0.0004894101238423512, 'samples': 930792, 'steps': 6463, 'loss/train': 2.8877882957458496}
+01/23/2022 08:22:12 - INFO - codeparrot_training - Step 38784: {'lr': 0.0004894101238423512, 'samples': 930816, 'steps': 6463, 'loss/train': 3.4338152408599854}
+01/23/2022 08:22:13 - INFO - codeparrot_training - Step 38785: {'lr': 0.0004894054114894055, 'samples': 930840, 'steps': 6464, 'loss/train': 3.553520679473877}
+01/23/2022 08:22:13 - INFO - codeparrot_training - Step 38786: {'lr': 0.0004894054114894055, 'samples': 930864, 'steps': 6464, 'loss/train': 2.9239444732666016}
+01/23/2022 08:22:14 - INFO - codeparrot_training - Step 38787: {'lr': 0.0004894054114894055, 'samples': 930888, 'steps': 6464, 'loss/train': 3.7048869132995605}
+01/23/2022 08:22:14 - INFO - codeparrot_training - Step 38788: {'lr': 0.0004894054114894055, 'samples': 930912, 'steps': 6464, 'loss/train': 3.0862619876861572}
+01/23/2022 08:22:14 - INFO - codeparrot_training - Step 38789: {'lr': 0.0004894054114894055, 'samples': 930936, 'steps': 6464, 'loss/train': 2.224236011505127}
+01/23/2022 08:22:15 - INFO - codeparrot_training - Step 38790: {'lr': 0.0004894054114894055, 'samples': 930960, 'steps': 6464, 'loss/train': 3.6132073402404785}
+01/23/2022 08:22:15 - INFO - codeparrot_training - Step 38791: {'lr': 0.0004894006981109232, 'samples': 930984, 'steps': 6465, 'loss/train': 3.8501741886138916}
+01/23/2022 08:22:16 - INFO - codeparrot_training - Step 38792: {'lr': 0.0004894006981109232, 'samples': 931008, 'steps': 6465, 'loss/train': 3.3962981700897217}
+01/23/2022 08:22:16 - INFO - codeparrot_training - Step 38793: {'lr': 0.0004894006981109232, 'samples': 931032, 'steps': 6465, 'loss/train': 3.1540756225585938}
+01/23/2022 08:22:16 - INFO - codeparrot_training - Step 38794: {'lr': 0.0004894006981109232, 'samples': 931056, 'steps': 6465, 'loss/train': 4.454113006591797}
+01/23/2022 08:22:17 - INFO - codeparrot_training - Step 38795: {'lr': 0.0004894006981109232, 'samples': 931080, 'steps': 6465, 'loss/train': 3.3532018661499023}
+01/23/2022 08:22:17 - INFO - codeparrot_training - Step 38796: {'lr': 0.0004894006981109232, 'samples': 931104, 'steps': 6465, 'loss/train': 3.504692792892456}
+01/23/2022 08:22:17 - INFO - codeparrot_training - Step 38797: {'lr': 0.0004893959837069243, 'samples': 931128, 'steps': 6466, 'loss/train': 1.6526727676391602}
+01/23/2022 08:22:19 - INFO - codeparrot_training - Step 38798: {'lr': 0.0004893959837069243, 'samples': 931152, 'steps': 6466, 'loss/train': 4.315201759338379}
+01/23/2022 08:22:19 - INFO - codeparrot_training - Step 38799: {'lr': 0.0004893959837069243, 'samples': 931176, 'steps': 6466, 'loss/train': 0.5433962345123291}
+01/23/2022 08:22:19 - INFO - codeparrot_training - Step 38800: {'lr': 0.0004893959837069243, 'samples': 931200, 'steps': 6466, 'loss/train': 3.0524849891662598}
+01/23/2022 08:22:20 - INFO - codeparrot_training - Step 38801: {'lr': 0.0004893959837069243, 'samples': 931224, 'steps': 6466, 'loss/train': 2.952908992767334}
+01/23/2022 08:22:20 - INFO - codeparrot_training - Step 38802: {'lr': 0.0004893959837069243, 'samples': 931248, 'steps': 6466, 'loss/train': 2.41058349609375}
+01/23/2022 08:22:20 - INFO - codeparrot_training - Step 38803: {'lr': 0.0004893912682774291, 'samples': 931272, 'steps': 6467, 'loss/train': 2.8284265995025635}
+01/23/2022 08:22:21 - INFO - codeparrot_training - Step 38804: {'lr': 0.0004893912682774291, 'samples': 931296, 'steps': 6467, 'loss/train': 4.447571277618408}
+01/23/2022 08:22:21 - INFO - codeparrot_training - Step 38805: {'lr': 0.0004893912682774291, 'samples': 931320, 'steps': 6467, 'loss/train': 3.862549066543579}
+01/23/2022 08:22:22 - INFO - codeparrot_training - Step 38806: {'lr': 0.0004893912682774291, 'samples': 931344, 'steps': 6467, 'loss/train': 3.3529248237609863}
+01/23/2022 08:22:22 - INFO - codeparrot_training - Step 38807: {'lr': 0.0004893912682774291, 'samples': 931368, 'steps': 6467, 'loss/train': 3.9694972038269043}
+01/23/2022 08:22:22 - INFO - codeparrot_training - Step 38808: {'lr': 0.0004893912682774291, 'samples': 931392, 'steps': 6467, 'loss/train': 4.756775856018066}
+01/23/2022 08:22:23 - INFO - codeparrot_training - Step 38809: {'lr': 0.0004893865518224576, 'samples': 931416, 'steps': 6468, 'loss/train': 1.2304844856262207}
+01/23/2022 08:22:23 - INFO - codeparrot_training - Step 38810: {'lr': 0.0004893865518224576, 'samples': 931440, 'steps': 6468, 'loss/train': 2.8902103900909424}
+01/23/2022 08:22:24 - INFO - codeparrot_training - Step 38811: {'lr': 0.0004893865518224576, 'samples': 931464, 'steps': 6468, 'loss/train': 2.8426499366760254}
+01/23/2022 08:22:24 - INFO - codeparrot_training - Step 38812: {'lr': 0.0004893865518224576, 'samples': 931488, 'steps': 6468, 'loss/train': 2.4604926109313965}
+01/23/2022 08:22:24 - INFO - codeparrot_training - Step 38813: {'lr': 0.0004893865518224576, 'samples': 931512, 'steps': 6468, 'loss/train': 2.4158525466918945}
+01/23/2022 08:22:25 - INFO - codeparrot_training - Step 38814: {'lr': 0.0004893865518224576, 'samples': 931536, 'steps': 6468, 'loss/train': 3.110269784927368}
+01/23/2022 08:22:25 - INFO - codeparrot_training - Step 38815: {'lr': 0.0004893818343420302, 'samples': 931560, 'steps': 6469, 'loss/train': 3.9228572845458984}
+01/23/2022 08:22:25 - INFO - codeparrot_training - Step 38816: {'lr': 0.0004893818343420302, 'samples': 931584, 'steps': 6469, 'loss/train': 3.4267594814300537}
+01/23/2022 08:22:26 - INFO - codeparrot_training - Step 38817: {'lr': 0.0004893818343420302, 'samples': 931608, 'steps': 6469, 'loss/train': 3.169466495513916}
+01/23/2022 08:22:26 - INFO - codeparrot_training - Step 38818: {'lr': 0.0004893818343420302, 'samples': 931632, 'steps': 6469, 'loss/train': 3.0702908039093018}
+01/23/2022 08:22:27 - INFO - codeparrot_training - Step 38819: {'lr': 0.0004893818343420302, 'samples': 931656, 'steps': 6469, 'loss/train': 2.777130126953125}
+01/23/2022 08:22:27 - INFO - codeparrot_training - Step 38820: {'lr': 0.0004893818343420302, 'samples': 931680, 'steps': 6469, 'loss/train': 1.7895735502243042}
+01/23/2022 08:22:27 - INFO - codeparrot_training - Step 38821: {'lr': 0.000489377115836167, 'samples': 931704, 'steps': 6470, 'loss/train': 3.375641107559204}
+01/23/2022 08:22:28 - INFO - codeparrot_training - Step 38822: {'lr': 0.000489377115836167, 'samples': 931728, 'steps': 6470, 'loss/train': 5.307866096496582}
+01/23/2022 08:22:28 - INFO - codeparrot_training - Step 38823: {'lr': 0.000489377115836167, 'samples': 931752, 'steps': 6470, 'loss/train': 3.13853120803833}
+01/23/2022 08:22:28 - INFO - codeparrot_training - Step 38824: {'lr': 0.000489377115836167, 'samples': 931776, 'steps': 6470, 'loss/train': 2.083646297454834}
+01/23/2022 08:22:29 - INFO - codeparrot_training - Step 38825: {'lr': 0.000489377115836167, 'samples': 931800, 'steps': 6470, 'loss/train': 3.663243532180786}
+01/23/2022 08:22:29 - INFO - codeparrot_training - Step 38826: {'lr': 0.000489377115836167, 'samples': 931824, 'steps': 6470, 'loss/train': 2.9247446060180664}
+01/23/2022 08:22:30 - INFO - codeparrot_training - Step 38827: {'lr': 0.0004893723963048882, 'samples': 931848, 'steps': 6471, 'loss/train': 3.8922746181488037}
+01/23/2022 08:22:30 - INFO - codeparrot_training - Step 38828: {'lr': 0.0004893723963048882, 'samples': 931872, 'steps': 6471, 'loss/train': 3.1335368156433105}
+01/23/2022 08:22:30 - INFO - codeparrot_training - Step 38829: {'lr': 0.0004893723963048882, 'samples': 931896, 'steps': 6471, 'loss/train': 3.2431652545928955}
+01/23/2022 08:22:31 - INFO - codeparrot_training - Step 38830: {'lr': 0.0004893723963048882, 'samples': 931920, 'steps': 6471, 'loss/train': 2.272441864013672}
+01/23/2022 08:22:31 - INFO - codeparrot_training - Step 38831: {'lr': 0.0004893723963048882, 'samples': 931944, 'steps': 6471, 'loss/train': 2.703583240509033}
+01/23/2022 08:22:31 - INFO - codeparrot_training - Step 38832: {'lr': 0.0004893723963048882, 'samples': 931968, 'steps': 6471, 'loss/train': 3.4553070068359375}
+01/23/2022 08:22:32 - INFO - codeparrot_training - Step 38833: {'lr': 0.0004893676757482142, 'samples': 931992, 'steps': 6472, 'loss/train': 1.6462393999099731}
+01/23/2022 08:22:32 - INFO - codeparrot_training - Step 38834: {'lr': 0.0004893676757482142, 'samples': 932016, 'steps': 6472, 'loss/train': 4.086414813995361}
+01/23/2022 08:22:33 - INFO - codeparrot_training - Step 38835: {'lr': 0.0004893676757482142, 'samples': 932040, 'steps': 6472, 'loss/train': 3.1305220127105713}
+01/23/2022 08:22:33 - INFO - codeparrot_training - Step 38836: {'lr': 0.0004893676757482142, 'samples': 932064, 'steps': 6472, 'loss/train': 3.763032913208008}
+01/23/2022 08:22:34 - INFO - codeparrot_training - Step 38837: {'lr': 0.0004893676757482142, 'samples': 932088, 'steps': 6472, 'loss/train': 3.6223177909851074}
+01/23/2022 08:22:34 - INFO - codeparrot_training - Step 38838: {'lr': 0.0004893676757482142, 'samples': 932112, 'steps': 6472, 'loss/train': 1.127502202987671}
+01/23/2022 08:22:35 - INFO - codeparrot_training - Step 38839: {'lr': 0.0004893629541661649, 'samples': 932136, 'steps': 6473, 'loss/train': 2.179426670074463}
+01/23/2022 08:22:35 - INFO - codeparrot_training - Step 38840: {'lr': 0.0004893629541661649, 'samples': 932160, 'steps': 6473, 'loss/train': 3.001182794570923}
+01/23/2022 08:22:36 - INFO - codeparrot_training - Step 38841: {'lr': 0.0004893629541661649, 'samples': 932184, 'steps': 6473, 'loss/train': 2.924971342086792}
+01/23/2022 08:22:36 - INFO - codeparrot_training - Step 38842: {'lr': 0.0004893629541661649, 'samples': 932208, 'steps': 6473, 'loss/train': 2.416865110397339}
+01/23/2022 08:22:36 - INFO - codeparrot_training - Step 38843: {'lr': 0.0004893629541661649, 'samples': 932232, 'steps': 6473, 'loss/train': 4.2117390632629395}
+01/23/2022 08:22:37 - INFO - codeparrot_training - Step 38844: {'lr': 0.0004893629541661649, 'samples': 932256, 'steps': 6473, 'loss/train': 2.3683254718780518}
+01/23/2022 08:22:37 - INFO - codeparrot_training - Step 38845: {'lr': 0.0004893582315587608, 'samples': 932280, 'steps': 6474, 'loss/train': 3.650562047958374}
+01/23/2022 08:22:38 - INFO - codeparrot_training - Step 38846: {'lr': 0.0004893582315587608, 'samples': 932304, 'steps': 6474, 'loss/train': 3.7589681148529053}
+01/23/2022 08:22:38 - INFO - codeparrot_training - Step 38847: {'lr': 0.0004893582315587608, 'samples': 932328, 'steps': 6474, 'loss/train': 2.647524356842041}
+01/23/2022 08:22:38 - INFO - codeparrot_training - Step 38848: {'lr': 0.0004893582315587608, 'samples': 932352, 'steps': 6474, 'loss/train': 3.419271469116211}
+01/23/2022 08:22:39 - INFO - codeparrot_training - Step 38849: {'lr': 0.0004893582315587608, 'samples': 932376, 'steps': 6474, 'loss/train': 2.2958428859710693}
+01/23/2022 08:22:39 - INFO - codeparrot_training - Step 38850: {'lr': 0.0004893582315587608, 'samples': 932400, 'steps': 6474, 'loss/train': 2.9909136295318604}
+01/23/2022 08:22:40 - INFO - codeparrot_training - Step 38851: {'lr': 0.0004893535079260221, 'samples': 932424, 'steps': 6475, 'loss/train': 4.061096668243408}
+01/23/2022 08:22:40 - INFO - codeparrot_training - Step 38852: {'lr': 0.0004893535079260221, 'samples': 932448, 'steps': 6475, 'loss/train': 2.8894896507263184}
+01/23/2022 08:22:40 - INFO - codeparrot_training - Step 38853: {'lr': 0.0004893535079260221, 'samples': 932472, 'steps': 6475, 'loss/train': 1.7812163829803467}
+01/23/2022 08:22:41 - INFO - codeparrot_training - Step 38854: {'lr': 0.0004893535079260221, 'samples': 932496, 'steps': 6475, 'loss/train': 3.0531392097473145}
+01/23/2022 08:22:41 - INFO - codeparrot_training - Step 38855: {'lr': 0.0004893535079260221, 'samples': 932520, 'steps': 6475, 'loss/train': 3.2945845127105713}
+01/23/2022 08:22:41 - INFO - codeparrot_training - Step 38856: {'lr': 0.0004893535079260221, 'samples': 932544, 'steps': 6475, 'loss/train': 5.337297439575195}
+01/23/2022 08:22:42 - INFO - codeparrot_training - Step 38857: {'lr': 0.000489348783267969, 'samples': 932568, 'steps': 6476, 'loss/train': 2.423247814178467}
+01/23/2022 08:22:42 - INFO - codeparrot_training - Step 38858: {'lr': 0.000489348783267969, 'samples': 932592, 'steps': 6476, 'loss/train': 3.282569646835327}
+01/23/2022 08:22:43 - INFO - codeparrot_training - Step 38859: {'lr': 0.000489348783267969, 'samples': 932616, 'steps': 6476, 'loss/train': 3.4176025390625}
+01/23/2022 08:22:43 - INFO - codeparrot_training - Step 38860: {'lr': 0.000489348783267969, 'samples': 932640, 'steps': 6476, 'loss/train': 3.054063081741333}
+01/23/2022 08:22:43 - INFO - codeparrot_training - Step 38861: {'lr': 0.000489348783267969, 'samples': 932664, 'steps': 6476, 'loss/train': 3.7912113666534424}
+01/23/2022 08:22:44 - INFO - codeparrot_training - Step 38862: {'lr': 0.000489348783267969, 'samples': 932688, 'steps': 6476, 'loss/train': 2.9695827960968018}
+01/23/2022 08:22:44 - INFO - codeparrot_training - Step 38863: {'lr': 0.0004893440575846215, 'samples': 932712, 'steps': 6477, 'loss/train': 3.0152111053466797}
+01/23/2022 08:22:45 - INFO - codeparrot_training - Step 38864: {'lr': 0.0004893440575846215, 'samples': 932736, 'steps': 6477, 'loss/train': 2.7534942626953125}
+01/23/2022 08:22:45 - INFO - codeparrot_training - Step 38865: {'lr': 0.0004893440575846215, 'samples': 932760, 'steps': 6477, 'loss/train': 2.177222967147827}
+01/23/2022 08:22:45 - INFO - codeparrot_training - Step 38866: {'lr': 0.0004893440575846215, 'samples': 932784, 'steps': 6477, 'loss/train': 2.85927414894104}
+01/23/2022 08:22:46 - INFO - codeparrot_training - Step 38867: {'lr': 0.0004893440575846215, 'samples': 932808, 'steps': 6477, 'loss/train': 3.112901210784912}
+01/23/2022 08:22:46 - INFO - codeparrot_training - Step 38868: {'lr': 0.0004893440575846215, 'samples': 932832, 'steps': 6477, 'loss/train': 2.743670701980591}
+01/23/2022 08:22:46 - INFO - codeparrot_training - Step 38869: {'lr': 0.0004893393308760002, 'samples': 932856, 'steps': 6478, 'loss/train': 3.295210599899292}
+01/23/2022 08:22:47 - INFO - codeparrot_training - Step 38870: {'lr': 0.0004893393308760002, 'samples': 932880, 'steps': 6478, 'loss/train': 3.1553430557250977}
+01/23/2022 08:22:47 - INFO - codeparrot_training - Step 38871: {'lr': 0.0004893393308760002, 'samples': 932904, 'steps': 6478, 'loss/train': 2.5944173336029053}
+01/23/2022 08:22:48 - INFO - codeparrot_training - Step 38872: {'lr': 0.0004893393308760002, 'samples': 932928, 'steps': 6478, 'loss/train': 3.978527069091797}
+01/23/2022 08:22:48 - INFO - codeparrot_training - Step 38873: {'lr': 0.0004893393308760002, 'samples': 932952, 'steps': 6478, 'loss/train': 3.4021072387695312}
+01/23/2022 08:22:48 - INFO - codeparrot_training - Step 38874: {'lr': 0.0004893393308760002, 'samples': 932976, 'steps': 6478, 'loss/train': 3.1414685249328613}
+01/23/2022 08:22:49 - INFO - codeparrot_training - Step 38875: {'lr': 0.0004893346031421253, 'samples': 933000, 'steps': 6479, 'loss/train': 1.8383803367614746}
+01/23/2022 08:22:50 - INFO - codeparrot_training - Step 38876: {'lr': 0.0004893346031421253, 'samples': 933024, 'steps': 6479, 'loss/train': 3.5609664916992188}
+01/23/2022 08:22:50 - INFO - codeparrot_training - Step 38877: {'lr': 0.0004893346031421253, 'samples': 933048, 'steps': 6479, 'loss/train': 3.075514793395996}
+01/23/2022 08:22:51 - INFO - codeparrot_training - Step 38878: {'lr': 0.0004893346031421253, 'samples': 933072, 'steps': 6479, 'loss/train': 2.3913803100585938}
+01/23/2022 08:22:51 - INFO - codeparrot_training - Step 38879: {'lr': 0.0004893346031421253, 'samples': 933096, 'steps': 6479, 'loss/train': 2.253059148788452}
+01/23/2022 08:22:51 - INFO - codeparrot_training - Step 38880: {'lr': 0.0004893346031421253, 'samples': 933120, 'steps': 6479, 'loss/train': 2.326528787612915}
+01/23/2022 08:22:52 - INFO - codeparrot_training - Step 38881: {'lr': 0.0004893298743830168, 'samples': 933144, 'steps': 6480, 'loss/train': 5.039220333099365}
+01/23/2022 08:22:52 - INFO - codeparrot_training - Step 38882: {'lr': 0.0004893298743830168, 'samples': 933168, 'steps': 6480, 'loss/train': 2.903369426727295}
+01/23/2022 08:22:52 - INFO - codeparrot_training - Step 38883: {'lr': 0.0004893298743830168, 'samples': 933192, 'steps': 6480, 'loss/train': 3.257283926010132}
+01/23/2022 08:22:53 - INFO - codeparrot_training - Step 38884: {'lr': 0.0004893298743830168, 'samples': 933216, 'steps': 6480, 'loss/train': 3.626098394393921}
+01/23/2022 08:22:53 - INFO - codeparrot_training - Step 38885: {'lr': 0.0004893298743830168, 'samples': 933240, 'steps': 6480, 'loss/train': 2.536405086517334}
+01/23/2022 08:22:54 - INFO - codeparrot_training - Step 38886: {'lr': 0.0004893298743830168, 'samples': 933264, 'steps': 6480, 'loss/train': 4.0158233642578125}
+01/23/2022 08:22:54 - INFO - codeparrot_training - Step 38887: {'lr': 0.0004893251445986952, 'samples': 933288, 'steps': 6481, 'loss/train': 2.9482288360595703}
+01/23/2022 08:22:54 - INFO - codeparrot_training - Step 38888: {'lr': 0.0004893251445986952, 'samples': 933312, 'steps': 6481, 'loss/train': 3.3658382892608643}
+01/23/2022 08:22:55 - INFO - codeparrot_training - Step 38889: {'lr': 0.0004893251445986952, 'samples': 933336, 'steps': 6481, 'loss/train': 3.5966427326202393}
+01/23/2022 08:22:55 - INFO - codeparrot_training - Step 38890: {'lr': 0.0004893251445986952, 'samples': 933360, 'steps': 6481, 'loss/train': 3.1047747135162354}
+01/23/2022 08:22:56 - INFO - codeparrot_training - Step 38891: {'lr': 0.0004893251445986952, 'samples': 933384, 'steps': 6481, 'loss/train': 3.4683868885040283}
+01/23/2022 08:22:56 - INFO - codeparrot_training - Step 38892: {'lr': 0.0004893251445986952, 'samples': 933408, 'steps': 6481, 'loss/train': 2.975180149078369}
+01/23/2022 08:22:56 - INFO - codeparrot_training - Step 38893: {'lr': 0.0004893204137891807, 'samples': 933432, 'steps': 6482, 'loss/train': 4.287125110626221}
+01/23/2022 08:22:57 - INFO - codeparrot_training - Step 38894: {'lr': 0.0004893204137891807, 'samples': 933456, 'steps': 6482, 'loss/train': 3.59468674659729}
+01/23/2022 08:22:57 - INFO - codeparrot_training - Step 38895: {'lr': 0.0004893204137891807, 'samples': 933480, 'steps': 6482, 'loss/train': 3.299471616744995}
+01/23/2022 08:22:57 - INFO - codeparrot_training - Step 38896: {'lr': 0.0004893204137891807, 'samples': 933504, 'steps': 6482, 'loss/train': 3.9173519611358643}
+01/23/2022 08:22:58 - INFO - codeparrot_training - Step 38897: {'lr': 0.0004893204137891807, 'samples': 933528, 'steps': 6482, 'loss/train': 3.3781731128692627}
+01/23/2022 08:22:58 - INFO - codeparrot_training - Step 38898: {'lr': 0.0004893204137891807, 'samples': 933552, 'steps': 6482, 'loss/train': 3.322413921356201}
+01/23/2022 08:22:59 - INFO - codeparrot_training - Step 38899: {'lr': 0.0004893156819544935, 'samples': 933576, 'steps': 6483, 'loss/train': 2.5860517024993896}
+01/23/2022 08:22:59 - INFO - codeparrot_training - Step 38900: {'lr': 0.0004893156819544935, 'samples': 933600, 'steps': 6483, 'loss/train': 2.382782220840454}
+01/23/2022 08:22:59 - INFO - codeparrot_training - Step 38901: {'lr': 0.0004893156819544935, 'samples': 933624, 'steps': 6483, 'loss/train': 2.910452127456665}
+01/23/2022 08:23:00 - INFO - codeparrot_training - Step 38902: {'lr': 0.0004893156819544935, 'samples': 933648, 'steps': 6483, 'loss/train': 3.4444100856781006}
+01/23/2022 08:23:00 - INFO - codeparrot_training - Step 38903: {'lr': 0.0004893156819544935, 'samples': 933672, 'steps': 6483, 'loss/train': 3.8086886405944824}
+01/23/2022 08:23:00 - INFO - codeparrot_training - Step 38904: {'lr': 0.0004893156819544935, 'samples': 933696, 'steps': 6483, 'loss/train': 3.3545432090759277}
+01/23/2022 08:23:01 - INFO - codeparrot_training - Step 38905: {'lr': 0.0004893109490946539, 'samples': 933720, 'steps': 6484, 'loss/train': 2.895346164703369}
+01/23/2022 08:23:01 - INFO - codeparrot_training - Step 38906: {'lr': 0.0004893109490946539, 'samples': 933744, 'steps': 6484, 'loss/train': 3.831143856048584}
+01/23/2022 08:23:02 - INFO - codeparrot_training - Step 38907: {'lr': 0.0004893109490946539, 'samples': 933768, 'steps': 6484, 'loss/train': 3.6038951873779297}
+01/23/2022 08:23:02 - INFO - codeparrot_training - Step 38908: {'lr': 0.0004893109490946539, 'samples': 933792, 'steps': 6484, 'loss/train': 3.126555919647217}
+01/23/2022 08:23:02 - INFO - codeparrot_training - Step 38909: {'lr': 0.0004893109490946539, 'samples': 933816, 'steps': 6484, 'loss/train': 3.4035418033599854}
+01/23/2022 08:23:03 - INFO - codeparrot_training - Step 38910: {'lr': 0.0004893109490946539, 'samples': 933840, 'steps': 6484, 'loss/train': 3.973771810531616}
+01/23/2022 08:23:03 - INFO - codeparrot_training - Step 38911: {'lr': 0.0004893062152096821, 'samples': 933864, 'steps': 6485, 'loss/train': 3.2474286556243896}
+01/23/2022 08:23:04 - INFO - codeparrot_training - Step 38912: {'lr': 0.0004893062152096821, 'samples': 933888, 'steps': 6485, 'loss/train': 4.440723896026611}
+01/23/2022 08:23:04 - INFO - codeparrot_training - Step 38913: {'lr': 0.0004893062152096821, 'samples': 933912, 'steps': 6485, 'loss/train': 4.816393852233887}
+01/23/2022 08:23:04 - INFO - codeparrot_training - Step 38914: {'lr': 0.0004893062152096821, 'samples': 933936, 'steps': 6485, 'loss/train': 5.640552520751953}
+01/23/2022 08:23:05 - INFO - codeparrot_training - Step 38915: {'lr': 0.0004893062152096821, 'samples': 933960, 'steps': 6485, 'loss/train': 3.6443979740142822}
+01/23/2022 08:23:06 - INFO - codeparrot_training - Step 38916: {'lr': 0.0004893062152096821, 'samples': 933984, 'steps': 6485, 'loss/train': 2.8850908279418945}
+01/23/2022 08:23:06 - INFO - codeparrot_training - Step 38917: {'lr': 0.0004893014802995985, 'samples': 934008, 'steps': 6486, 'loss/train': 3.6309163570404053}
+01/23/2022 08:23:07 - INFO - codeparrot_training - Step 38918: {'lr': 0.0004893014802995985, 'samples': 934032, 'steps': 6486, 'loss/train': 3.9266741275787354}
+01/23/2022 08:23:07 - INFO - codeparrot_training - Step 38919: {'lr': 0.0004893014802995985, 'samples': 934056, 'steps': 6486, 'loss/train': 0.70063316822052}
+01/23/2022 08:23:07 - INFO - codeparrot_training - Step 38920: {'lr': 0.0004893014802995985, 'samples': 934080, 'steps': 6486, 'loss/train': 3.348419189453125}
+01/23/2022 08:23:08 - INFO - codeparrot_training - Step 38921: {'lr': 0.0004893014802995985, 'samples': 934104, 'steps': 6486, 'loss/train': 2.9127750396728516}
+01/23/2022 08:23:08 - INFO - codeparrot_training - Step 38922: {'lr': 0.0004893014802995985, 'samples': 934128, 'steps': 6486, 'loss/train': 3.4886701107025146}
+01/23/2022 08:23:08 - INFO - codeparrot_training - Step 38923: {'lr': 0.0004892967443644235, 'samples': 934152, 'steps': 6487, 'loss/train': 2.508640766143799}
+01/23/2022 08:23:09 - INFO - codeparrot_training - Step 38924: {'lr': 0.0004892967443644235, 'samples': 934176, 'steps': 6487, 'loss/train': 2.784447193145752}
+01/23/2022 08:23:09 - INFO - codeparrot_training - Step 38925: {'lr': 0.0004892967443644235, 'samples': 934200, 'steps': 6487, 'loss/train': 3.252133846282959}
+01/23/2022 08:23:10 - INFO - codeparrot_training - Step 38926: {'lr': 0.0004892967443644235, 'samples': 934224, 'steps': 6487, 'loss/train': 2.9814655780792236}
+01/23/2022 08:23:10 - INFO - codeparrot_training - Step 38927: {'lr': 0.0004892967443644235, 'samples': 934248, 'steps': 6487, 'loss/train': 4.041112422943115}
+01/23/2022 08:23:10 - INFO - codeparrot_training - Step 38928: {'lr': 0.0004892967443644235, 'samples': 934272, 'steps': 6487, 'loss/train': 3.285799980163574}
+01/23/2022 08:23:11 - INFO - codeparrot_training - Step 38929: {'lr': 0.0004892920074041771, 'samples': 934296, 'steps': 6488, 'loss/train': 4.002261638641357}
+01/23/2022 08:23:11 - INFO - codeparrot_training - Step 38930: {'lr': 0.0004892920074041771, 'samples': 934320, 'steps': 6488, 'loss/train': 0.5520836710929871}
+01/23/2022 08:23:12 - INFO - codeparrot_training - Step 38931: {'lr': 0.0004892920074041771, 'samples': 934344, 'steps': 6488, 'loss/train': 3.6740050315856934}
+01/23/2022 08:23:12 - INFO - codeparrot_training - Step 38932: {'lr': 0.0004892920074041771, 'samples': 934368, 'steps': 6488, 'loss/train': 2.7591140270233154}
+01/23/2022 08:23:12 - INFO - codeparrot_training - Step 38933: {'lr': 0.0004892920074041771, 'samples': 934392, 'steps': 6488, 'loss/train': 3.248534917831421}
+01/23/2022 08:23:13 - INFO - codeparrot_training - Step 38934: {'lr': 0.0004892920074041771, 'samples': 934416, 'steps': 6488, 'loss/train': 2.344391107559204}
+01/23/2022 08:23:13 - INFO - codeparrot_training - Step 38935: {'lr': 0.0004892872694188797, 'samples': 934440, 'steps': 6489, 'loss/train': 2.9902102947235107}
+01/23/2022 08:23:13 - INFO - codeparrot_training - Step 38936: {'lr': 0.0004892872694188797, 'samples': 934464, 'steps': 6489, 'loss/train': 3.9867334365844727}
+01/23/2022 08:23:14 - INFO - codeparrot_training - Step 38937: {'lr': 0.0004892872694188797, 'samples': 934488, 'steps': 6489, 'loss/train': 2.689467430114746}
+01/23/2022 08:23:14 - INFO - codeparrot_training - Step 38938: {'lr': 0.0004892872694188797, 'samples': 934512, 'steps': 6489, 'loss/train': 2.9017086029052734}
+01/23/2022 08:23:15 - INFO - codeparrot_training - Step 38939: {'lr': 0.0004892872694188797, 'samples': 934536, 'steps': 6489, 'loss/train': 2.316284418106079}
+01/23/2022 08:23:15 - INFO - codeparrot_training - Step 38940: {'lr': 0.0004892872694188797, 'samples': 934560, 'steps': 6489, 'loss/train': 1.8944032192230225}
+01/23/2022 08:23:15 - INFO - codeparrot_training - Step 38941: {'lr': 0.0004892825304085517, 'samples': 934584, 'steps': 6490, 'loss/train': 2.9005796909332275}
+01/23/2022 08:23:16 - INFO - codeparrot_training - Step 38942: {'lr': 0.0004892825304085517, 'samples': 934608, 'steps': 6490, 'loss/train': 4.5663557052612305}
+01/23/2022 08:23:16 - INFO - codeparrot_training - Step 38943: {'lr': 0.0004892825304085517, 'samples': 934632, 'steps': 6490, 'loss/train': 3.713061809539795}
+01/23/2022 08:23:16 - INFO - codeparrot_training - Step 38944: {'lr': 0.0004892825304085517, 'samples': 934656, 'steps': 6490, 'loss/train': 1.8487786054611206}
+01/23/2022 08:23:17 - INFO - codeparrot_training - Step 38945: {'lr': 0.0004892825304085517, 'samples': 934680, 'steps': 6490, 'loss/train': 3.1148533821105957}
+01/23/2022 08:23:17 - INFO - codeparrot_training - Step 38946: {'lr': 0.0004892825304085517, 'samples': 934704, 'steps': 6490, 'loss/train': 2.737846851348877}
+01/23/2022 08:23:18 - INFO - codeparrot_training - Step 38947: {'lr': 0.0004892777903732133, 'samples': 934728, 'steps': 6491, 'loss/train': 3.592200994491577}
+01/23/2022 08:23:18 - INFO - codeparrot_training - Step 38948: {'lr': 0.0004892777903732133, 'samples': 934752, 'steps': 6491, 'loss/train': 2.6748924255371094}
+01/23/2022 08:23:18 - INFO - codeparrot_training - Step 38949: {'lr': 0.0004892777903732133, 'samples': 934776, 'steps': 6491, 'loss/train': 3.316619873046875}
+01/23/2022 08:23:19 - INFO - codeparrot_training - Step 38950: {'lr': 0.0004892777903732133, 'samples': 934800, 'steps': 6491, 'loss/train': 3.0748071670532227}
+01/23/2022 08:23:19 - INFO - codeparrot_training - Step 38951: {'lr': 0.0004892777903732133, 'samples': 934824, 'steps': 6491, 'loss/train': 3.1785330772399902}
+01/23/2022 08:23:20 - INFO - codeparrot_training - Step 38952: {'lr': 0.0004892777903732133, 'samples': 934848, 'steps': 6491, 'loss/train': 0.7273881435394287}
+01/23/2022 08:23:20 - INFO - codeparrot_training - Step 38953: {'lr': 0.0004892730493128848, 'samples': 934872, 'steps': 6492, 'loss/train': 4.084521770477295}
+01/23/2022 08:23:20 - INFO - codeparrot_training - Step 38954: {'lr': 0.0004892730493128848, 'samples': 934896, 'steps': 6492, 'loss/train': 3.154726982116699}
+01/23/2022 08:23:21 - INFO - codeparrot_training - Step 38955: {'lr': 0.0004892730493128848, 'samples': 934920, 'steps': 6492, 'loss/train': 2.3778674602508545}
+01/23/2022 08:23:22 - INFO - codeparrot_training - Step 38956: {'lr': 0.0004892730493128848, 'samples': 934944, 'steps': 6492, 'loss/train': 3.1778645515441895}
+01/23/2022 08:23:22 - INFO - codeparrot_training - Step 38957: {'lr': 0.0004892730493128848, 'samples': 934968, 'steps': 6492, 'loss/train': 3.027038812637329}
+01/23/2022 08:23:23 - INFO - codeparrot_training - Step 38958: {'lr': 0.0004892730493128848, 'samples': 934992, 'steps': 6492, 'loss/train': 2.468350410461426}
+01/23/2022 08:23:23 - INFO - codeparrot_training - Step 38959: {'lr': 0.0004892683072275865, 'samples': 935016, 'steps': 6493, 'loss/train': 2.9695277214050293}
+01/23/2022 08:23:23 - INFO - codeparrot_training - Step 38960: {'lr': 0.0004892683072275865, 'samples': 935040, 'steps': 6493, 'loss/train': 3.139239549636841}
+01/23/2022 08:23:24 - INFO - codeparrot_training - Step 38961: {'lr': 0.0004892683072275865, 'samples': 935064, 'steps': 6493, 'loss/train': 2.6443142890930176}
+01/23/2022 08:23:24 - INFO - codeparrot_training - Step 38962: {'lr': 0.0004892683072275865, 'samples': 935088, 'steps': 6493, 'loss/train': 3.998425245285034}
+01/23/2022 08:23:25 - INFO - codeparrot_training - Step 38963: {'lr': 0.0004892683072275865, 'samples': 935112, 'steps': 6493, 'loss/train': 3.2447407245635986}
+01/23/2022 08:23:25 - INFO - codeparrot_training - Step 38964: {'lr': 0.0004892683072275865, 'samples': 935136, 'steps': 6493, 'loss/train': 3.300205945968628}
+01/23/2022 08:23:25 - INFO - codeparrot_training - Step 38965: {'lr': 0.0004892635641173389, 'samples': 935160, 'steps': 6494, 'loss/train': 2.6813838481903076}
+01/23/2022 08:23:26 - INFO - codeparrot_training - Step 38966: {'lr': 0.0004892635641173389, 'samples': 935184, 'steps': 6494, 'loss/train': 3.7363176345825195}
+01/23/2022 08:23:26 - INFO - codeparrot_training - Step 38967: {'lr': 0.0004892635641173389, 'samples': 935208, 'steps': 6494, 'loss/train': 3.9169981479644775}
+01/23/2022 08:23:26 - INFO - codeparrot_training - Step 38968: {'lr': 0.0004892635641173389, 'samples': 935232, 'steps': 6494, 'loss/train': 3.306304931640625}
+01/23/2022 08:23:27 - INFO - codeparrot_training - Step 38969: {'lr': 0.0004892635641173389, 'samples': 935256, 'steps': 6494, 'loss/train': 2.569910764694214}
+01/23/2022 08:23:27 - INFO - codeparrot_training - Step 38970: {'lr': 0.0004892635641173389, 'samples': 935280, 'steps': 6494, 'loss/train': 2.595763683319092}
+01/23/2022 08:23:28 - INFO - codeparrot_training - Step 38971: {'lr': 0.0004892588199821619, 'samples': 935304, 'steps': 6495, 'loss/train': 0.46242615580558777}
+01/23/2022 08:23:28 - INFO - codeparrot_training - Step 38972: {'lr': 0.0004892588199821619, 'samples': 935328, 'steps': 6495, 'loss/train': 2.2126288414001465}
+01/23/2022 08:23:28 - INFO - codeparrot_training - Step 38973: {'lr': 0.0004892588199821619, 'samples': 935352, 'steps': 6495, 'loss/train': 3.0755927562713623}
+01/23/2022 08:23:29 - INFO - codeparrot_training - Step 38974: {'lr': 0.0004892588199821619, 'samples': 935376, 'steps': 6495, 'loss/train': 3.9965803623199463}
+01/23/2022 08:23:29 - INFO - codeparrot_training - Step 38975: {'lr': 0.0004892588199821619, 'samples': 935400, 'steps': 6495, 'loss/train': 3.0220978260040283}
+01/23/2022 08:23:30 - INFO - codeparrot_training - Step 38976: {'lr': 0.0004892588199821619, 'samples': 935424, 'steps': 6495, 'loss/train': 3.470982789993286}
+01/23/2022 08:23:30 - INFO - codeparrot_training - Step 38977: {'lr': 0.0004892540748220763, 'samples': 935448, 'steps': 6496, 'loss/train': 3.736644983291626}
+01/23/2022 08:23:30 - INFO - codeparrot_training - Step 38978: {'lr': 0.0004892540748220763, 'samples': 935472, 'steps': 6496, 'loss/train': 4.492185115814209}
+01/23/2022 08:23:31 - INFO - codeparrot_training - Step 38979: {'lr': 0.0004892540748220763, 'samples': 935496, 'steps': 6496, 'loss/train': 3.867025852203369}
+01/23/2022 08:23:31 - INFO - codeparrot_training - Step 38980: {'lr': 0.0004892540748220763, 'samples': 935520, 'steps': 6496, 'loss/train': 2.6435251235961914}
+01/23/2022 08:23:32 - INFO - codeparrot_training - Step 38981: {'lr': 0.0004892540748220763, 'samples': 935544, 'steps': 6496, 'loss/train': 2.9853460788726807}
+01/23/2022 08:23:32 - INFO - codeparrot_training - Step 38982: {'lr': 0.0004892540748220763, 'samples': 935568, 'steps': 6496, 'loss/train': 3.3330023288726807}
+01/23/2022 08:23:32 - INFO - codeparrot_training - Step 38983: {'lr': 0.0004892493286371022, 'samples': 935592, 'steps': 6497, 'loss/train': 3.604947566986084}
+01/23/2022 08:23:33 - INFO - codeparrot_training - Step 38984: {'lr': 0.0004892493286371022, 'samples': 935616, 'steps': 6497, 'loss/train': 3.325223684310913}
+01/23/2022 08:23:33 - INFO - codeparrot_training - Step 38985: {'lr': 0.0004892493286371022, 'samples': 935640, 'steps': 6497, 'loss/train': 3.888036012649536}
+01/23/2022 08:23:33 - INFO - codeparrot_training - Step 38986: {'lr': 0.0004892493286371022, 'samples': 935664, 'steps': 6497, 'loss/train': 2.7254035472869873}
+01/23/2022 08:23:34 - INFO - codeparrot_training - Step 38987: {'lr': 0.0004892493286371022, 'samples': 935688, 'steps': 6497, 'loss/train': 2.6150989532470703}
+01/23/2022 08:23:34 - INFO - codeparrot_training - Step 38988: {'lr': 0.0004892493286371022, 'samples': 935712, 'steps': 6497, 'loss/train': 2.956801176071167}
+01/23/2022 08:23:35 - INFO - codeparrot_training - Step 38989: {'lr': 0.00048924458142726, 'samples': 935736, 'steps': 6498, 'loss/train': 3.34716534614563}
+01/23/2022 08:23:35 - INFO - codeparrot_training - Step 38990: {'lr': 0.00048924458142726, 'samples': 935760, 'steps': 6498, 'loss/train': 3.314286470413208}
+01/23/2022 08:23:35 - INFO - codeparrot_training - Step 38991: {'lr': 0.00048924458142726, 'samples': 935784, 'steps': 6498, 'loss/train': 2.997450590133667}
+01/23/2022 08:23:36 - INFO - codeparrot_training - Step 38992: {'lr': 0.00048924458142726, 'samples': 935808, 'steps': 6498, 'loss/train': 2.32065749168396}
+01/23/2022 08:23:36 - INFO - codeparrot_training - Step 38993: {'lr': 0.00048924458142726, 'samples': 935832, 'steps': 6498, 'loss/train': 3.1127583980560303}
+01/23/2022 08:23:36 - INFO - codeparrot_training - Step 38994: {'lr': 0.00048924458142726, 'samples': 935856, 'steps': 6498, 'loss/train': 6.154850959777832}
+01/23/2022 08:23:38 - INFO - codeparrot_training - Step 38995: {'lr': 0.0004892398331925698, 'samples': 935880, 'steps': 6499, 'loss/train': 3.7040252685546875}
+01/23/2022 08:23:38 - INFO - codeparrot_training - Step 38996: {'lr': 0.0004892398331925698, 'samples': 935904, 'steps': 6499, 'loss/train': 2.8373143672943115}
+01/23/2022 08:23:38 - INFO - codeparrot_training - Step 38997: {'lr': 0.0004892398331925698, 'samples': 935928, 'steps': 6499, 'loss/train': 3.3800978660583496}
+01/23/2022 08:23:39 - INFO - codeparrot_training - Step 38998: {'lr': 0.0004892398331925698, 'samples': 935952, 'steps': 6499, 'loss/train': 4.658259391784668}
+01/23/2022 08:23:39 - INFO - codeparrot_training - Step 38999: {'lr': 0.0004892398331925698, 'samples': 935976, 'steps': 6499, 'loss/train': 3.707747220993042}
+01/23/2022 08:23:39 - INFO - codeparrot_training - Step 39000: {'lr': 0.0004892398331925698, 'samples': 936000, 'steps': 6499, 'loss/train': 3.601071834564209}
+01/23/2022 08:23:40 - INFO - codeparrot_training - Step 39001: {'lr': 0.0004892350839330522, 'samples': 936024, 'steps': 6500, 'loss/train': 2.9819533824920654}
+01/23/2022 08:23:40 - INFO - codeparrot_training - Step 39002: {'lr': 0.0004892350839330522, 'samples': 936048, 'steps': 6500, 'loss/train': 3.6728169918060303}
+01/23/2022 08:23:41 - INFO - codeparrot_training - Step 39003: {'lr': 0.0004892350839330522, 'samples': 936072, 'steps': 6500, 'loss/train': 4.011600494384766}
+01/23/2022 08:23:41 - INFO - codeparrot_training - Step 39004: {'lr': 0.0004892350839330522, 'samples': 936096, 'steps': 6500, 'loss/train': 2.1730215549468994}
+01/23/2022 08:23:41 - INFO - codeparrot_training - Step 39005: {'lr': 0.0004892350839330522, 'samples': 936120, 'steps': 6500, 'loss/train': 2.1609578132629395}
+01/23/2022 08:23:42 - INFO - codeparrot_training - Step 39006: {'lr': 0.0004892350839330522, 'samples': 936144, 'steps': 6500, 'loss/train': 3.0513007640838623}
+01/23/2022 08:23:42 - INFO - codeparrot_training - Step 39007: {'lr': 0.0004892303336487275, 'samples': 936168, 'steps': 6501, 'loss/train': 3.6708455085754395}
+01/23/2022 08:23:43 - INFO - codeparrot_training - Step 39008: {'lr': 0.0004892303336487275, 'samples': 936192, 'steps': 6501, 'loss/train': 1.713176965713501}
+01/23/2022 08:23:43 - INFO - codeparrot_training - Step 39009: {'lr': 0.0004892303336487275, 'samples': 936216, 'steps': 6501, 'loss/train': 4.724745750427246}
+01/23/2022 08:23:43 - INFO - codeparrot_training - Step 39010: {'lr': 0.0004892303336487275, 'samples': 936240, 'steps': 6501, 'loss/train': 3.2632827758789062}
+01/23/2022 08:23:44 - INFO - codeparrot_training - Step 39011: {'lr': 0.0004892303336487275, 'samples': 936264, 'steps': 6501, 'loss/train': 3.0275161266326904}
+01/23/2022 08:23:44 - INFO - codeparrot_training - Step 39012: {'lr': 0.0004892303336487275, 'samples': 936288, 'steps': 6501, 'loss/train': 3.3429436683654785}
+01/23/2022 08:23:44 - INFO - codeparrot_training - Step 39013: {'lr': 0.000489225582339616, 'samples': 936312, 'steps': 6502, 'loss/train': 3.3782660961151123}
+01/23/2022 08:23:45 - INFO - codeparrot_training - Step 39014: {'lr': 0.000489225582339616, 'samples': 936336, 'steps': 6502, 'loss/train': 3.9251415729522705}
+01/23/2022 08:23:45 - INFO - codeparrot_training - Step 39015: {'lr': 0.000489225582339616, 'samples': 936360, 'steps': 6502, 'loss/train': 2.9964852333068848}
+01/23/2022 08:23:46 - INFO - codeparrot_training - Step 39016: {'lr': 0.000489225582339616, 'samples': 936384, 'steps': 6502, 'loss/train': 3.7134881019592285}
+01/23/2022 08:23:46 - INFO - codeparrot_training - Step 39017: {'lr': 0.000489225582339616, 'samples': 936408, 'steps': 6502, 'loss/train': 3.6953885555267334}
+01/23/2022 08:23:46 - INFO - codeparrot_training - Step 39018: {'lr': 0.000489225582339616, 'samples': 936432, 'steps': 6502, 'loss/train': 3.07155442237854}
+01/23/2022 08:23:47 - INFO - codeparrot_training - Step 39019: {'lr': 0.000489220830005738, 'samples': 936456, 'steps': 6503, 'loss/train': 2.6243226528167725}
+01/23/2022 08:23:47 - INFO - codeparrot_training - Step 39020: {'lr': 0.000489220830005738, 'samples': 936480, 'steps': 6503, 'loss/train': 3.3387882709503174}
+01/23/2022 08:23:48 - INFO - codeparrot_training - Step 39021: {'lr': 0.000489220830005738, 'samples': 936504, 'steps': 6503, 'loss/train': 4.375588417053223}
+01/23/2022 08:23:48 - INFO - codeparrot_training - Step 39022: {'lr': 0.000489220830005738, 'samples': 936528, 'steps': 6503, 'loss/train': 4.010272026062012}
+01/23/2022 08:23:48 - INFO - codeparrot_training - Step 39023: {'lr': 0.000489220830005738, 'samples': 936552, 'steps': 6503, 'loss/train': 3.700294256210327}
+01/23/2022 08:23:49 - INFO - codeparrot_training - Step 39024: {'lr': 0.000489220830005738, 'samples': 936576, 'steps': 6503, 'loss/train': 2.4142520427703857}
+01/23/2022 08:23:49 - INFO - codeparrot_training - Step 39025: {'lr': 0.0004892160766471141, 'samples': 936600, 'steps': 6504, 'loss/train': 4.3196563720703125}
+01/23/2022 08:23:49 - INFO - codeparrot_training - Step 39026: {'lr': 0.0004892160766471141, 'samples': 936624, 'steps': 6504, 'loss/train': 2.169215679168701}
+01/23/2022 08:23:50 - INFO - codeparrot_training - Step 39027: {'lr': 0.0004892160766471141, 'samples': 936648, 'steps': 6504, 'loss/train': 3.2423768043518066}
+01/23/2022 08:23:50 - INFO - codeparrot_training - Step 39028: {'lr': 0.0004892160766471141, 'samples': 936672, 'steps': 6504, 'loss/train': 2.822842597961426}
+01/23/2022 08:23:51 - INFO - codeparrot_training - Step 39029: {'lr': 0.0004892160766471141, 'samples': 936696, 'steps': 6504, 'loss/train': 3.6219489574432373}
+01/23/2022 08:23:51 - INFO - codeparrot_training - Step 39030: {'lr': 0.0004892160766471141, 'samples': 936720, 'steps': 6504, 'loss/train': 3.5815837383270264}
+01/23/2022 08:23:51 - INFO - codeparrot_training - Step 39031: {'lr': 0.0004892113222637643, 'samples': 936744, 'steps': 6505, 'loss/train': 2.823336362838745}
+01/23/2022 08:23:52 - INFO - codeparrot_training - Step 39032: {'lr': 0.0004892113222637643, 'samples': 936768, 'steps': 6505, 'loss/train': 2.9066028594970703}
+01/23/2022 08:23:52 - INFO - codeparrot_training - Step 39033: {'lr': 0.0004892113222637643, 'samples': 936792, 'steps': 6505, 'loss/train': 2.5959084033966064}
+01/23/2022 08:23:55 - INFO - codeparrot_training - Step 39034: {'lr': 0.0004892113222637643, 'samples': 936816, 'steps': 6505, 'loss/train': 2.3259854316711426}
+01/23/2022 08:23:55 - INFO - codeparrot_training - Step 39035: {'lr': 0.0004892113222637643, 'samples': 936840, 'steps': 6505, 'loss/train': 3.937258720397949}
+01/23/2022 08:23:56 - INFO - codeparrot_training - Step 39036: {'lr': 0.0004892113222637643, 'samples': 936864, 'steps': 6505, 'loss/train': 2.650022506713867}
+01/23/2022 08:23:56 - INFO - codeparrot_training - Step 39037: {'lr': 0.0004892065668557093, 'samples': 936888, 'steps': 6506, 'loss/train': 3.084451198577881}
+01/23/2022 08:23:56 - INFO - codeparrot_training - Step 39038: {'lr': 0.0004892065668557093, 'samples': 936912, 'steps': 6506, 'loss/train': 2.478200674057007}
+01/23/2022 08:23:57 - INFO - codeparrot_training - Step 39039: {'lr': 0.0004892065668557093, 'samples': 936936, 'steps': 6506, 'loss/train': 3.650277614593506}
+01/23/2022 08:23:57 - INFO - codeparrot_training - Step 39040: {'lr': 0.0004892065668557093, 'samples': 936960, 'steps': 6506, 'loss/train': 3.0496108531951904}
+01/23/2022 08:23:58 - INFO - codeparrot_training - Step 39041: {'lr': 0.0004892065668557093, 'samples': 936984, 'steps': 6506, 'loss/train': 3.0641489028930664}
+01/23/2022 08:23:58 - INFO - codeparrot_training - Step 39042: {'lr': 0.0004892065668557093, 'samples': 937008, 'steps': 6506, 'loss/train': 3.7832164764404297}
+01/23/2022 08:23:58 - INFO - codeparrot_training - Step 39043: {'lr': 0.0004892018104229692, 'samples': 937032, 'steps': 6507, 'loss/train': 3.5957138538360596}
+01/23/2022 08:23:59 - INFO - codeparrot_training - Step 39044: {'lr': 0.0004892018104229692, 'samples': 937056, 'steps': 6507, 'loss/train': 3.4060049057006836}
+01/23/2022 08:23:59 - INFO - codeparrot_training - Step 39045: {'lr': 0.0004892018104229692, 'samples': 937080, 'steps': 6507, 'loss/train': 2.635688304901123}
+01/23/2022 08:24:00 - INFO - codeparrot_training - Step 39046: {'lr': 0.0004892018104229692, 'samples': 937104, 'steps': 6507, 'loss/train': 4.598985195159912}
+01/23/2022 08:24:00 - INFO - codeparrot_training - Step 39047: {'lr': 0.0004892018104229692, 'samples': 937128, 'steps': 6507, 'loss/train': 2.3277552127838135}
+01/23/2022 08:24:00 - INFO - codeparrot_training - Step 39048: {'lr': 0.0004892018104229692, 'samples': 937152, 'steps': 6507, 'loss/train': 3.5089752674102783}
+01/23/2022 08:24:01 - INFO - codeparrot_training - Step 39049: {'lr': 0.0004891970529655646, 'samples': 937176, 'steps': 6508, 'loss/train': 2.9135549068450928}
+01/23/2022 08:24:01 - INFO - codeparrot_training - Step 39050: {'lr': 0.0004891970529655646, 'samples': 937200, 'steps': 6508, 'loss/train': 2.865877389907837}
+01/23/2022 08:24:01 - INFO - codeparrot_training - Step 39051: {'lr': 0.0004891970529655646, 'samples': 937224, 'steps': 6508, 'loss/train': 3.399592876434326}
+01/23/2022 08:24:02 - INFO - codeparrot_training - Step 39052: {'lr': 0.0004891970529655646, 'samples': 937248, 'steps': 6508, 'loss/train': 2.834918975830078}
+01/23/2022 08:24:02 - INFO - codeparrot_training - Step 39053: {'lr': 0.0004891970529655646, 'samples': 937272, 'steps': 6508, 'loss/train': 3.769096612930298}
+01/23/2022 08:24:03 - INFO - codeparrot_training - Step 39054: {'lr': 0.0004891970529655646, 'samples': 937296, 'steps': 6508, 'loss/train': 4.056525707244873}
+01/23/2022 08:24:03 - INFO - codeparrot_training - Step 39055: {'lr': 0.0004891922944835158, 'samples': 937320, 'steps': 6509, 'loss/train': 4.156972885131836}
+01/23/2022 08:24:03 - INFO - codeparrot_training - Step 39056: {'lr': 0.0004891922944835158, 'samples': 937344, 'steps': 6509, 'loss/train': 3.4235475063323975}
+01/23/2022 08:24:04 - INFO - codeparrot_training - Step 39057: {'lr': 0.0004891922944835158, 'samples': 937368, 'steps': 6509, 'loss/train': 3.2262887954711914}
+01/23/2022 08:24:04 - INFO - codeparrot_training - Step 39058: {'lr': 0.0004891922944835158, 'samples': 937392, 'steps': 6509, 'loss/train': 3.0736501216888428}
+01/23/2022 08:24:05 - INFO - codeparrot_training - Step 39059: {'lr': 0.0004891922944835158, 'samples': 937416, 'steps': 6509, 'loss/train': 2.4471399784088135}
+01/23/2022 08:24:05 - INFO - codeparrot_training - Step 39060: {'lr': 0.0004891922944835158, 'samples': 937440, 'steps': 6509, 'loss/train': 1.5778363943099976}
+01/23/2022 08:24:05 - INFO - codeparrot_training - Step 39061: {'lr': 0.000489187534976843, 'samples': 937464, 'steps': 6510, 'loss/train': 2.259246826171875}
+01/23/2022 08:24:06 - INFO - codeparrot_training - Step 39062: {'lr': 0.000489187534976843, 'samples': 937488, 'steps': 6510, 'loss/train': 2.365201711654663}
+01/23/2022 08:24:06 - INFO - codeparrot_training - Step 39063: {'lr': 0.000489187534976843, 'samples': 937512, 'steps': 6510, 'loss/train': 3.3095786571502686}
+01/23/2022 08:24:06 - INFO - codeparrot_training - Step 39064: {'lr': 0.000489187534976843, 'samples': 937536, 'steps': 6510, 'loss/train': 2.499480724334717}
+01/23/2022 08:24:07 - INFO - codeparrot_training - Step 39065: {'lr': 0.000489187534976843, 'samples': 937560, 'steps': 6510, 'loss/train': 2.7357852458953857}
+01/23/2022 08:24:07 - INFO - codeparrot_training - Step 39066: {'lr': 0.000489187534976843, 'samples': 937584, 'steps': 6510, 'loss/train': 2.990896224975586}
+01/23/2022 08:24:08 - INFO - codeparrot_training - Step 39067: {'lr': 0.0004891827744455668, 'samples': 937608, 'steps': 6511, 'loss/train': 3.4626216888427734}
+01/23/2022 08:24:08 - INFO - codeparrot_training - Step 39068: {'lr': 0.0004891827744455668, 'samples': 937632, 'steps': 6511, 'loss/train': 2.431201457977295}
+01/23/2022 08:24:08 - INFO - codeparrot_training - Step 39069: {'lr': 0.0004891827744455668, 'samples': 937656, 'steps': 6511, 'loss/train': 1.8026885986328125}
+01/23/2022 08:24:09 - INFO - codeparrot_training - Step 39070: {'lr': 0.0004891827744455668, 'samples': 937680, 'steps': 6511, 'loss/train': 1.6267896890640259}
+01/23/2022 08:24:09 - INFO - codeparrot_training - Step 39071: {'lr': 0.0004891827744455668, 'samples': 937704, 'steps': 6511, 'loss/train': 2.145327091217041}
+01/23/2022 08:24:09 - INFO - codeparrot_training - Step 39072: {'lr': 0.0004891827744455668, 'samples': 937728, 'steps': 6511, 'loss/train': 3.6713900566101074}
+01/23/2022 08:24:10 - INFO - codeparrot_training - Step 39073: {'lr': 0.0004891780128897077, 'samples': 937752, 'steps': 6512, 'loss/train': 0.8099073171615601}
+01/23/2022 08:24:11 - INFO - codeparrot_training - Step 39074: {'lr': 0.0004891780128897077, 'samples': 937776, 'steps': 6512, 'loss/train': 3.4083569049835205}
+01/23/2022 08:24:11 - INFO - codeparrot_training - Step 39075: {'lr': 0.0004891780128897077, 'samples': 937800, 'steps': 6512, 'loss/train': 3.3191070556640625}
+01/23/2022 08:24:12 - INFO - codeparrot_training - Step 39076: {'lr': 0.0004891780128897077, 'samples': 937824, 'steps': 6512, 'loss/train': 3.316289186477661}
+01/23/2022 08:24:12 - INFO - codeparrot_training - Step 39077: {'lr': 0.0004891780128897077, 'samples': 937848, 'steps': 6512, 'loss/train': 3.8867931365966797}
+01/23/2022 08:24:13 - INFO - codeparrot_training - Step 39078: {'lr': 0.0004891780128897077, 'samples': 937872, 'steps': 6512, 'loss/train': 2.9686365127563477}
+01/23/2022 08:24:13 - INFO - codeparrot_training - Step 39079: {'lr': 0.0004891732503092858, 'samples': 937896, 'steps': 6513, 'loss/train': 2.600116729736328}
+01/23/2022 08:24:13 - INFO - codeparrot_training - Step 39080: {'lr': 0.0004891732503092858, 'samples': 937920, 'steps': 6513, 'loss/train': 2.704653263092041}
+01/23/2022 08:24:14 - INFO - codeparrot_training - Step 39081: {'lr': 0.0004891732503092858, 'samples': 937944, 'steps': 6513, 'loss/train': 2.579841136932373}
+01/23/2022 08:24:14 - INFO - codeparrot_training - Step 39082: {'lr': 0.0004891732503092858, 'samples': 937968, 'steps': 6513, 'loss/train': 3.283367395401001}
+01/23/2022 08:24:15 - INFO - codeparrot_training - Step 39083: {'lr': 0.0004891732503092858, 'samples': 937992, 'steps': 6513, 'loss/train': 2.9705052375793457}
+01/23/2022 08:24:15 - INFO - codeparrot_training - Step 39084: {'lr': 0.0004891732503092858, 'samples': 938016, 'steps': 6513, 'loss/train': 2.8900678157806396}
+01/23/2022 08:24:15 - INFO - codeparrot_training - Step 39085: {'lr': 0.0004891684867043216, 'samples': 938040, 'steps': 6514, 'loss/train': 3.2166287899017334}
+01/23/2022 08:24:16 - INFO - codeparrot_training - Step 39086: {'lr': 0.0004891684867043216, 'samples': 938064, 'steps': 6514, 'loss/train': 3.8004114627838135}
+01/23/2022 08:24:16 - INFO - codeparrot_training - Step 39087: {'lr': 0.0004891684867043216, 'samples': 938088, 'steps': 6514, 'loss/train': 3.828885078430176}
+01/23/2022 08:24:17 - INFO - codeparrot_training - Step 39088: {'lr': 0.0004891684867043216, 'samples': 938112, 'steps': 6514, 'loss/train': 3.5047051906585693}
+01/23/2022 08:24:17 - INFO - codeparrot_training - Step 39089: {'lr': 0.0004891684867043216, 'samples': 938136, 'steps': 6514, 'loss/train': 2.1761090755462646}
+01/23/2022 08:24:17 - INFO - codeparrot_training - Step 39090: {'lr': 0.0004891684867043216, 'samples': 938160, 'steps': 6514, 'loss/train': 3.080559492111206}
+01/23/2022 08:24:18 - INFO - codeparrot_training - Step 39091: {'lr': 0.0004891637220748356, 'samples': 938184, 'steps': 6515, 'loss/train': 3.475106716156006}
+01/23/2022 08:24:18 - INFO - codeparrot_training - Step 39092: {'lr': 0.0004891637220748356, 'samples': 938208, 'steps': 6515, 'loss/train': 3.361220598220825}
+01/23/2022 08:24:18 - INFO - codeparrot_training - Step 39093: {'lr': 0.0004891637220748356, 'samples': 938232, 'steps': 6515, 'loss/train': 2.07442307472229}
+01/23/2022 08:24:19 - INFO - codeparrot_training - Step 39094: {'lr': 0.0004891637220748356, 'samples': 938256, 'steps': 6515, 'loss/train': 3.1725361347198486}
+01/23/2022 08:24:19 - INFO - codeparrot_training - Step 39095: {'lr': 0.0004891637220748356, 'samples': 938280, 'steps': 6515, 'loss/train': 3.9187402725219727}
+01/23/2022 08:24:20 - INFO - codeparrot_training - Step 39096: {'lr': 0.0004891637220748356, 'samples': 938304, 'steps': 6515, 'loss/train': 3.225205898284912}
+01/23/2022 08:24:20 - INFO - codeparrot_training - Step 39097: {'lr': 0.0004891589564208482, 'samples': 938328, 'steps': 6516, 'loss/train': 4.052952766418457}
+01/23/2022 08:24:20 - INFO - codeparrot_training - Step 39098: {'lr': 0.0004891589564208482, 'samples': 938352, 'steps': 6516, 'loss/train': 2.945675849914551}
+01/23/2022 08:24:21 - INFO - codeparrot_training - Step 39099: {'lr': 0.0004891589564208482, 'samples': 938376, 'steps': 6516, 'loss/train': 3.747231960296631}
+01/23/2022 08:24:21 - INFO - codeparrot_training - Step 39100: {'lr': 0.0004891589564208482, 'samples': 938400, 'steps': 6516, 'loss/train': 4.077655792236328}
+01/23/2022 08:24:21 - INFO - codeparrot_training - Step 39101: {'lr': 0.0004891589564208482, 'samples': 938424, 'steps': 6516, 'loss/train': 3.229374408721924}
+01/23/2022 08:24:22 - INFO - codeparrot_training - Step 39102: {'lr': 0.0004891589564208482, 'samples': 938448, 'steps': 6516, 'loss/train': 2.9376883506774902}
+01/23/2022 08:24:22 - INFO - codeparrot_training - Step 39103: {'lr': 0.0004891541897423798, 'samples': 938472, 'steps': 6517, 'loss/train': 3.6955370903015137}
+01/23/2022 08:24:23 - INFO - codeparrot_training - Step 39104: {'lr': 0.0004891541897423798, 'samples': 938496, 'steps': 6517, 'loss/train': 3.682662010192871}
+01/23/2022 08:24:23 - INFO - codeparrot_training - Step 39105: {'lr': 0.0004891541897423798, 'samples': 938520, 'steps': 6517, 'loss/train': 2.7815065383911133}
+01/23/2022 08:24:23 - INFO - codeparrot_training - Step 39106: {'lr': 0.0004891541897423798, 'samples': 938544, 'steps': 6517, 'loss/train': 3.721973419189453}
+01/23/2022 08:24:24 - INFO - codeparrot_training - Step 39107: {'lr': 0.0004891541897423798, 'samples': 938568, 'steps': 6517, 'loss/train': 2.9643712043762207}
+01/23/2022 08:24:24 - INFO - codeparrot_training - Step 39108: {'lr': 0.0004891541897423798, 'samples': 938592, 'steps': 6517, 'loss/train': 3.0919277667999268}
+01/23/2022 08:24:25 - INFO - codeparrot_training - Step 39109: {'lr': 0.0004891494220394507, 'samples': 938616, 'steps': 6518, 'loss/train': 3.636679172515869}
+01/23/2022 08:24:25 - INFO - codeparrot_training - Step 39110: {'lr': 0.0004891494220394507, 'samples': 938640, 'steps': 6518, 'loss/train': 2.977193832397461}
+01/23/2022 08:24:25 - INFO - codeparrot_training - Step 39111: {'lr': 0.0004891494220394507, 'samples': 938664, 'steps': 6518, 'loss/train': 1.6148834228515625}
+01/23/2022 08:24:26 - INFO - codeparrot_training - Step 39112: {'lr': 0.0004891494220394507, 'samples': 938688, 'steps': 6518, 'loss/train': 2.456171989440918}
+01/23/2022 08:24:26 - INFO - codeparrot_training - Step 39113: {'lr': 0.0004891494220394507, 'samples': 938712, 'steps': 6518, 'loss/train': 2.59468936920166}
+01/23/2022 08:24:27 - INFO - codeparrot_training - Step 39114: {'lr': 0.0004891494220394507, 'samples': 938736, 'steps': 6518, 'loss/train': 2.4107868671417236}
+01/23/2022 08:24:28 - INFO - codeparrot_training - Step 39115: {'lr': 0.0004891446533120815, 'samples': 938760, 'steps': 6519, 'loss/train': 3.309251070022583}
+01/23/2022 08:24:28 - INFO - codeparrot_training - Step 39116: {'lr': 0.0004891446533120815, 'samples': 938784, 'steps': 6519, 'loss/train': 2.7868404388427734}
+01/23/2022 08:24:28 - INFO - codeparrot_training - Step 39117: {'lr': 0.0004891446533120815, 'samples': 938808, 'steps': 6519, 'loss/train': 5.67004919052124}
+01/23/2022 08:24:29 - INFO - codeparrot_training - Step 39118: {'lr': 0.0004891446533120815, 'samples': 938832, 'steps': 6519, 'loss/train': 2.6071829795837402}
+01/23/2022 08:24:29 - INFO - codeparrot_training - Step 39119: {'lr': 0.0004891446533120815, 'samples': 938856, 'steps': 6519, 'loss/train': 3.425738573074341}
+01/23/2022 08:24:29 - INFO - codeparrot_training - Step 39120: {'lr': 0.0004891446533120815, 'samples': 938880, 'steps': 6519, 'loss/train': 0.9670765995979309}
+01/23/2022 08:24:30 - INFO - codeparrot_training - Step 39121: {'lr': 0.0004891398835602925, 'samples': 938904, 'steps': 6520, 'loss/train': 0.9275011420249939}
+01/23/2022 08:24:30 - INFO - codeparrot_training - Step 39122: {'lr': 0.0004891398835602925, 'samples': 938928, 'steps': 6520, 'loss/train': 3.0346133708953857}
+01/23/2022 08:24:31 - INFO - codeparrot_training - Step 39123: {'lr': 0.0004891398835602925, 'samples': 938952, 'steps': 6520, 'loss/train': 3.8918676376342773}
+01/23/2022 08:24:31 - INFO - codeparrot_training - Step 39124: {'lr': 0.0004891398835602925, 'samples': 938976, 'steps': 6520, 'loss/train': 3.5400021076202393}
+01/23/2022 08:24:31 - INFO - codeparrot_training - Step 39125: {'lr': 0.0004891398835602925, 'samples': 939000, 'steps': 6520, 'loss/train': 3.7843353748321533}
+01/23/2022 08:24:32 - INFO - codeparrot_training - Step 39126: {'lr': 0.0004891398835602925, 'samples': 939024, 'steps': 6520, 'loss/train': 4.135538101196289}
+01/23/2022 08:24:32 - INFO - codeparrot_training - Step 39127: {'lr': 0.0004891351127841041, 'samples': 939048, 'steps': 6521, 'loss/train': 2.9937477111816406}
+01/23/2022 08:24:33 - INFO - codeparrot_training - Step 39128: {'lr': 0.0004891351127841041, 'samples': 939072, 'steps': 6521, 'loss/train': 3.4318594932556152}
+01/23/2022 08:24:33 - INFO - codeparrot_training - Step 39129: {'lr': 0.0004891351127841041, 'samples': 939096, 'steps': 6521, 'loss/train': 3.6747617721557617}
+01/23/2022 08:24:33 - INFO - codeparrot_training - Step 39130: {'lr': 0.0004891351127841041, 'samples': 939120, 'steps': 6521, 'loss/train': 2.8288333415985107}
+01/23/2022 08:24:34 - INFO - codeparrot_training - Step 39131: {'lr': 0.0004891351127841041, 'samples': 939144, 'steps': 6521, 'loss/train': 2.961115598678589}
+01/23/2022 08:24:34 - INFO - codeparrot_training - Step 39132: {'lr': 0.0004891351127841041, 'samples': 939168, 'steps': 6521, 'loss/train': 3.383082866668701}
+01/23/2022 08:24:34 - INFO - codeparrot_training - Step 39133: {'lr': 0.0004891303409835369, 'samples': 939192, 'steps': 6522, 'loss/train': 2.795271635055542}
+01/23/2022 08:24:35 - INFO - codeparrot_training - Step 39134: {'lr': 0.0004891303409835369, 'samples': 939216, 'steps': 6522, 'loss/train': 3.682159662246704}
+01/23/2022 08:24:35 - INFO - codeparrot_training - Step 39135: {'lr': 0.0004891303409835369, 'samples': 939240, 'steps': 6522, 'loss/train': 2.8251142501831055}
+01/23/2022 08:24:36 - INFO - codeparrot_training - Step 39136: {'lr': 0.0004891303409835369, 'samples': 939264, 'steps': 6522, 'loss/train': 3.8328967094421387}
+01/23/2022 08:24:36 - INFO - codeparrot_training - Step 39137: {'lr': 0.0004891303409835369, 'samples': 939288, 'steps': 6522, 'loss/train': 2.4862945079803467}
+01/23/2022 08:24:36 - INFO - codeparrot_training - Step 39138: {'lr': 0.0004891303409835369, 'samples': 939312, 'steps': 6522, 'loss/train': 2.362865924835205}
+01/23/2022 08:24:37 - INFO - codeparrot_training - Step 39139: {'lr': 0.0004891255681586113, 'samples': 939336, 'steps': 6523, 'loss/train': 2.969085454940796}
+01/23/2022 08:24:37 - INFO - codeparrot_training - Step 39140: {'lr': 0.0004891255681586113, 'samples': 939360, 'steps': 6523, 'loss/train': 3.137305498123169}
+01/23/2022 08:24:37 - INFO - codeparrot_training - Step 39141: {'lr': 0.0004891255681586113, 'samples': 939384, 'steps': 6523, 'loss/train': 2.1702699661254883}
+01/23/2022 08:24:38 - INFO - codeparrot_training - Step 39142: {'lr': 0.0004891255681586113, 'samples': 939408, 'steps': 6523, 'loss/train': 3.129179000854492}
+01/23/2022 08:24:38 - INFO - codeparrot_training - Step 39143: {'lr': 0.0004891255681586113, 'samples': 939432, 'steps': 6523, 'loss/train': 3.4345905780792236}
+01/23/2022 08:24:39 - INFO - codeparrot_training - Step 39144: {'lr': 0.0004891255681586113, 'samples': 939456, 'steps': 6523, 'loss/train': 4.52192497253418}
+01/23/2022 08:24:39 - INFO - codeparrot_training - Step 39145: {'lr': 0.0004891207943093476, 'samples': 939480, 'steps': 6524, 'loss/train': 2.947676658630371}
+01/23/2022 08:24:39 - INFO - codeparrot_training - Step 39146: {'lr': 0.0004891207943093476, 'samples': 939504, 'steps': 6524, 'loss/train': 2.4530177116394043}
+01/23/2022 08:24:40 - INFO - codeparrot_training - Step 39147: {'lr': 0.0004891207943093476, 'samples': 939528, 'steps': 6524, 'loss/train': 2.7081611156463623}
+01/23/2022 08:24:40 - INFO - codeparrot_training - Step 39148: {'lr': 0.0004891207943093476, 'samples': 939552, 'steps': 6524, 'loss/train': 3.4979851245880127}
+01/23/2022 08:24:41 - INFO - codeparrot_training - Step 39149: {'lr': 0.0004891207943093476, 'samples': 939576, 'steps': 6524, 'loss/train': 2.603226900100708}
+01/23/2022 08:24:41 - INFO - codeparrot_training - Step 39150: {'lr': 0.0004891207943093476, 'samples': 939600, 'steps': 6524, 'loss/train': 3.3854598999023438}
+01/23/2022 08:24:41 - INFO - codeparrot_training - Step 39151: {'lr': 0.0004891160194357663, 'samples': 939624, 'steps': 6525, 'loss/train': 3.8636884689331055}
+01/23/2022 08:24:42 - INFO - codeparrot_training - Step 39152: {'lr': 0.0004891160194357663, 'samples': 939648, 'steps': 6525, 'loss/train': 3.1098599433898926}
+01/23/2022 08:24:43 - INFO - codeparrot_training - Step 39153: {'lr': 0.0004891160194357663, 'samples': 939672, 'steps': 6525, 'loss/train': 3.3723201751708984}
+01/23/2022 08:24:43 - INFO - codeparrot_training - Step 39154: {'lr': 0.0004891160194357663, 'samples': 939696, 'steps': 6525, 'loss/train': 3.0431630611419678}
+01/23/2022 08:24:44 - INFO - codeparrot_training - Step 39155: {'lr': 0.0004891160194357663, 'samples': 939720, 'steps': 6525, 'loss/train': 3.445808172225952}
+01/23/2022 08:24:44 - INFO - codeparrot_training - Step 39156: {'lr': 0.0004891160194357663, 'samples': 939744, 'steps': 6525, 'loss/train': 3.826186180114746}
+01/23/2022 08:24:44 - INFO - codeparrot_training - Step 39157: {'lr': 0.0004891112435378881, 'samples': 939768, 'steps': 6526, 'loss/train': 3.920912504196167}
+01/23/2022 08:24:45 - INFO - codeparrot_training - Step 39158: {'lr': 0.0004891112435378881, 'samples': 939792, 'steps': 6526, 'loss/train': 4.038571834564209}
+01/23/2022 08:24:45 - INFO - codeparrot_training - Step 39159: {'lr': 0.0004891112435378881, 'samples': 939816, 'steps': 6526, 'loss/train': 5.1516547203063965}
+01/23/2022 08:24:45 - INFO - codeparrot_training - Step 39160: {'lr': 0.0004891112435378881, 'samples': 939840, 'steps': 6526, 'loss/train': 3.636204242706299}
+01/23/2022 08:24:46 - INFO - codeparrot_training - Step 39161: {'lr': 0.0004891112435378881, 'samples': 939864, 'steps': 6526, 'loss/train': 5.447638511657715}
+01/23/2022 08:24:46 - INFO - codeparrot_training - Step 39162: {'lr': 0.0004891112435378881, 'samples': 939888, 'steps': 6526, 'loss/train': 3.3519551753997803}
+01/23/2022 08:24:47 - INFO - codeparrot_training - Step 39163: {'lr': 0.0004891064666157331, 'samples': 939912, 'steps': 6527, 'loss/train': 2.8537988662719727}
+01/23/2022 08:24:47 - INFO - codeparrot_training - Step 39164: {'lr': 0.0004891064666157331, 'samples': 939936, 'steps': 6527, 'loss/train': 2.7723374366760254}
+01/23/2022 08:24:47 - INFO - codeparrot_training - Step 39165: {'lr': 0.0004891064666157331, 'samples': 939960, 'steps': 6527, 'loss/train': 2.5464305877685547}
+01/23/2022 08:24:48 - INFO - codeparrot_training - Step 39166: {'lr': 0.0004891064666157331, 'samples': 939984, 'steps': 6527, 'loss/train': 2.1894421577453613}
+01/23/2022 08:24:48 - INFO - codeparrot_training - Step 39167: {'lr': 0.0004891064666157331, 'samples': 940008, 'steps': 6527, 'loss/train': 3.020317792892456}
+01/23/2022 08:24:48 - INFO - codeparrot_training - Step 39168: {'lr': 0.0004891064666157331, 'samples': 940032, 'steps': 6527, 'loss/train': 1.9449101686477661}
+01/23/2022 08:24:49 - INFO - codeparrot_training - Step 39169: {'lr': 0.0004891016886693219, 'samples': 940056, 'steps': 6528, 'loss/train': 4.0292534828186035}
+01/23/2022 08:24:49 - INFO - codeparrot_training - Step 39170: {'lr': 0.0004891016886693219, 'samples': 940080, 'steps': 6528, 'loss/train': 2.6529619693756104}
+01/23/2022 08:24:50 - INFO - codeparrot_training - Step 39171: {'lr': 0.0004891016886693219, 'samples': 940104, 'steps': 6528, 'loss/train': 4.165742874145508}
+01/23/2022 08:24:50 - INFO - codeparrot_training - Step 39172: {'lr': 0.0004891016886693219, 'samples': 940128, 'steps': 6528, 'loss/train': 2.759321689605713}
+01/23/2022 08:24:50 - INFO - codeparrot_training - Step 39173: {'lr': 0.0004891016886693219, 'samples': 940152, 'steps': 6528, 'loss/train': 3.5519540309906006}
+01/23/2022 08:24:51 - INFO - codeparrot_training - Step 39174: {'lr': 0.0004891016886693219, 'samples': 940176, 'steps': 6528, 'loss/train': 2.954737424850464}
+01/23/2022 08:24:51 - INFO - codeparrot_training - Step 39175: {'lr': 0.0004890969096986751, 'samples': 940200, 'steps': 6529, 'loss/train': 3.369725227355957}
+01/23/2022 08:24:52 - INFO - codeparrot_training - Step 39176: {'lr': 0.0004890969096986751, 'samples': 940224, 'steps': 6529, 'loss/train': 3.554361581802368}
+01/23/2022 08:24:52 - INFO - codeparrot_training - Step 39177: {'lr': 0.0004890969096986751, 'samples': 940248, 'steps': 6529, 'loss/train': 3.527127265930176}
+01/23/2022 08:24:52 - INFO - codeparrot_training - Step 39178: {'lr': 0.0004890969096986751, 'samples': 940272, 'steps': 6529, 'loss/train': 2.388362169265747}
+01/23/2022 08:24:53 - INFO - codeparrot_training - Step 39179: {'lr': 0.0004890969096986751, 'samples': 940296, 'steps': 6529, 'loss/train': 4.104141712188721}
+01/23/2022 08:24:53 - INFO - codeparrot_training - Step 39180: {'lr': 0.0004890969096986751, 'samples': 940320, 'steps': 6529, 'loss/train': 2.7259178161621094}
+01/23/2022 08:24:54 - INFO - codeparrot_training - Step 39181: {'lr': 0.000489092129703813, 'samples': 940344, 'steps': 6530, 'loss/train': 2.2068445682525635}
+01/23/2022 08:24:54 - INFO - codeparrot_training - Step 39182: {'lr': 0.000489092129703813, 'samples': 940368, 'steps': 6530, 'loss/train': 3.863715887069702}
+01/23/2022 08:24:54 - INFO - codeparrot_training - Step 39183: {'lr': 0.000489092129703813, 'samples': 940392, 'steps': 6530, 'loss/train': 3.0754292011260986}
+01/23/2022 08:24:55 - INFO - codeparrot_training - Step 39184: {'lr': 0.000489092129703813, 'samples': 940416, 'steps': 6530, 'loss/train': 3.1997010707855225}
+01/23/2022 08:24:55 - INFO - codeparrot_training - Step 39185: {'lr': 0.000489092129703813, 'samples': 940440, 'steps': 6530, 'loss/train': 3.329169273376465}
+01/23/2022 08:24:55 - INFO - codeparrot_training - Step 39186: {'lr': 0.000489092129703813, 'samples': 940464, 'steps': 6530, 'loss/train': 3.0338399410247803}
+01/23/2022 08:24:56 - INFO - codeparrot_training - Step 39187: {'lr': 0.0004890873486847561, 'samples': 940488, 'steps': 6531, 'loss/train': 3.248319149017334}
+01/23/2022 08:24:56 - INFO - codeparrot_training - Step 39188: {'lr': 0.0004890873486847561, 'samples': 940512, 'steps': 6531, 'loss/train': 3.63550066947937}
+01/23/2022 08:24:57 - INFO - codeparrot_training - Step 39189: {'lr': 0.0004890873486847561, 'samples': 940536, 'steps': 6531, 'loss/train': 3.6914005279541016}
+01/23/2022 08:24:57 - INFO - codeparrot_training - Step 39190: {'lr': 0.0004890873486847561, 'samples': 940560, 'steps': 6531, 'loss/train': 3.647219657897949}
+01/23/2022 08:24:58 - INFO - codeparrot_training - Step 39191: {'lr': 0.0004890873486847561, 'samples': 940584, 'steps': 6531, 'loss/train': 2.3955376148223877}
+01/23/2022 08:24:58 - INFO - codeparrot_training - Step 39192: {'lr': 0.0004890873486847561, 'samples': 940608, 'steps': 6531, 'loss/train': 4.083267688751221}
+01/23/2022 08:24:59 - INFO - codeparrot_training - Step 39193: {'lr': 0.000489082566641525, 'samples': 940632, 'steps': 6532, 'loss/train': 3.417757749557495}
+01/23/2022 08:24:59 - INFO - codeparrot_training - Step 39194: {'lr': 0.000489082566641525, 'samples': 940656, 'steps': 6532, 'loss/train': 2.748868465423584}
+01/23/2022 08:24:59 - INFO - codeparrot_training - Step 39195: {'lr': 0.000489082566641525, 'samples': 940680, 'steps': 6532, 'loss/train': 4.329190254211426}
+01/23/2022 08:25:00 - INFO - codeparrot_training - Step 39196: {'lr': 0.000489082566641525, 'samples': 940704, 'steps': 6532, 'loss/train': 2.303649425506592}
+01/23/2022 08:25:00 - INFO - codeparrot_training - Step 39197: {'lr': 0.000489082566641525, 'samples': 940728, 'steps': 6532, 'loss/train': 2.711116313934326}
+01/23/2022 08:25:01 - INFO - codeparrot_training - Step 39198: {'lr': 0.000489082566641525, 'samples': 940752, 'steps': 6532, 'loss/train': 2.602043867111206}
+01/23/2022 08:25:01 - INFO - codeparrot_training - Step 39199: {'lr': 0.00048907778357414, 'samples': 940776, 'steps': 6533, 'loss/train': 1.9942946434020996}
+01/23/2022 08:25:01 - INFO - codeparrot_training - Step 39200: {'lr': 0.00048907778357414, 'samples': 940800, 'steps': 6533, 'loss/train': 4.03064489364624}
+01/23/2022 08:25:02 - INFO - codeparrot_training - Step 39201: {'lr': 0.00048907778357414, 'samples': 940824, 'steps': 6533, 'loss/train': 1.9251554012298584}
+01/23/2022 08:25:02 - INFO - codeparrot_training - Step 39202: {'lr': 0.00048907778357414, 'samples': 940848, 'steps': 6533, 'loss/train': 2.838017702102661}
+01/23/2022 08:25:03 - INFO - codeparrot_training - Step 39203: {'lr': 0.00048907778357414, 'samples': 940872, 'steps': 6533, 'loss/train': 4.388152122497559}
+01/23/2022 08:25:03 - INFO - codeparrot_training - Step 39204: {'lr': 0.00048907778357414, 'samples': 940896, 'steps': 6533, 'loss/train': 4.175238609313965}
+01/23/2022 08:25:03 - INFO - codeparrot_training - Step 39205: {'lr': 0.0004890729994826218, 'samples': 940920, 'steps': 6534, 'loss/train': 2.7356162071228027}
+01/23/2022 08:25:04 - INFO - codeparrot_training - Step 39206: {'lr': 0.0004890729994826218, 'samples': 940944, 'steps': 6534, 'loss/train': 2.295320749282837}
+01/23/2022 08:25:04 - INFO - codeparrot_training - Step 39207: {'lr': 0.0004890729994826218, 'samples': 940968, 'steps': 6534, 'loss/train': 3.635976791381836}
+01/23/2022 08:25:04 - INFO - codeparrot_training - Step 39208: {'lr': 0.0004890729994826218, 'samples': 940992, 'steps': 6534, 'loss/train': 2.590236186981201}
+01/23/2022 08:25:05 - INFO - codeparrot_training - Step 39209: {'lr': 0.0004890729994826218, 'samples': 941016, 'steps': 6534, 'loss/train': 4.52496862411499}
+01/23/2022 08:25:05 - INFO - codeparrot_training - Step 39210: {'lr': 0.0004890729994826218, 'samples': 941040, 'steps': 6534, 'loss/train': 3.1678621768951416}
+01/23/2022 08:25:06 - INFO - codeparrot_training - Step 39211: {'lr': 0.0004890682143669908, 'samples': 941064, 'steps': 6535, 'loss/train': 2.908327579498291}
+01/23/2022 08:25:06 - INFO - codeparrot_training - Step 39212: {'lr': 0.0004890682143669908, 'samples': 941088, 'steps': 6535, 'loss/train': 2.4184861183166504}
+01/23/2022 08:25:06 - INFO - codeparrot_training - Step 39213: {'lr': 0.0004890682143669908, 'samples': 941112, 'steps': 6535, 'loss/train': 3.212529420852661}
+01/23/2022 08:25:07 - INFO - codeparrot_training - Step 39214: {'lr': 0.0004890682143669908, 'samples': 941136, 'steps': 6535, 'loss/train': 3.940927267074585}
+01/23/2022 08:25:07 - INFO - codeparrot_training - Step 39215: {'lr': 0.0004890682143669908, 'samples': 941160, 'steps': 6535, 'loss/train': 3.366302967071533}
+01/23/2022 08:25:08 - INFO - codeparrot_training - Step 39216: {'lr': 0.0004890682143669908, 'samples': 941184, 'steps': 6535, 'loss/train': 5.098115921020508}
+01/23/2022 08:25:08 - INFO - codeparrot_training - Step 39217: {'lr': 0.0004890634282272673, 'samples': 941208, 'steps': 6536, 'loss/train': 1.422756314277649}
+01/23/2022 08:25:08 - INFO - codeparrot_training - Step 39218: {'lr': 0.0004890634282272673, 'samples': 941232, 'steps': 6536, 'loss/train': 3.8572499752044678}
+01/23/2022 08:25:09 - INFO - codeparrot_training - Step 39219: {'lr': 0.0004890634282272673, 'samples': 941256, 'steps': 6536, 'loss/train': 3.4178547859191895}
+01/23/2022 08:25:09 - INFO - codeparrot_training - Step 39220: {'lr': 0.0004890634282272673, 'samples': 941280, 'steps': 6536, 'loss/train': 2.9708850383758545}
+01/23/2022 08:25:09 - INFO - codeparrot_training - Step 39221: {'lr': 0.0004890634282272673, 'samples': 941304, 'steps': 6536, 'loss/train': 3.263442039489746}
+01/23/2022 08:25:10 - INFO - codeparrot_training - Step 39222: {'lr': 0.0004890634282272673, 'samples': 941328, 'steps': 6536, 'loss/train': 3.2819325923919678}
+01/23/2022 08:25:10 - INFO - codeparrot_training - Step 39223: {'lr': 0.0004890586410634722, 'samples': 941352, 'steps': 6537, 'loss/train': 3.3370141983032227}
+01/23/2022 08:25:11 - INFO - codeparrot_training - Step 39224: {'lr': 0.0004890586410634722, 'samples': 941376, 'steps': 6537, 'loss/train': 3.8802387714385986}
+01/23/2022 08:25:11 - INFO - codeparrot_training - Step 39225: {'lr': 0.0004890586410634722, 'samples': 941400, 'steps': 6537, 'loss/train': 4.353004455566406}
+01/23/2022 08:25:11 - INFO - codeparrot_training - Step 39226: {'lr': 0.0004890586410634722, 'samples': 941424, 'steps': 6537, 'loss/train': 2.9441096782684326}
+01/23/2022 08:25:12 - INFO - codeparrot_training - Step 39227: {'lr': 0.0004890586410634722, 'samples': 941448, 'steps': 6537, 'loss/train': 1.996368646621704}
+01/23/2022 08:25:12 - INFO - codeparrot_training - Step 39228: {'lr': 0.0004890586410634722, 'samples': 941472, 'steps': 6537, 'loss/train': 3.980069398880005}
+01/23/2022 08:25:13 - INFO - codeparrot_training - Step 39229: {'lr': 0.0004890538528756256, 'samples': 941496, 'steps': 6538, 'loss/train': 2.7561724185943604}
+01/23/2022 08:25:13 - INFO - codeparrot_training - Step 39230: {'lr': 0.0004890538528756256, 'samples': 941520, 'steps': 6538, 'loss/train': 5.992818355560303}
+01/23/2022 08:25:13 - INFO - codeparrot_training - Step 39231: {'lr': 0.0004890538528756256, 'samples': 941544, 'steps': 6538, 'loss/train': 2.749002695083618}
+01/23/2022 08:25:16 - INFO - codeparrot_training - Step 39232: {'lr': 0.0004890538528756256, 'samples': 941568, 'steps': 6538, 'loss/train': 2.2302212715148926}
+01/23/2022 08:25:16 - INFO - codeparrot_training - Step 39233: {'lr': 0.0004890538528756256, 'samples': 941592, 'steps': 6538, 'loss/train': 3.4393181800842285}
+01/23/2022 08:25:17 - INFO - codeparrot_training - Step 39234: {'lr': 0.0004890538528756256, 'samples': 941616, 'steps': 6538, 'loss/train': 3.53908634185791}
+01/23/2022 08:25:17 - INFO - codeparrot_training - Step 39235: {'lr': 0.0004890490636637484, 'samples': 941640, 'steps': 6539, 'loss/train': 4.001498222351074}
+01/23/2022 08:25:18 - INFO - codeparrot_training - Step 39236: {'lr': 0.0004890490636637484, 'samples': 941664, 'steps': 6539, 'loss/train': 4.907427787780762}
+01/23/2022 08:25:18 - INFO - codeparrot_training - Step 39237: {'lr': 0.0004890490636637484, 'samples': 941688, 'steps': 6539, 'loss/train': 3.7552754878997803}
+01/23/2022 08:25:18 - INFO - codeparrot_training - Step 39238: {'lr': 0.0004890490636637484, 'samples': 941712, 'steps': 6539, 'loss/train': 3.0433897972106934}
+01/23/2022 08:25:19 - INFO - codeparrot_training - Step 39239: {'lr': 0.0004890490636637484, 'samples': 941736, 'steps': 6539, 'loss/train': 1.6125813722610474}
+01/23/2022 08:25:19 - INFO - codeparrot_training - Step 39240: {'lr': 0.0004890490636637484, 'samples': 941760, 'steps': 6539, 'loss/train': 3.544696807861328}
+01/23/2022 08:25:19 - INFO - codeparrot_training - Step 39241: {'lr': 0.0004890442734278608, 'samples': 941784, 'steps': 6540, 'loss/train': 2.6960835456848145}
+01/23/2022 08:25:20 - INFO - codeparrot_training - Step 39242: {'lr': 0.0004890442734278608, 'samples': 941808, 'steps': 6540, 'loss/train': 3.6674981117248535}
+01/23/2022 08:25:20 - INFO - codeparrot_training - Step 39243: {'lr': 0.0004890442734278608, 'samples': 941832, 'steps': 6540, 'loss/train': 3.0548593997955322}
+01/23/2022 08:25:21 - INFO - codeparrot_training - Step 39244: {'lr': 0.0004890442734278608, 'samples': 941856, 'steps': 6540, 'loss/train': 3.3730030059814453}
+01/23/2022 08:25:21 - INFO - codeparrot_training - Step 39245: {'lr': 0.0004890442734278608, 'samples': 941880, 'steps': 6540, 'loss/train': 2.6314852237701416}
+01/23/2022 08:25:21 - INFO - codeparrot_training - Step 39246: {'lr': 0.0004890442734278608, 'samples': 941904, 'steps': 6540, 'loss/train': 2.0491297245025635}
+01/23/2022 08:25:22 - INFO - codeparrot_training - Step 39247: {'lr': 0.0004890394821679834, 'samples': 941928, 'steps': 6541, 'loss/train': 4.940652847290039}
+01/23/2022 08:25:22 - INFO - codeparrot_training - Step 39248: {'lr': 0.0004890394821679834, 'samples': 941952, 'steps': 6541, 'loss/train': 1.5810532569885254}
+01/23/2022 08:25:23 - INFO - codeparrot_training - Step 39249: {'lr': 0.0004890394821679834, 'samples': 941976, 'steps': 6541, 'loss/train': 3.7880072593688965}
+01/23/2022 08:25:23 - INFO - codeparrot_training - Step 39250: {'lr': 0.0004890394821679834, 'samples': 942000, 'steps': 6541, 'loss/train': 2.7705631256103516}
+01/23/2022 08:25:23 - INFO - codeparrot_training - Step 39251: {'lr': 0.0004890394821679834, 'samples': 942024, 'steps': 6541, 'loss/train': 2.756967782974243}
+01/23/2022 08:25:24 - INFO - codeparrot_training - Step 39252: {'lr': 0.0004890394821679834, 'samples': 942048, 'steps': 6541, 'loss/train': 3.3573343753814697}
+01/23/2022 08:25:24 - INFO - codeparrot_training - Step 39253: {'lr': 0.0004890346898841369, 'samples': 942072, 'steps': 6542, 'loss/train': 1.1063586473464966}
+01/23/2022 08:25:24 - INFO - codeparrot_training - Step 39254: {'lr': 0.0004890346898841369, 'samples': 942096, 'steps': 6542, 'loss/train': 4.138541221618652}
+01/23/2022 08:25:25 - INFO - codeparrot_training - Step 39255: {'lr': 0.0004890346898841369, 'samples': 942120, 'steps': 6542, 'loss/train': 3.3824446201324463}
+01/23/2022 08:25:25 - INFO - codeparrot_training - Step 39256: {'lr': 0.0004890346898841369, 'samples': 942144, 'steps': 6542, 'loss/train': 3.6520512104034424}
+01/23/2022 08:25:26 - INFO - codeparrot_training - Step 39257: {'lr': 0.0004890346898841369, 'samples': 942168, 'steps': 6542, 'loss/train': 3.3276782035827637}
+01/23/2022 08:25:26 - INFO - codeparrot_training - Step 39258: {'lr': 0.0004890346898841369, 'samples': 942192, 'steps': 6542, 'loss/train': 3.494051218032837}
+01/23/2022 08:25:26 - INFO - codeparrot_training - Step 39259: {'lr': 0.0004890298965763416, 'samples': 942216, 'steps': 6543, 'loss/train': 0.7514816522598267}
+01/23/2022 08:25:27 - INFO - codeparrot_training - Step 39260: {'lr': 0.0004890298965763416, 'samples': 942240, 'steps': 6543, 'loss/train': 2.8473079204559326}
+01/23/2022 08:25:27 - INFO - codeparrot_training - Step 39261: {'lr': 0.0004890298965763416, 'samples': 942264, 'steps': 6543, 'loss/train': 3.622958183288574}
+01/23/2022 08:25:27 - INFO - codeparrot_training - Step 39262: {'lr': 0.0004890298965763416, 'samples': 942288, 'steps': 6543, 'loss/train': 2.8083150386810303}
+01/23/2022 08:25:28 - INFO - codeparrot_training - Step 39263: {'lr': 0.0004890298965763416, 'samples': 942312, 'steps': 6543, 'loss/train': 3.316075325012207}
+01/23/2022 08:25:28 - INFO - codeparrot_training - Step 39264: {'lr': 0.0004890298965763416, 'samples': 942336, 'steps': 6543, 'loss/train': 2.8991243839263916}
+01/23/2022 08:25:29 - INFO - codeparrot_training - Step 39265: {'lr': 0.0004890251022446181, 'samples': 942360, 'steps': 6544, 'loss/train': 2.899083137512207}
+01/23/2022 08:25:29 - INFO - codeparrot_training - Step 39266: {'lr': 0.0004890251022446181, 'samples': 942384, 'steps': 6544, 'loss/train': 2.5222766399383545}
+01/23/2022 08:25:29 - INFO - codeparrot_training - Step 39267: {'lr': 0.0004890251022446181, 'samples': 942408, 'steps': 6544, 'loss/train': 2.2005720138549805}
+01/23/2022 08:25:30 - INFO - codeparrot_training - Step 39268: {'lr': 0.0004890251022446181, 'samples': 942432, 'steps': 6544, 'loss/train': 3.1782164573669434}
+01/23/2022 08:25:30 - INFO - codeparrot_training - Step 39269: {'lr': 0.0004890251022446181, 'samples': 942456, 'steps': 6544, 'loss/train': 3.7653255462646484}
+01/23/2022 08:25:30 - INFO - codeparrot_training - Step 39270: {'lr': 0.0004890251022446181, 'samples': 942480, 'steps': 6544, 'loss/train': 3.2876429557800293}
+01/23/2022 08:25:33 - INFO - codeparrot_training - Step 39271: {'lr': 0.000489020306888987, 'samples': 942504, 'steps': 6545, 'loss/train': 3.9574105739593506}
+01/23/2022 08:25:33 - INFO - codeparrot_training - Step 39272: {'lr': 0.000489020306888987, 'samples': 942528, 'steps': 6545, 'loss/train': 2.4164879322052}
+01/23/2022 08:25:34 - INFO - codeparrot_training - Step 39273: {'lr': 0.000489020306888987, 'samples': 942552, 'steps': 6545, 'loss/train': 3.084165096282959}
+01/23/2022 08:25:34 - INFO - codeparrot_training - Step 39274: {'lr': 0.000489020306888987, 'samples': 942576, 'steps': 6545, 'loss/train': 2.34128737449646}
+01/23/2022 08:25:34 - INFO - codeparrot_training - Step 39275: {'lr': 0.000489020306888987, 'samples': 942600, 'steps': 6545, 'loss/train': 2.260530471801758}
+01/23/2022 08:25:35 - INFO - codeparrot_training - Step 39276: {'lr': 0.000489020306888987, 'samples': 942624, 'steps': 6545, 'loss/train': 2.353102684020996}
+01/23/2022 08:25:35 - INFO - codeparrot_training - Step 39277: {'lr': 0.0004890155105094688, 'samples': 942648, 'steps': 6546, 'loss/train': 2.0295703411102295}
+01/23/2022 08:25:35 - INFO - codeparrot_training - Step 39278: {'lr': 0.0004890155105094688, 'samples': 942672, 'steps': 6546, 'loss/train': 2.363706111907959}
+01/23/2022 08:25:36 - INFO - codeparrot_training - Step 39279: {'lr': 0.0004890155105094688, 'samples': 942696, 'steps': 6546, 'loss/train': 1.9831182956695557}
+01/23/2022 08:25:36 - INFO - codeparrot_training - Step 39280: {'lr': 0.0004890155105094688, 'samples': 942720, 'steps': 6546, 'loss/train': 2.1120543479919434}
+01/23/2022 08:25:37 - INFO - codeparrot_training - Step 39281: {'lr': 0.0004890155105094688, 'samples': 942744, 'steps': 6546, 'loss/train': 2.01114559173584}
+01/23/2022 08:25:37 - INFO - codeparrot_training - Step 39282: {'lr': 0.0004890155105094688, 'samples': 942768, 'steps': 6546, 'loss/train': 2.0247790813446045}
+01/23/2022 08:25:37 - INFO - codeparrot_training - Step 39283: {'lr': 0.0004890107131060841, 'samples': 942792, 'steps': 6547, 'loss/train': 1.7621902227401733}
+01/23/2022 08:25:38 - INFO - codeparrot_training - Step 39284: {'lr': 0.0004890107131060841, 'samples': 942816, 'steps': 6547, 'loss/train': 1.7940438985824585}
+01/23/2022 08:25:38 - INFO - codeparrot_training - Step 39285: {'lr': 0.0004890107131060841, 'samples': 942840, 'steps': 6547, 'loss/train': 1.6856672763824463}
+01/23/2022 08:25:38 - INFO - codeparrot_training - Step 39286: {'lr': 0.0004890107131060841, 'samples': 942864, 'steps': 6547, 'loss/train': 1.5781887769699097}
+01/23/2022 08:25:39 - INFO - codeparrot_training - Step 39287: {'lr': 0.0004890107131060841, 'samples': 942888, 'steps': 6547, 'loss/train': 1.572725772857666}
+01/23/2022 08:25:39 - INFO - codeparrot_training - Step 39288: {'lr': 0.0004890107131060841, 'samples': 942912, 'steps': 6547, 'loss/train': 1.7530122995376587}
+01/23/2022 08:25:40 - INFO - codeparrot_training - Step 39289: {'lr': 0.0004890059146788532, 'samples': 942936, 'steps': 6548, 'loss/train': 3.003816843032837}
+01/23/2022 08:25:40 - INFO - codeparrot_training - Step 39290: {'lr': 0.0004890059146788532, 'samples': 942960, 'steps': 6548, 'loss/train': 3.0069355964660645}
+01/23/2022 08:25:40 - INFO - codeparrot_training - Step 39291: {'lr': 0.0004890059146788532, 'samples': 942984, 'steps': 6548, 'loss/train': 1.605518102645874}
+01/23/2022 08:25:41 - INFO - codeparrot_training - Step 39292: {'lr': 0.0004890059146788532, 'samples': 943008, 'steps': 6548, 'loss/train': 1.3311513662338257}
+01/23/2022 08:25:41 - INFO - codeparrot_training - Step 39293: {'lr': 0.0004890059146788532, 'samples': 943032, 'steps': 6548, 'loss/train': 1.6376621723175049}
+01/23/2022 08:25:41 - INFO - codeparrot_training - Step 39294: {'lr': 0.0004890059146788532, 'samples': 943056, 'steps': 6548, 'loss/train': 1.5334742069244385}
+01/23/2022 08:25:42 - INFO - codeparrot_training - Step 39295: {'lr': 0.000489001115227797, 'samples': 943080, 'steps': 6549, 'loss/train': 1.6102607250213623}
+01/23/2022 08:25:42 - INFO - codeparrot_training - Step 39296: {'lr': 0.000489001115227797, 'samples': 943104, 'steps': 6549, 'loss/train': 1.5710971355438232}
+01/23/2022 08:25:43 - INFO - codeparrot_training - Step 39297: {'lr': 0.000489001115227797, 'samples': 943128, 'steps': 6549, 'loss/train': 1.5178031921386719}
+01/23/2022 08:25:43 - INFO - codeparrot_training - Step 39298: {'lr': 0.000489001115227797, 'samples': 943152, 'steps': 6549, 'loss/train': 1.5835461616516113}
+01/23/2022 08:25:43 - INFO - codeparrot_training - Step 39299: {'lr': 0.000489001115227797, 'samples': 943176, 'steps': 6549, 'loss/train': 1.5509699583053589}
+01/23/2022 08:25:44 - INFO - codeparrot_training - Step 39300: {'lr': 0.000489001115227797, 'samples': 943200, 'steps': 6549, 'loss/train': 1.6569252014160156}
+01/23/2022 08:25:44 - INFO - codeparrot_training - Step 39301: {'lr': 0.000488996314752936, 'samples': 943224, 'steps': 6550, 'loss/train': 1.3460016250610352}
+01/23/2022 08:25:45 - INFO - codeparrot_training - Step 39302: {'lr': 0.000488996314752936, 'samples': 943248, 'steps': 6550, 'loss/train': 1.3740395307540894}
+01/23/2022 08:25:45 - INFO - codeparrot_training - Step 39303: {'lr': 0.000488996314752936, 'samples': 943272, 'steps': 6550, 'loss/train': 1.3426997661590576}
+01/23/2022 08:25:45 - INFO - codeparrot_training - Step 39304: {'lr': 0.000488996314752936, 'samples': 943296, 'steps': 6550, 'loss/train': 2.3223776817321777}
+01/23/2022 08:25:46 - INFO - codeparrot_training - Step 39305: {'lr': 0.000488996314752936, 'samples': 943320, 'steps': 6550, 'loss/train': 4.411060810089111}
+01/23/2022 08:25:46 - INFO - codeparrot_training - Step 39306: {'lr': 0.000488996314752936, 'samples': 943344, 'steps': 6550, 'loss/train': 4.006638526916504}
+01/23/2022 08:25:47 - INFO - codeparrot_training - Step 39307: {'lr': 0.0004889915132542906, 'samples': 943368, 'steps': 6551, 'loss/train': 3.476055383682251}
+01/23/2022 08:25:47 - INFO - codeparrot_training - Step 39308: {'lr': 0.0004889915132542906, 'samples': 943392, 'steps': 6551, 'loss/train': 3.4641242027282715}
+01/23/2022 08:25:47 - INFO - codeparrot_training - Step 39309: {'lr': 0.0004889915132542906, 'samples': 943416, 'steps': 6551, 'loss/train': 3.791215419769287}
+01/23/2022 08:25:48 - INFO - codeparrot_training - Step 39310: {'lr': 0.0004889915132542906, 'samples': 943440, 'steps': 6551, 'loss/train': 4.159421920776367}
+01/23/2022 08:25:48 - INFO - codeparrot_training - Step 39311: {'lr': 0.0004889915132542906, 'samples': 943464, 'steps': 6551, 'loss/train': 2.960364818572998}
+01/23/2022 08:25:48 - INFO - codeparrot_training - Step 39312: {'lr': 0.0004889915132542906, 'samples': 943488, 'steps': 6551, 'loss/train': 3.121450662612915}
+01/23/2022 08:25:49 - INFO - codeparrot_training - Step 39313: {'lr': 0.0004889867107318814, 'samples': 943512, 'steps': 6552, 'loss/train': 3.0941808223724365}
+01/23/2022 08:25:49 - INFO - codeparrot_training - Step 39314: {'lr': 0.0004889867107318814, 'samples': 943536, 'steps': 6552, 'loss/train': 4.090391159057617}
+01/23/2022 08:25:50 - INFO - codeparrot_training - Step 39315: {'lr': 0.0004889867107318814, 'samples': 943560, 'steps': 6552, 'loss/train': 2.9871439933776855}
+01/23/2022 08:25:50 - INFO - codeparrot_training - Step 39316: {'lr': 0.0004889867107318814, 'samples': 943584, 'steps': 6552, 'loss/train': 3.1008009910583496}
+01/23/2022 08:25:50 - INFO - codeparrot_training - Step 39317: {'lr': 0.0004889867107318814, 'samples': 943608, 'steps': 6552, 'loss/train': 3.6593143939971924}
+01/23/2022 08:25:51 - INFO - codeparrot_training - Step 39318: {'lr': 0.0004889867107318814, 'samples': 943632, 'steps': 6552, 'loss/train': 2.3937888145446777}
+01/23/2022 08:25:51 - INFO - codeparrot_training - Step 39319: {'lr': 0.0004889819071857291, 'samples': 943656, 'steps': 6553, 'loss/train': 3.004699230194092}
+01/23/2022 08:25:52 - INFO - codeparrot_training - Step 39320: {'lr': 0.0004889819071857291, 'samples': 943680, 'steps': 6553, 'loss/train': 3.10282826423645}
+01/23/2022 08:25:52 - INFO - codeparrot_training - Step 39321: {'lr': 0.0004889819071857291, 'samples': 943704, 'steps': 6553, 'loss/train': 3.5169944763183594}
+01/23/2022 08:25:52 - INFO - codeparrot_training - Step 39322: {'lr': 0.0004889819071857291, 'samples': 943728, 'steps': 6553, 'loss/train': 3.839998483657837}
+01/23/2022 08:25:53 - INFO - codeparrot_training - Step 39323: {'lr': 0.0004889819071857291, 'samples': 943752, 'steps': 6553, 'loss/train': 1.666954517364502}
+01/23/2022 08:25:53 - INFO - codeparrot_training - Step 39324: {'lr': 0.0004889819071857291, 'samples': 943776, 'steps': 6553, 'loss/train': 1.9880118370056152}
+01/23/2022 08:25:53 - INFO - codeparrot_training - Step 39325: {'lr': 0.0004889771026158541, 'samples': 943800, 'steps': 6554, 'loss/train': 1.4642125368118286}
+01/23/2022 08:25:54 - INFO - codeparrot_training - Step 39326: {'lr': 0.0004889771026158541, 'samples': 943824, 'steps': 6554, 'loss/train': 1.2414017915725708}
+01/23/2022 08:25:54 - INFO - codeparrot_training - Step 39327: {'lr': 0.0004889771026158541, 'samples': 943848, 'steps': 6554, 'loss/train': 2.9689817428588867}
+01/23/2022 08:25:55 - INFO - codeparrot_training - Step 39328: {'lr': 0.0004889771026158541, 'samples': 943872, 'steps': 6554, 'loss/train': 2.6785709857940674}
+01/23/2022 08:25:55 - INFO - codeparrot_training - Step 39329: {'lr': 0.0004889771026158541, 'samples': 943896, 'steps': 6554, 'loss/train': 3.410048723220825}
+01/23/2022 08:25:55 - INFO - codeparrot_training - Step 39330: {'lr': 0.0004889771026158541, 'samples': 943920, 'steps': 6554, 'loss/train': 1.6497374773025513}
+01/23/2022 08:25:56 - INFO - codeparrot_training - Step 39331: {'lr': 0.0004889722970222772, 'samples': 943944, 'steps': 6555, 'loss/train': 2.808699131011963}
+01/23/2022 08:25:57 - INFO - codeparrot_training - Step 39332: {'lr': 0.0004889722970222772, 'samples': 943968, 'steps': 6555, 'loss/train': 3.1877827644348145}
+01/23/2022 08:25:57 - INFO - codeparrot_training - Step 39333: {'lr': 0.0004889722970222772, 'samples': 943992, 'steps': 6555, 'loss/train': 2.225607395172119}
+01/23/2022 08:25:58 - INFO - codeparrot_training - Step 39334: {'lr': 0.0004889722970222772, 'samples': 944016, 'steps': 6555, 'loss/train': 3.1820998191833496}
+01/23/2022 08:25:58 - INFO - codeparrot_training - Step 39335: {'lr': 0.0004889722970222772, 'samples': 944040, 'steps': 6555, 'loss/train': 3.3809847831726074}
+01/23/2022 08:25:59 - INFO - codeparrot_training - Step 39336: {'lr': 0.0004889722970222772, 'samples': 944064, 'steps': 6555, 'loss/train': 3.9144601821899414}
+01/23/2022 08:25:59 - INFO - codeparrot_training - Step 39337: {'lr': 0.0004889674904050188, 'samples': 944088, 'steps': 6556, 'loss/train': 3.6291136741638184}
+01/23/2022 08:25:59 - INFO - codeparrot_training - Step 39338: {'lr': 0.0004889674904050188, 'samples': 944112, 'steps': 6556, 'loss/train': 3.417271614074707}
+01/23/2022 08:26:00 - INFO - codeparrot_training - Step 39339: {'lr': 0.0004889674904050188, 'samples': 944136, 'steps': 6556, 'loss/train': 3.154315948486328}
+01/23/2022 08:26:00 - INFO - codeparrot_training - Step 39340: {'lr': 0.0004889674904050188, 'samples': 944160, 'steps': 6556, 'loss/train': 3.287968158721924}
+01/23/2022 08:26:00 - INFO - codeparrot_training - Step 39341: {'lr': 0.0004889674904050188, 'samples': 944184, 'steps': 6556, 'loss/train': 3.92545223236084}
+01/23/2022 08:26:01 - INFO - codeparrot_training - Step 39342: {'lr': 0.0004889674904050188, 'samples': 944208, 'steps': 6556, 'loss/train': 2.7249085903167725}
+01/23/2022 08:26:01 - INFO - codeparrot_training - Step 39343: {'lr': 0.0004889626827640994, 'samples': 944232, 'steps': 6557, 'loss/train': 3.6261143684387207}
+01/23/2022 08:26:02 - INFO - codeparrot_training - Step 39344: {'lr': 0.0004889626827640994, 'samples': 944256, 'steps': 6557, 'loss/train': 2.6044914722442627}
+01/23/2022 08:26:02 - INFO - codeparrot_training - Step 39345: {'lr': 0.0004889626827640994, 'samples': 944280, 'steps': 6557, 'loss/train': 3.7463324069976807}
+01/23/2022 08:26:02 - INFO - codeparrot_training - Step 39346: {'lr': 0.0004889626827640994, 'samples': 944304, 'steps': 6557, 'loss/train': 5.577484607696533}
+01/23/2022 08:26:03 - INFO - codeparrot_training - Step 39347: {'lr': 0.0004889626827640994, 'samples': 944328, 'steps': 6557, 'loss/train': 2.1946232318878174}
+01/23/2022 08:26:03 - INFO - codeparrot_training - Step 39348: {'lr': 0.0004889626827640994, 'samples': 944352, 'steps': 6557, 'loss/train': 3.1425564289093018}
+01/23/2022 08:26:04 - INFO - codeparrot_training - Step 39349: {'lr': 0.00048895787409954, 'samples': 944376, 'steps': 6558, 'loss/train': 4.192150115966797}
+01/23/2022 08:26:04 - INFO - codeparrot_training - Step 39350: {'lr': 0.00048895787409954, 'samples': 944400, 'steps': 6558, 'loss/train': 4.541668891906738}
+01/23/2022 08:26:04 - INFO - codeparrot_training - Step 39351: {'lr': 0.00048895787409954, 'samples': 944424, 'steps': 6558, 'loss/train': 3.380622625350952}
+01/23/2022 08:26:05 - INFO - codeparrot_training - Step 39352: {'lr': 0.00048895787409954, 'samples': 944448, 'steps': 6558, 'loss/train': 4.9103474617004395}
+01/23/2022 08:26:05 - INFO - codeparrot_training - Step 39353: {'lr': 0.00048895787409954, 'samples': 944472, 'steps': 6558, 'loss/train': 3.1113128662109375}
+01/23/2022 08:26:05 - INFO - codeparrot_training - Step 39354: {'lr': 0.00048895787409954, 'samples': 944496, 'steps': 6558, 'loss/train': 3.1461851596832275}
+01/23/2022 08:26:06 - INFO - codeparrot_training - Step 39355: {'lr': 0.0004889530644113608, 'samples': 944520, 'steps': 6559, 'loss/train': 2.970012664794922}
+01/23/2022 08:26:06 - INFO - codeparrot_training - Step 39356: {'lr': 0.0004889530644113608, 'samples': 944544, 'steps': 6559, 'loss/train': 3.9289684295654297}
+01/23/2022 08:26:07 - INFO - codeparrot_training - Step 39357: {'lr': 0.0004889530644113608, 'samples': 944568, 'steps': 6559, 'loss/train': 3.3668506145477295}
+01/23/2022 08:26:07 - INFO - codeparrot_training - Step 39358: {'lr': 0.0004889530644113608, 'samples': 944592, 'steps': 6559, 'loss/train': 1.8048347234725952}
+01/23/2022 08:26:07 - INFO - codeparrot_training - Step 39359: {'lr': 0.0004889530644113608, 'samples': 944616, 'steps': 6559, 'loss/train': 3.585874080657959}
+01/23/2022 08:26:08 - INFO - codeparrot_training - Step 39360: {'lr': 0.0004889530644113608, 'samples': 944640, 'steps': 6559, 'loss/train': 4.976003170013428}
+01/23/2022 08:26:08 - INFO - codeparrot_training - Step 39361: {'lr': 0.0004889482536995825, 'samples': 944664, 'steps': 6560, 'loss/train': 2.288273572921753}
+01/23/2022 08:26:09 - INFO - codeparrot_training - Step 39362: {'lr': 0.0004889482536995825, 'samples': 944688, 'steps': 6560, 'loss/train': 2.5408318042755127}
+01/23/2022 08:26:09 - INFO - codeparrot_training - Step 39363: {'lr': 0.0004889482536995825, 'samples': 944712, 'steps': 6560, 'loss/train': 2.6989176273345947}
+01/23/2022 08:26:09 - INFO - codeparrot_training - Step 39364: {'lr': 0.0004889482536995825, 'samples': 944736, 'steps': 6560, 'loss/train': 2.1841070652008057}
+01/23/2022 08:26:10 - INFO - codeparrot_training - Step 39365: {'lr': 0.0004889482536995825, 'samples': 944760, 'steps': 6560, 'loss/train': 2.013079881668091}
+01/23/2022 08:26:10 - INFO - codeparrot_training - Step 39366: {'lr': 0.0004889482536995825, 'samples': 944784, 'steps': 6560, 'loss/train': 1.8912492990493774}
+01/23/2022 08:26:10 - INFO - codeparrot_training - Step 39367: {'lr': 0.0004889434419642259, 'samples': 944808, 'steps': 6561, 'loss/train': 1.4756205081939697}
+01/23/2022 08:26:11 - INFO - codeparrot_training - Step 39368: {'lr': 0.0004889434419642259, 'samples': 944832, 'steps': 6561, 'loss/train': 1.2945584058761597}
+01/23/2022 08:26:11 - INFO - codeparrot_training - Step 39369: {'lr': 0.0004889434419642259, 'samples': 944856, 'steps': 6561, 'loss/train': 1.6135460138320923}
+01/23/2022 08:26:12 - INFO - codeparrot_training - Step 39370: {'lr': 0.0004889434419642259, 'samples': 944880, 'steps': 6561, 'loss/train': 2.472020387649536}
+01/23/2022 08:26:12 - INFO - codeparrot_training - Step 39371: {'lr': 0.0004889434419642259, 'samples': 944904, 'steps': 6561, 'loss/train': 2.923487424850464}
+01/23/2022 08:26:12 - INFO - codeparrot_training - Step 39372: {'lr': 0.0004889434419642259, 'samples': 944928, 'steps': 6561, 'loss/train': 1.6573374271392822}
+01/23/2022 08:26:13 - INFO - codeparrot_training - Step 39373: {'lr': 0.0004889386292053114, 'samples': 944952, 'steps': 6562, 'loss/train': 4.795544147491455}
+01/23/2022 08:26:14 - INFO - codeparrot_training - Step 39374: {'lr': 0.0004889386292053114, 'samples': 944976, 'steps': 6562, 'loss/train': 3.2198400497436523}
+01/23/2022 08:26:14 - INFO - codeparrot_training - Step 39375: {'lr': 0.0004889386292053114, 'samples': 945000, 'steps': 6562, 'loss/train': 3.9517509937286377}
+01/23/2022 08:26:15 - INFO - codeparrot_training - Step 39376: {'lr': 0.0004889386292053114, 'samples': 945024, 'steps': 6562, 'loss/train': 3.850275993347168}
+01/23/2022 08:26:15 - INFO - codeparrot_training - Step 39377: {'lr': 0.0004889386292053114, 'samples': 945048, 'steps': 6562, 'loss/train': 3.5318005084991455}
+01/23/2022 08:26:15 - INFO - codeparrot_training - Step 39378: {'lr': 0.0004889386292053114, 'samples': 945072, 'steps': 6562, 'loss/train': 1.775570034980774}
+01/23/2022 08:26:16 - INFO - codeparrot_training - Step 39379: {'lr': 0.0004889338154228596, 'samples': 945096, 'steps': 6563, 'loss/train': 3.1156766414642334}
+01/23/2022 08:26:16 - INFO - codeparrot_training - Step 39380: {'lr': 0.0004889338154228596, 'samples': 945120, 'steps': 6563, 'loss/train': 2.9608306884765625}
+01/23/2022 08:26:16 - INFO - codeparrot_training - Step 39381: {'lr': 0.0004889338154228596, 'samples': 945144, 'steps': 6563, 'loss/train': 2.8860559463500977}
+01/23/2022 08:26:17 - INFO - codeparrot_training - Step 39382: {'lr': 0.0004889338154228596, 'samples': 945168, 'steps': 6563, 'loss/train': 2.4836573600769043}
+01/23/2022 08:26:17 - INFO - codeparrot_training - Step 39383: {'lr': 0.0004889338154228596, 'samples': 945192, 'steps': 6563, 'loss/train': 2.8617701530456543}
+01/23/2022 08:26:18 - INFO - codeparrot_training - Step 39384: {'lr': 0.0004889338154228596, 'samples': 945216, 'steps': 6563, 'loss/train': 4.098076820373535}
+01/23/2022 08:26:18 - INFO - codeparrot_training - Step 39385: {'lr': 0.0004889290006168913, 'samples': 945240, 'steps': 6564, 'loss/train': 2.936919689178467}
+01/23/2022 08:26:18 - INFO - codeparrot_training - Step 39386: {'lr': 0.0004889290006168913, 'samples': 945264, 'steps': 6564, 'loss/train': 2.3886702060699463}
+01/23/2022 08:26:19 - INFO - codeparrot_training - Step 39387: {'lr': 0.0004889290006168913, 'samples': 945288, 'steps': 6564, 'loss/train': 3.1836681365966797}
+01/23/2022 08:26:19 - INFO - codeparrot_training - Step 39388: {'lr': 0.0004889290006168913, 'samples': 945312, 'steps': 6564, 'loss/train': 3.246121644973755}
+01/23/2022 08:26:20 - INFO - codeparrot_training - Step 39389: {'lr': 0.0004889290006168913, 'samples': 945336, 'steps': 6564, 'loss/train': 3.915313959121704}
+01/23/2022 08:26:20 - INFO - codeparrot_training - Step 39390: {'lr': 0.0004889290006168913, 'samples': 945360, 'steps': 6564, 'loss/train': 2.6131224632263184}
+01/23/2022 08:26:20 - INFO - codeparrot_training - Step 39391: {'lr': 0.0004889241847874271, 'samples': 945384, 'steps': 6565, 'loss/train': 2.8215718269348145}
+01/23/2022 08:26:21 - INFO - codeparrot_training - Step 39392: {'lr': 0.0004889241847874271, 'samples': 945408, 'steps': 6565, 'loss/train': 3.0232112407684326}
+01/23/2022 08:26:21 - INFO - codeparrot_training - Step 39393: {'lr': 0.0004889241847874271, 'samples': 945432, 'steps': 6565, 'loss/train': 2.9829771518707275}
+01/23/2022 08:26:21 - INFO - codeparrot_training - Step 39394: {'lr': 0.0004889241847874271, 'samples': 945456, 'steps': 6565, 'loss/train': 3.380221128463745}
+01/23/2022 08:26:22 - INFO - codeparrot_training - Step 39395: {'lr': 0.0004889241847874271, 'samples': 945480, 'steps': 6565, 'loss/train': 2.8248586654663086}
+01/23/2022 08:26:22 - INFO - codeparrot_training - Step 39396: {'lr': 0.0004889241847874271, 'samples': 945504, 'steps': 6565, 'loss/train': 3.5660386085510254}
+01/23/2022 08:26:23 - INFO - codeparrot_training - Step 39397: {'lr': 0.0004889193679344874, 'samples': 945528, 'steps': 6566, 'loss/train': 4.01016092300415}
+01/23/2022 08:26:23 - INFO - codeparrot_training - Step 39398: {'lr': 0.0004889193679344874, 'samples': 945552, 'steps': 6566, 'loss/train': 4.101414203643799}
+01/23/2022 08:26:23 - INFO - codeparrot_training - Step 39399: {'lr': 0.0004889193679344874, 'samples': 945576, 'steps': 6566, 'loss/train': 3.1728675365448}
+01/23/2022 08:26:24 - INFO - codeparrot_training - Step 39400: {'lr': 0.0004889193679344874, 'samples': 945600, 'steps': 6566, 'loss/train': 2.941969394683838}
+01/23/2022 08:26:24 - INFO - codeparrot_training - Step 39401: {'lr': 0.0004889193679344874, 'samples': 945624, 'steps': 6566, 'loss/train': 1.3400193452835083}
+01/23/2022 08:26:25 - INFO - codeparrot_training - Step 39402: {'lr': 0.0004889193679344874, 'samples': 945648, 'steps': 6566, 'loss/train': 3.9980154037475586}
+01/23/2022 08:26:25 - INFO - codeparrot_training - Step 39403: {'lr': 0.0004889145500580932, 'samples': 945672, 'steps': 6567, 'loss/train': 3.630131721496582}
+01/23/2022 08:26:25 - INFO - codeparrot_training - Step 39404: {'lr': 0.0004889145500580932, 'samples': 945696, 'steps': 6567, 'loss/train': 2.4155635833740234}
+01/23/2022 08:26:26 - INFO - codeparrot_training - Step 39405: {'lr': 0.0004889145500580932, 'samples': 945720, 'steps': 6567, 'loss/train': 3.4758572578430176}
+01/23/2022 08:26:26 - INFO - codeparrot_training - Step 39406: {'lr': 0.0004889145500580932, 'samples': 945744, 'steps': 6567, 'loss/train': 0.5920477509498596}
+01/23/2022 08:26:26 - INFO - codeparrot_training - Step 39407: {'lr': 0.0004889145500580932, 'samples': 945768, 'steps': 6567, 'loss/train': 3.157531499862671}
+01/23/2022 08:26:27 - INFO - codeparrot_training - Step 39408: {'lr': 0.0004889145500580932, 'samples': 945792, 'steps': 6567, 'loss/train': 3.5291035175323486}
+01/23/2022 08:26:27 - INFO - codeparrot_training - Step 39409: {'lr': 0.0004889097311582647, 'samples': 945816, 'steps': 6568, 'loss/train': 4.074158668518066}
+01/23/2022 08:26:28 - INFO - codeparrot_training - Step 39410: {'lr': 0.0004889097311582647, 'samples': 945840, 'steps': 6568, 'loss/train': 1.5899596214294434}
+01/23/2022 08:26:28 - INFO - codeparrot_training - Step 39411: {'lr': 0.0004889097311582647, 'samples': 945864, 'steps': 6568, 'loss/train': 2.871488571166992}
+01/23/2022 08:26:29 - INFO - codeparrot_training - Step 39412: {'lr': 0.0004889097311582647, 'samples': 945888, 'steps': 6568, 'loss/train': 4.163006782531738}
+01/23/2022 08:26:29 - INFO - codeparrot_training - Step 39413: {'lr': 0.0004889097311582647, 'samples': 945912, 'steps': 6568, 'loss/train': 2.762707471847534}
+01/23/2022 08:26:30 - INFO - codeparrot_training - Step 39414: {'lr': 0.0004889097311582647, 'samples': 945936, 'steps': 6568, 'loss/train': 2.629348039627075}
+01/23/2022 08:26:30 - INFO - codeparrot_training - Step 39415: {'lr': 0.000488904911235023, 'samples': 945960, 'steps': 6569, 'loss/train': 2.121138095855713}
+01/23/2022 08:26:31 - INFO - codeparrot_training - Step 39416: {'lr': 0.000488904911235023, 'samples': 945984, 'steps': 6569, 'loss/train': 3.0221242904663086}
+01/23/2022 08:26:31 - INFO - codeparrot_training - Step 39417: {'lr': 0.000488904911235023, 'samples': 946008, 'steps': 6569, 'loss/train': 2.9159984588623047}
+01/23/2022 08:26:31 - INFO - codeparrot_training - Step 39418: {'lr': 0.000488904911235023, 'samples': 946032, 'steps': 6569, 'loss/train': 4.092543125152588}
+01/23/2022 08:26:32 - INFO - codeparrot_training - Step 39419: {'lr': 0.000488904911235023, 'samples': 946056, 'steps': 6569, 'loss/train': 3.982084035873413}
+01/23/2022 08:26:32 - INFO - codeparrot_training - Step 39420: {'lr': 0.000488904911235023, 'samples': 946080, 'steps': 6569, 'loss/train': 3.987621545791626}
+01/23/2022 08:26:33 - INFO - codeparrot_training - Step 39421: {'lr': 0.0004889000902883883, 'samples': 946104, 'steps': 6570, 'loss/train': 3.603787660598755}
+01/23/2022 08:26:33 - INFO - codeparrot_training - Step 39422: {'lr': 0.0004889000902883883, 'samples': 946128, 'steps': 6570, 'loss/train': 2.983677387237549}
+01/23/2022 08:26:33 - INFO - codeparrot_training - Step 39423: {'lr': 0.0004889000902883883, 'samples': 946152, 'steps': 6570, 'loss/train': 5.418353080749512}
+01/23/2022 08:26:34 - INFO - codeparrot_training - Step 39424: {'lr': 0.0004889000902883883, 'samples': 946176, 'steps': 6570, 'loss/train': 3.9826366901397705}
+01/23/2022 08:26:34 - INFO - codeparrot_training - Step 39425: {'lr': 0.0004889000902883883, 'samples': 946200, 'steps': 6570, 'loss/train': 4.009023189544678}
+01/23/2022 08:26:34 - INFO - codeparrot_training - Step 39426: {'lr': 0.0004889000902883883, 'samples': 946224, 'steps': 6570, 'loss/train': 4.667829990386963}
+01/23/2022 08:26:35 - INFO - codeparrot_training - Step 39427: {'lr': 0.0004888952683183816, 'samples': 946248, 'steps': 6571, 'loss/train': 4.441924095153809}
+01/23/2022 08:26:35 - INFO - codeparrot_training - Step 39428: {'lr': 0.0004888952683183816, 'samples': 946272, 'steps': 6571, 'loss/train': 1.6560255289077759}
+01/23/2022 08:26:36 - INFO - codeparrot_training - Step 39429: {'lr': 0.0004888952683183816, 'samples': 946296, 'steps': 6571, 'loss/train': 3.631377696990967}
+01/23/2022 08:26:36 - INFO - codeparrot_training - Step 39430: {'lr': 0.0004888952683183816, 'samples': 946320, 'steps': 6571, 'loss/train': 4.814955711364746}
+01/23/2022 08:26:36 - INFO - codeparrot_training - Step 39431: {'lr': 0.0004888952683183816, 'samples': 946344, 'steps': 6571, 'loss/train': 4.015174388885498}
+01/23/2022 08:26:37 - INFO - codeparrot_training - Step 39432: {'lr': 0.0004888952683183816, 'samples': 946368, 'steps': 6571, 'loss/train': 4.306601047515869}
+01/23/2022 08:26:37 - INFO - codeparrot_training - Step 39433: {'lr': 0.0004888904453250233, 'samples': 946392, 'steps': 6572, 'loss/train': 4.208254814147949}
+01/23/2022 08:26:38 - INFO - codeparrot_training - Step 39434: {'lr': 0.0004888904453250233, 'samples': 946416, 'steps': 6572, 'loss/train': 3.5904152393341064}
+01/23/2022 08:26:38 - INFO - codeparrot_training - Step 39435: {'lr': 0.0004888904453250233, 'samples': 946440, 'steps': 6572, 'loss/train': 2.894926071166992}
+01/23/2022 08:26:38 - INFO - codeparrot_training - Step 39436: {'lr': 0.0004888904453250233, 'samples': 946464, 'steps': 6572, 'loss/train': 1.61025869846344}
+01/23/2022 08:26:39 - INFO - codeparrot_training - Step 39437: {'lr': 0.0004888904453250233, 'samples': 946488, 'steps': 6572, 'loss/train': 3.2669458389282227}
+01/23/2022 08:26:39 - INFO - codeparrot_training - Step 39438: {'lr': 0.0004888904453250233, 'samples': 946512, 'steps': 6572, 'loss/train': 3.8875951766967773}
+01/23/2022 08:26:39 - INFO - codeparrot_training - Step 39439: {'lr': 0.0004888856213083343, 'samples': 946536, 'steps': 6573, 'loss/train': 3.4393815994262695}
+01/23/2022 08:26:40 - INFO - codeparrot_training - Step 39440: {'lr': 0.0004888856213083343, 'samples': 946560, 'steps': 6573, 'loss/train': 3.1709976196289062}
+01/23/2022 08:26:40 - INFO - codeparrot_training - Step 39441: {'lr': 0.0004888856213083343, 'samples': 946584, 'steps': 6573, 'loss/train': 2.2305150032043457}
+01/23/2022 08:26:41 - INFO - codeparrot_training - Step 39442: {'lr': 0.0004888856213083343, 'samples': 946608, 'steps': 6573, 'loss/train': 3.305680274963379}
+01/23/2022 08:26:41 - INFO - codeparrot_training - Step 39443: {'lr': 0.0004888856213083343, 'samples': 946632, 'steps': 6573, 'loss/train': 3.6540582180023193}
+01/23/2022 08:26:41 - INFO - codeparrot_training - Step 39444: {'lr': 0.0004888856213083343, 'samples': 946656, 'steps': 6573, 'loss/train': 2.7756175994873047}
+01/23/2022 08:26:42 - INFO - codeparrot_training - Step 39445: {'lr': 0.0004888807962683353, 'samples': 946680, 'steps': 6574, 'loss/train': 2.081913948059082}
+01/23/2022 08:26:42 - INFO - codeparrot_training - Step 39446: {'lr': 0.0004888807962683353, 'samples': 946704, 'steps': 6574, 'loss/train': 2.3533499240875244}
+01/23/2022 08:26:43 - INFO - codeparrot_training - Step 39447: {'lr': 0.0004888807962683353, 'samples': 946728, 'steps': 6574, 'loss/train': 3.908698320388794}
+01/23/2022 08:26:43 - INFO - codeparrot_training - Step 39448: {'lr': 0.0004888807962683353, 'samples': 946752, 'steps': 6574, 'loss/train': 2.52559494972229}
+01/23/2022 08:26:43 - INFO - codeparrot_training - Step 39449: {'lr': 0.0004888807962683353, 'samples': 946776, 'steps': 6574, 'loss/train': 2.8395462036132812}
+01/23/2022 08:26:44 - INFO - codeparrot_training - Step 39450: {'lr': 0.0004888807962683353, 'samples': 946800, 'steps': 6574, 'loss/train': 2.554753541946411}
+01/23/2022 08:26:44 - INFO - codeparrot_training - Step 39451: {'lr': 0.0004888759702050466, 'samples': 946824, 'steps': 6575, 'loss/train': 2.997220754623413}
+01/23/2022 08:26:44 - INFO - codeparrot_training - Step 39452: {'lr': 0.0004888759702050466, 'samples': 946848, 'steps': 6575, 'loss/train': 1.526603102684021}
+01/23/2022 08:26:45 - INFO - codeparrot_training - Step 39453: {'lr': 0.0004888759702050466, 'samples': 946872, 'steps': 6575, 'loss/train': 3.4559168815612793}
+01/23/2022 08:26:48 - INFO - codeparrot_training - Step 39454: {'lr': 0.0004888759702050466, 'samples': 946896, 'steps': 6575, 'loss/train': 2.635401725769043}
+01/23/2022 08:26:48 - INFO - codeparrot_training - Step 39455: {'lr': 0.0004888759702050466, 'samples': 946920, 'steps': 6575, 'loss/train': 2.951209783554077}
+01/23/2022 08:26:48 - INFO - codeparrot_training - Step 39456: {'lr': 0.0004888759702050466, 'samples': 946944, 'steps': 6575, 'loss/train': 3.6766891479492188}
+01/23/2022 08:26:49 - INFO - codeparrot_training - Step 39457: {'lr': 0.0004888711431184892, 'samples': 946968, 'steps': 6576, 'loss/train': 3.331228733062744}
+01/23/2022 08:26:49 - INFO - codeparrot_training - Step 39458: {'lr': 0.0004888711431184892, 'samples': 946992, 'steps': 6576, 'loss/train': 3.7854630947113037}
+01/23/2022 08:26:50 - INFO - codeparrot_training - Step 39459: {'lr': 0.0004888711431184892, 'samples': 947016, 'steps': 6576, 'loss/train': 3.579988956451416}
+01/23/2022 08:26:50 - INFO - codeparrot_training - Step 39460: {'lr': 0.0004888711431184892, 'samples': 947040, 'steps': 6576, 'loss/train': 3.1991350650787354}
+01/23/2022 08:26:50 - INFO - codeparrot_training - Step 39461: {'lr': 0.0004888711431184892, 'samples': 947064, 'steps': 6576, 'loss/train': 4.219009876251221}
+01/23/2022 08:26:51 - INFO - codeparrot_training - Step 39462: {'lr': 0.0004888711431184892, 'samples': 947088, 'steps': 6576, 'loss/train': 2.6763229370117188}
+01/23/2022 08:26:51 - INFO - codeparrot_training - Step 39463: {'lr': 0.0004888663150086835, 'samples': 947112, 'steps': 6577, 'loss/train': 3.3051583766937256}
+01/23/2022 08:26:51 - INFO - codeparrot_training - Step 39464: {'lr': 0.0004888663150086835, 'samples': 947136, 'steps': 6577, 'loss/train': 3.855036973953247}
+01/23/2022 08:26:52 - INFO - codeparrot_training - Step 39465: {'lr': 0.0004888663150086835, 'samples': 947160, 'steps': 6577, 'loss/train': 2.924755573272705}
+01/23/2022 08:26:52 - INFO - codeparrot_training - Step 39466: {'lr': 0.0004888663150086835, 'samples': 947184, 'steps': 6577, 'loss/train': 2.1921422481536865}
+01/23/2022 08:26:53 - INFO - codeparrot_training - Step 39467: {'lr': 0.0004888663150086835, 'samples': 947208, 'steps': 6577, 'loss/train': 3.122788190841675}
+01/23/2022 08:26:53 - INFO - codeparrot_training - Step 39468: {'lr': 0.0004888663150086835, 'samples': 947232, 'steps': 6577, 'loss/train': 3.123659133911133}
+01/23/2022 08:26:53 - INFO - codeparrot_training - Step 39469: {'lr': 0.0004888614858756505, 'samples': 947256, 'steps': 6578, 'loss/train': 3.106525182723999}
+01/23/2022 08:26:54 - INFO - codeparrot_training - Step 39470: {'lr': 0.0004888614858756505, 'samples': 947280, 'steps': 6578, 'loss/train': 2.4451096057891846}
+01/23/2022 08:26:54 - INFO - codeparrot_training - Step 39471: {'lr': 0.0004888614858756505, 'samples': 947304, 'steps': 6578, 'loss/train': 3.2773494720458984}
+01/23/2022 08:26:55 - INFO - codeparrot_training - Step 39472: {'lr': 0.0004888614858756505, 'samples': 947328, 'steps': 6578, 'loss/train': 3.8774282932281494}
+01/23/2022 08:26:55 - INFO - codeparrot_training - Step 39473: {'lr': 0.0004888614858756505, 'samples': 947352, 'steps': 6578, 'loss/train': 3.5242221355438232}
+01/23/2022 08:26:55 - INFO - codeparrot_training - Step 39474: {'lr': 0.0004888614858756505, 'samples': 947376, 'steps': 6578, 'loss/train': 3.6783182621002197}
+01/23/2022 08:26:56 - INFO - codeparrot_training - Step 39475: {'lr': 0.0004888566557194107, 'samples': 947400, 'steps': 6579, 'loss/train': 3.0560414791107178}
+01/23/2022 08:26:56 - INFO - codeparrot_training - Step 39476: {'lr': 0.0004888566557194107, 'samples': 947424, 'steps': 6579, 'loss/train': 3.1156766414642334}
+01/23/2022 08:26:56 - INFO - codeparrot_training - Step 39477: {'lr': 0.0004888566557194107, 'samples': 947448, 'steps': 6579, 'loss/train': 3.3071682453155518}
+01/23/2022 08:26:57 - INFO - codeparrot_training - Step 39478: {'lr': 0.0004888566557194107, 'samples': 947472, 'steps': 6579, 'loss/train': 2.8109488487243652}
+01/23/2022 08:26:57 - INFO - codeparrot_training - Step 39479: {'lr': 0.0004888566557194107, 'samples': 947496, 'steps': 6579, 'loss/train': 3.8240113258361816}
+01/23/2022 08:26:58 - INFO - codeparrot_training - Step 39480: {'lr': 0.0004888566557194107, 'samples': 947520, 'steps': 6579, 'loss/train': 4.122755527496338}
+01/23/2022 08:26:58 - INFO - codeparrot_training - Step 39481: {'lr': 0.0004888518245399849, 'samples': 947544, 'steps': 6580, 'loss/train': 3.0159530639648438}
+01/23/2022 08:26:58 - INFO - codeparrot_training - Step 39482: {'lr': 0.0004888518245399849, 'samples': 947568, 'steps': 6580, 'loss/train': 3.445997953414917}
+01/23/2022 08:26:59 - INFO - codeparrot_training - Step 39483: {'lr': 0.0004888518245399849, 'samples': 947592, 'steps': 6580, 'loss/train': 3.301321029663086}
+01/23/2022 08:26:59 - INFO - codeparrot_training - Step 39484: {'lr': 0.0004888518245399849, 'samples': 947616, 'steps': 6580, 'loss/train': 2.252297878265381}
+01/23/2022 08:26:59 - INFO - codeparrot_training - Step 39485: {'lr': 0.0004888518245399849, 'samples': 947640, 'steps': 6580, 'loss/train': 3.032694101333618}
+01/23/2022 08:27:00 - INFO - codeparrot_training - Step 39486: {'lr': 0.0004888518245399849, 'samples': 947664, 'steps': 6580, 'loss/train': 2.1533164978027344}
+01/23/2022 08:27:00 - INFO - codeparrot_training - Step 39487: {'lr': 0.0004888469923373937, 'samples': 947688, 'steps': 6581, 'loss/train': 2.6769449710845947}
+01/23/2022 08:27:01 - INFO - codeparrot_training - Step 39488: {'lr': 0.0004888469923373937, 'samples': 947712, 'steps': 6581, 'loss/train': 3.1941258907318115}
+01/23/2022 08:27:01 - INFO - codeparrot_training - Step 39489: {'lr': 0.0004888469923373937, 'samples': 947736, 'steps': 6581, 'loss/train': 2.118013620376587}
+01/23/2022 08:27:01 - INFO - codeparrot_training - Step 39490: {'lr': 0.0004888469923373937, 'samples': 947760, 'steps': 6581, 'loss/train': 3.5010814666748047}
+01/23/2022 08:27:02 - INFO - codeparrot_training - Step 39491: {'lr': 0.0004888469923373937, 'samples': 947784, 'steps': 6581, 'loss/train': 2.385258436203003}
+01/23/2022 08:27:02 - INFO - codeparrot_training - Step 39492: {'lr': 0.0004888469923373937, 'samples': 947808, 'steps': 6581, 'loss/train': 3.8076953887939453}
+01/23/2022 08:27:03 - INFO - codeparrot_training - Step 39493: {'lr': 0.0004888421591116578, 'samples': 947832, 'steps': 6582, 'loss/train': 2.615622043609619}
+01/23/2022 08:27:03 - INFO - codeparrot_training - Step 39494: {'lr': 0.0004888421591116578, 'samples': 947856, 'steps': 6582, 'loss/train': 3.009289026260376}
+01/23/2022 08:27:04 - INFO - codeparrot_training - Step 39495: {'lr': 0.0004888421591116578, 'samples': 947880, 'steps': 6582, 'loss/train': 2.678691864013672}
+01/23/2022 08:27:04 - INFO - codeparrot_training - Step 39496: {'lr': 0.0004888421591116578, 'samples': 947904, 'steps': 6582, 'loss/train': 2.0367095470428467}
+01/23/2022 08:27:05 - INFO - codeparrot_training - Step 39497: {'lr': 0.0004888421591116578, 'samples': 947928, 'steps': 6582, 'loss/train': 3.0955045223236084}
+01/23/2022 08:27:05 - INFO - codeparrot_training - Step 39498: {'lr': 0.0004888421591116578, 'samples': 947952, 'steps': 6582, 'loss/train': 3.5932469367980957}
+01/23/2022 08:27:06 - INFO - codeparrot_training - Step 39499: {'lr': 0.000488837324862798, 'samples': 947976, 'steps': 6583, 'loss/train': 3.2359869480133057}
+01/23/2022 08:27:06 - INFO - codeparrot_training - Step 39500: {'lr': 0.000488837324862798, 'samples': 948000, 'steps': 6583, 'loss/train': 2.882493734359741}
+01/23/2022 08:27:06 - INFO - codeparrot_training - Step 39501: {'lr': 0.000488837324862798, 'samples': 948024, 'steps': 6583, 'loss/train': 3.7544291019439697}
+01/23/2022 08:27:07 - INFO - codeparrot_training - Step 39502: {'lr': 0.000488837324862798, 'samples': 948048, 'steps': 6583, 'loss/train': 4.139290809631348}
+01/23/2022 08:27:07 - INFO - codeparrot_training - Step 39503: {'lr': 0.000488837324862798, 'samples': 948072, 'steps': 6583, 'loss/train': 2.8595194816589355}
+01/23/2022 08:27:07 - INFO - codeparrot_training - Step 39504: {'lr': 0.000488837324862798, 'samples': 948096, 'steps': 6583, 'loss/train': 4.03502893447876}
+01/23/2022 08:27:08 - INFO - codeparrot_training - Step 39505: {'lr': 0.0004888324895908349, 'samples': 948120, 'steps': 6584, 'loss/train': 3.0091164112091064}
+01/23/2022 08:27:08 - INFO - codeparrot_training - Step 39506: {'lr': 0.0004888324895908349, 'samples': 948144, 'steps': 6584, 'loss/train': 0.4613305330276489}
+01/23/2022 08:27:09 - INFO - codeparrot_training - Step 39507: {'lr': 0.0004888324895908349, 'samples': 948168, 'steps': 6584, 'loss/train': 3.018129348754883}
+01/23/2022 08:27:09 - INFO - codeparrot_training - Step 39508: {'lr': 0.0004888324895908349, 'samples': 948192, 'steps': 6584, 'loss/train': 3.622781753540039}
+01/23/2022 08:27:09 - INFO - codeparrot_training - Step 39509: {'lr': 0.0004888324895908349, 'samples': 948216, 'steps': 6584, 'loss/train': 3.2388386726379395}
+01/23/2022 08:27:10 - INFO - codeparrot_training - Step 39510: {'lr': 0.0004888324895908349, 'samples': 948240, 'steps': 6584, 'loss/train': 3.246419906616211}
+01/23/2022 08:27:10 - INFO - codeparrot_training - Step 39511: {'lr': 0.0004888276532957892, 'samples': 948264, 'steps': 6585, 'loss/train': 3.5823495388031006}
+01/23/2022 08:27:10 - INFO - codeparrot_training - Step 39512: {'lr': 0.0004888276532957892, 'samples': 948288, 'steps': 6585, 'loss/train': 4.338104248046875}
+01/23/2022 08:27:11 - INFO - codeparrot_training - Step 39513: {'lr': 0.0004888276532957892, 'samples': 948312, 'steps': 6585, 'loss/train': 2.5861270427703857}
+01/23/2022 08:27:11 - INFO - codeparrot_training - Step 39514: {'lr': 0.0004888276532957892, 'samples': 948336, 'steps': 6585, 'loss/train': 2.445862293243408}
+01/23/2022 08:27:12 - INFO - codeparrot_training - Step 39515: {'lr': 0.0004888276532957892, 'samples': 948360, 'steps': 6585, 'loss/train': 4.0128302574157715}
+01/23/2022 08:27:12 - INFO - codeparrot_training - Step 39516: {'lr': 0.0004888276532957892, 'samples': 948384, 'steps': 6585, 'loss/train': 2.8402209281921387}
+01/23/2022 08:27:12 - INFO - codeparrot_training - Step 39517: {'lr': 0.0004888228159776818, 'samples': 948408, 'steps': 6586, 'loss/train': 3.095761299133301}
+01/23/2022 08:27:13 - INFO - codeparrot_training - Step 39518: {'lr': 0.0004888228159776818, 'samples': 948432, 'steps': 6586, 'loss/train': 3.070746421813965}
+01/23/2022 08:27:13 - INFO - codeparrot_training - Step 39519: {'lr': 0.0004888228159776818, 'samples': 948456, 'steps': 6586, 'loss/train': 2.5289456844329834}
+01/23/2022 08:27:13 - INFO - codeparrot_training - Step 39520: {'lr': 0.0004888228159776818, 'samples': 948480, 'steps': 6586, 'loss/train': 3.052000045776367}
+01/23/2022 08:27:14 - INFO - codeparrot_training - Step 39521: {'lr': 0.0004888228159776818, 'samples': 948504, 'steps': 6586, 'loss/train': 2.446962833404541}
+01/23/2022 08:27:14 - INFO - codeparrot_training - Step 39522: {'lr': 0.0004888228159776818, 'samples': 948528, 'steps': 6586, 'loss/train': 3.5778989791870117}
+01/23/2022 08:27:15 - INFO - codeparrot_training - Step 39523: {'lr': 0.0004888179776365331, 'samples': 948552, 'steps': 6587, 'loss/train': 4.396428108215332}
+01/23/2022 08:27:15 - INFO - codeparrot_training - Step 39524: {'lr': 0.0004888179776365331, 'samples': 948576, 'steps': 6587, 'loss/train': 2.6337249279022217}
+01/23/2022 08:27:15 - INFO - codeparrot_training - Step 39525: {'lr': 0.0004888179776365331, 'samples': 948600, 'steps': 6587, 'loss/train': 4.615782260894775}
+01/23/2022 08:27:16 - INFO - codeparrot_training - Step 39526: {'lr': 0.0004888179776365331, 'samples': 948624, 'steps': 6587, 'loss/train': 2.1522090435028076}
+01/23/2022 08:27:16 - INFO - codeparrot_training - Step 39527: {'lr': 0.0004888179776365331, 'samples': 948648, 'steps': 6587, 'loss/train': 2.75506854057312}
+01/23/2022 08:27:17 - INFO - codeparrot_training - Step 39528: {'lr': 0.0004888179776365331, 'samples': 948672, 'steps': 6587, 'loss/train': 2.2520759105682373}
+01/23/2022 08:27:17 - INFO - codeparrot_training - Step 39529: {'lr': 0.0004888131382723641, 'samples': 948696, 'steps': 6588, 'loss/train': 3.4085352420806885}
+01/23/2022 08:27:17 - INFO - codeparrot_training - Step 39530: {'lr': 0.0004888131382723641, 'samples': 948720, 'steps': 6588, 'loss/train': 2.6323986053466797}
+01/23/2022 08:27:18 - INFO - codeparrot_training - Step 39531: {'lr': 0.0004888131382723641, 'samples': 948744, 'steps': 6588, 'loss/train': 3.1926157474517822}
+01/23/2022 08:27:18 - INFO - codeparrot_training - Step 39532: {'lr': 0.0004888131382723641, 'samples': 948768, 'steps': 6588, 'loss/train': 3.2555112838745117}
+01/23/2022 08:27:18 - INFO - codeparrot_training - Step 39533: {'lr': 0.0004888131382723641, 'samples': 948792, 'steps': 6588, 'loss/train': 2.162386655807495}
+01/23/2022 08:27:20 - INFO - codeparrot_training - Step 39534: {'lr': 0.0004888131382723641, 'samples': 948816, 'steps': 6588, 'loss/train': 1.9532700777053833}
+01/23/2022 08:27:20 - INFO - codeparrot_training - Step 39535: {'lr': 0.0004888082978851954, 'samples': 948840, 'steps': 6589, 'loss/train': 3.9461965560913086}
+01/23/2022 08:27:20 - INFO - codeparrot_training - Step 39536: {'lr': 0.0004888082978851954, 'samples': 948864, 'steps': 6589, 'loss/train': 2.343686580657959}
+01/23/2022 08:27:21 - INFO - codeparrot_training - Step 39537: {'lr': 0.0004888082978851954, 'samples': 948888, 'steps': 6589, 'loss/train': 2.345708131790161}
+01/23/2022 08:27:21 - INFO - codeparrot_training - Step 39538: {'lr': 0.0004888082978851954, 'samples': 948912, 'steps': 6589, 'loss/train': 2.7945950031280518}
+01/23/2022 08:27:21 - INFO - codeparrot_training - Step 39539: {'lr': 0.0004888082978851954, 'samples': 948936, 'steps': 6589, 'loss/train': 3.3515403270721436}
+01/23/2022 08:27:22 - INFO - codeparrot_training - Step 39540: {'lr': 0.0004888082978851954, 'samples': 948960, 'steps': 6589, 'loss/train': 3.8459177017211914}
+01/23/2022 08:27:22 - INFO - codeparrot_training - Step 39541: {'lr': 0.000488803456475048, 'samples': 948984, 'steps': 6590, 'loss/train': 3.6526167392730713}
+01/23/2022 08:27:23 - INFO - codeparrot_training - Step 39542: {'lr': 0.000488803456475048, 'samples': 949008, 'steps': 6590, 'loss/train': 3.030672788619995}
+01/23/2022 08:27:23 - INFO - codeparrot_training - Step 39543: {'lr': 0.000488803456475048, 'samples': 949032, 'steps': 6590, 'loss/train': 3.5282206535339355}
+01/23/2022 08:27:23 - INFO - codeparrot_training - Step 39544: {'lr': 0.000488803456475048, 'samples': 949056, 'steps': 6590, 'loss/train': 3.1778221130371094}
+01/23/2022 08:27:24 - INFO - codeparrot_training - Step 39545: {'lr': 0.000488803456475048, 'samples': 949080, 'steps': 6590, 'loss/train': 2.574775218963623}
+01/23/2022 08:27:24 - INFO - codeparrot_training - Step 39546: {'lr': 0.000488803456475048, 'samples': 949104, 'steps': 6590, 'loss/train': 4.209159851074219}
+01/23/2022 08:27:25 - INFO - codeparrot_training - Step 39547: {'lr': 0.0004887986140419422, 'samples': 949128, 'steps': 6591, 'loss/train': 3.0977847576141357}
+01/23/2022 08:27:25 - INFO - codeparrot_training - Step 39548: {'lr': 0.0004887986140419422, 'samples': 949152, 'steps': 6591, 'loss/train': 2.8869917392730713}
+01/23/2022 08:27:25 - INFO - codeparrot_training - Step 39549: {'lr': 0.0004887986140419422, 'samples': 949176, 'steps': 6591, 'loss/train': 3.860635757446289}
+01/23/2022 08:27:26 - INFO - codeparrot_training - Step 39550: {'lr': 0.0004887986140419422, 'samples': 949200, 'steps': 6591, 'loss/train': 4.374944686889648}
+01/23/2022 08:27:26 - INFO - codeparrot_training - Step 39551: {'lr': 0.0004887986140419422, 'samples': 949224, 'steps': 6591, 'loss/train': 3.020859479904175}
+01/23/2022 08:27:26 - INFO - codeparrot_training - Step 39552: {'lr': 0.0004887986140419422, 'samples': 949248, 'steps': 6591, 'loss/train': 1.7071356773376465}
+01/23/2022 08:27:27 - INFO - codeparrot_training - Step 39553: {'lr': 0.000488793770585899, 'samples': 949272, 'steps': 6592, 'loss/train': 3.0089516639709473}
+01/23/2022 08:27:27 - INFO - codeparrot_training - Step 39554: {'lr': 0.000488793770585899, 'samples': 949296, 'steps': 6592, 'loss/train': 2.8441967964172363}
+01/23/2022 08:27:28 - INFO - codeparrot_training - Step 39555: {'lr': 0.000488793770585899, 'samples': 949320, 'steps': 6592, 'loss/train': 2.9921183586120605}
+01/23/2022 08:27:28 - INFO - codeparrot_training - Step 39556: {'lr': 0.000488793770585899, 'samples': 949344, 'steps': 6592, 'loss/train': 3.0547091960906982}
+01/23/2022 08:27:28 - INFO - codeparrot_training - Step 39557: {'lr': 0.000488793770585899, 'samples': 949368, 'steps': 6592, 'loss/train': 3.3019626140594482}
+01/23/2022 08:27:29 - INFO - codeparrot_training - Step 39558: {'lr': 0.000488793770585899, 'samples': 949392, 'steps': 6592, 'loss/train': 2.713526964187622}
+01/23/2022 08:27:29 - INFO - codeparrot_training - Step 39559: {'lr': 0.0004887889261069392, 'samples': 949416, 'steps': 6593, 'loss/train': 2.5103847980499268}
+01/23/2022 08:27:30 - INFO - codeparrot_training - Step 39560: {'lr': 0.0004887889261069392, 'samples': 949440, 'steps': 6593, 'loss/train': 3.367967128753662}
+01/23/2022 08:27:30 - INFO - codeparrot_training - Step 39561: {'lr': 0.0004887889261069392, 'samples': 949464, 'steps': 6593, 'loss/train': 3.162057399749756}
+01/23/2022 08:27:30 - INFO - codeparrot_training - Step 39562: {'lr': 0.0004887889261069392, 'samples': 949488, 'steps': 6593, 'loss/train': 2.2834677696228027}
+01/23/2022 08:27:31 - INFO - codeparrot_training - Step 39563: {'lr': 0.0004887889261069392, 'samples': 949512, 'steps': 6593, 'loss/train': 3.1009654998779297}
+01/23/2022 08:27:31 - INFO - codeparrot_training - Step 39564: {'lr': 0.0004887889261069392, 'samples': 949536, 'steps': 6593, 'loss/train': 3.4486618041992188}
+01/23/2022 08:27:31 - INFO - codeparrot_training - Step 39565: {'lr': 0.0004887840806050834, 'samples': 949560, 'steps': 6594, 'loss/train': 2.636906385421753}
+01/23/2022 08:27:32 - INFO - codeparrot_training - Step 39566: {'lr': 0.0004887840806050834, 'samples': 949584, 'steps': 6594, 'loss/train': 1.9629610776901245}
+01/23/2022 08:27:32 - INFO - codeparrot_training - Step 39567: {'lr': 0.0004887840806050834, 'samples': 949608, 'steps': 6594, 'loss/train': 3.432258367538452}
+01/23/2022 08:27:33 - INFO - codeparrot_training - Step 39568: {'lr': 0.0004887840806050834, 'samples': 949632, 'steps': 6594, 'loss/train': 2.741351842880249}
+01/23/2022 08:27:33 - INFO - codeparrot_training - Step 39569: {'lr': 0.0004887840806050834, 'samples': 949656, 'steps': 6594, 'loss/train': 4.375844478607178}
+01/23/2022 08:27:33 - INFO - codeparrot_training - Step 39570: {'lr': 0.0004887840806050834, 'samples': 949680, 'steps': 6594, 'loss/train': 3.0668234825134277}
+01/23/2022 08:27:34 - INFO - codeparrot_training - Step 39571: {'lr': 0.0004887792340803524, 'samples': 949704, 'steps': 6595, 'loss/train': 2.4352431297302246}
+01/23/2022 08:27:34 - INFO - codeparrot_training - Step 39572: {'lr': 0.0004887792340803524, 'samples': 949728, 'steps': 6595, 'loss/train': 3.7245640754699707}
+01/23/2022 08:27:34 - INFO - codeparrot_training - Step 39573: {'lr': 0.0004887792340803524, 'samples': 949752, 'steps': 6595, 'loss/train': 2.748079299926758}
+01/23/2022 08:27:35 - INFO - codeparrot_training - Step 39574: {'lr': 0.0004887792340803524, 'samples': 949776, 'steps': 6595, 'loss/train': 2.8202075958251953}
+01/23/2022 08:27:36 - INFO - codeparrot_training - Step 39575: {'lr': 0.0004887792340803524, 'samples': 949800, 'steps': 6595, 'loss/train': 3.33016300201416}
+01/23/2022 08:27:36 - INFO - codeparrot_training - Step 39576: {'lr': 0.0004887792340803524, 'samples': 949824, 'steps': 6595, 'loss/train': 3.152240753173828}
+01/23/2022 08:27:37 - INFO - codeparrot_training - Step 39577: {'lr': 0.000488774386532767, 'samples': 949848, 'steps': 6596, 'loss/train': 3.7984678745269775}
+01/23/2022 08:27:37 - INFO - codeparrot_training - Step 39578: {'lr': 0.000488774386532767, 'samples': 949872, 'steps': 6596, 'loss/train': 3.845282793045044}
+01/23/2022 08:27:37 - INFO - codeparrot_training - Step 39579: {'lr': 0.000488774386532767, 'samples': 949896, 'steps': 6596, 'loss/train': 2.7544379234313965}
+01/23/2022 08:27:38 - INFO - codeparrot_training - Step 39580: {'lr': 0.000488774386532767, 'samples': 949920, 'steps': 6596, 'loss/train': 2.8876826763153076}
+01/23/2022 08:27:38 - INFO - codeparrot_training - Step 39581: {'lr': 0.000488774386532767, 'samples': 949944, 'steps': 6596, 'loss/train': 3.8829145431518555}
+01/23/2022 08:27:39 - INFO - codeparrot_training - Step 39582: {'lr': 0.000488774386532767, 'samples': 949968, 'steps': 6596, 'loss/train': 2.9708425998687744}
+01/23/2022 08:27:39 - INFO - codeparrot_training - Step 39583: {'lr': 0.0004887695379623481, 'samples': 949992, 'steps': 6597, 'loss/train': 1.3491309881210327}
+01/23/2022 08:27:39 - INFO - codeparrot_training - Step 39584: {'lr': 0.0004887695379623481, 'samples': 950016, 'steps': 6597, 'loss/train': 3.304015874862671}
+01/23/2022 08:27:40 - INFO - codeparrot_training - Step 39585: {'lr': 0.0004887695379623481, 'samples': 950040, 'steps': 6597, 'loss/train': 1.0561325550079346}
+01/23/2022 08:27:40 - INFO - codeparrot_training - Step 39586: {'lr': 0.0004887695379623481, 'samples': 950064, 'steps': 6597, 'loss/train': 0.8092907667160034}
+01/23/2022 08:27:41 - INFO - codeparrot_training - Step 39587: {'lr': 0.0004887695379623481, 'samples': 950088, 'steps': 6597, 'loss/train': 0.8273599743843079}
+01/23/2022 08:27:41 - INFO - codeparrot_training - Step 39588: {'lr': 0.0004887695379623481, 'samples': 950112, 'steps': 6597, 'loss/train': 2.787769079208374}
+01/23/2022 08:27:41 - INFO - codeparrot_training - Step 39589: {'lr': 0.000488764688369116, 'samples': 950136, 'steps': 6598, 'loss/train': 4.067763805389404}
+01/23/2022 08:27:42 - INFO - codeparrot_training - Step 39590: {'lr': 0.000488764688369116, 'samples': 950160, 'steps': 6598, 'loss/train': 3.73348331451416}
+01/23/2022 08:27:42 - INFO - codeparrot_training - Step 39591: {'lr': 0.000488764688369116, 'samples': 950184, 'steps': 6598, 'loss/train': 3.336317539215088}
+01/23/2022 08:27:42 - INFO - codeparrot_training - Step 39592: {'lr': 0.000488764688369116, 'samples': 950208, 'steps': 6598, 'loss/train': 3.6403794288635254}
+01/23/2022 08:27:43 - INFO - codeparrot_training - Step 39593: {'lr': 0.000488764688369116, 'samples': 950232, 'steps': 6598, 'loss/train': 2.5769667625427246}
+01/23/2022 08:27:43 - INFO - codeparrot_training - Step 39594: {'lr': 0.000488764688369116, 'samples': 950256, 'steps': 6598, 'loss/train': 3.4257688522338867}
+01/23/2022 08:27:44 - INFO - codeparrot_training - Step 39595: {'lr': 0.000488759837753092, 'samples': 950280, 'steps': 6599, 'loss/train': 3.5614969730377197}
+01/23/2022 08:27:44 - INFO - codeparrot_training - Step 39596: {'lr': 0.000488759837753092, 'samples': 950304, 'steps': 6599, 'loss/train': 3.202650785446167}
+01/23/2022 08:27:44 - INFO - codeparrot_training - Step 39597: {'lr': 0.000488759837753092, 'samples': 950328, 'steps': 6599, 'loss/train': 3.2686269283294678}
+01/23/2022 08:27:45 - INFO - codeparrot_training - Step 39598: {'lr': 0.000488759837753092, 'samples': 950352, 'steps': 6599, 'loss/train': 3.1399781703948975}
+01/23/2022 08:27:45 - INFO - codeparrot_training - Step 39599: {'lr': 0.000488759837753092, 'samples': 950376, 'steps': 6599, 'loss/train': 3.3391788005828857}
+01/23/2022 08:27:45 - INFO - codeparrot_training - Step 39600: {'lr': 0.000488759837753092, 'samples': 950400, 'steps': 6599, 'loss/train': 1.7753721475601196}
+01/23/2022 08:27:46 - INFO - codeparrot_training - Step 39601: {'lr': 0.0004887549861142967, 'samples': 950424, 'steps': 6600, 'loss/train': 2.9471864700317383}
+01/23/2022 08:27:46 - INFO - codeparrot_training - Step 39602: {'lr': 0.0004887549861142967, 'samples': 950448, 'steps': 6600, 'loss/train': 2.5053019523620605}
+01/23/2022 08:27:47 - INFO - codeparrot_training - Step 39603: {'lr': 0.0004887549861142967, 'samples': 950472, 'steps': 6600, 'loss/train': 3.2195205688476562}
+01/23/2022 08:27:47 - INFO - codeparrot_training - Step 39604: {'lr': 0.0004887549861142967, 'samples': 950496, 'steps': 6600, 'loss/train': 1.7440052032470703}
+01/23/2022 08:27:47 - INFO - codeparrot_training - Step 39605: {'lr': 0.0004887549861142967, 'samples': 950520, 'steps': 6600, 'loss/train': 2.304807424545288}
+01/23/2022 08:27:48 - INFO - codeparrot_training - Step 39606: {'lr': 0.0004887549861142967, 'samples': 950544, 'steps': 6600, 'loss/train': 2.8927066326141357}
+01/23/2022 08:27:48 - INFO - codeparrot_training - Step 39607: {'lr': 0.0004887501334527507, 'samples': 950568, 'steps': 6601, 'loss/train': 1.5773766040802002}
+01/23/2022 08:27:49 - INFO - codeparrot_training - Step 39608: {'lr': 0.0004887501334527507, 'samples': 950592, 'steps': 6601, 'loss/train': 2.050217628479004}
+01/23/2022 08:27:49 - INFO - codeparrot_training - Step 39609: {'lr': 0.0004887501334527507, 'samples': 950616, 'steps': 6601, 'loss/train': 3.5501928329467773}
+01/23/2022 08:27:49 - INFO - codeparrot_training - Step 39610: {'lr': 0.0004887501334527507, 'samples': 950640, 'steps': 6601, 'loss/train': 3.281006336212158}
+01/23/2022 08:27:50 - INFO - codeparrot_training - Step 39611: {'lr': 0.0004887501334527507, 'samples': 950664, 'steps': 6601, 'loss/train': 2.768867254257202}
+01/23/2022 08:27:50 - INFO - codeparrot_training - Step 39612: {'lr': 0.0004887501334527507, 'samples': 950688, 'steps': 6601, 'loss/train': 4.074234485626221}
+01/23/2022 08:27:50 - INFO - codeparrot_training - Step 39613: {'lr': 0.000488745279768475, 'samples': 950712, 'steps': 6602, 'loss/train': 2.7694294452667236}
+01/23/2022 08:27:51 - INFO - codeparrot_training - Step 39614: {'lr': 0.000488745279768475, 'samples': 950736, 'steps': 6602, 'loss/train': 2.3473575115203857}
+01/23/2022 08:27:52 - INFO - codeparrot_training - Step 39615: {'lr': 0.000488745279768475, 'samples': 950760, 'steps': 6602, 'loss/train': 2.6453909873962402}
+01/23/2022 08:27:52 - INFO - codeparrot_training - Step 39616: {'lr': 0.000488745279768475, 'samples': 950784, 'steps': 6602, 'loss/train': 1.6389124393463135}
+01/23/2022 08:27:53 - INFO - codeparrot_training - Step 39617: {'lr': 0.000488745279768475, 'samples': 950808, 'steps': 6602, 'loss/train': 2.645829200744629}
+01/23/2022 08:27:53 - INFO - codeparrot_training - Step 39618: {'lr': 0.000488745279768475, 'samples': 950832, 'steps': 6602, 'loss/train': 3.2772743701934814}
+01/23/2022 08:27:53 - INFO - codeparrot_training - Step 39619: {'lr': 0.0004887404250614904, 'samples': 950856, 'steps': 6603, 'loss/train': 2.6699893474578857}
+01/23/2022 08:27:54 - INFO - codeparrot_training - Step 39620: {'lr': 0.0004887404250614904, 'samples': 950880, 'steps': 6603, 'loss/train': 3.619605541229248}
+01/23/2022 08:27:54 - INFO - codeparrot_training - Step 39621: {'lr': 0.0004887404250614904, 'samples': 950904, 'steps': 6603, 'loss/train': 3.5268776416778564}
+01/23/2022 08:27:55 - INFO - codeparrot_training - Step 39622: {'lr': 0.0004887404250614904, 'samples': 950928, 'steps': 6603, 'loss/train': 3.38533616065979}
+01/23/2022 08:27:55 - INFO - codeparrot_training - Step 39623: {'lr': 0.0004887404250614904, 'samples': 950952, 'steps': 6603, 'loss/train': 2.1541051864624023}
+01/23/2022 08:27:55 - INFO - codeparrot_training - Step 39624: {'lr': 0.0004887404250614904, 'samples': 950976, 'steps': 6603, 'loss/train': 2.7339272499084473}
+01/23/2022 08:27:56 - INFO - codeparrot_training - Step 39625: {'lr': 0.0004887355693318176, 'samples': 951000, 'steps': 6604, 'loss/train': 4.221538543701172}
+01/23/2022 08:27:56 - INFO - codeparrot_training - Step 39626: {'lr': 0.0004887355693318176, 'samples': 951024, 'steps': 6604, 'loss/train': 1.8694554567337036}
+01/23/2022 08:27:56 - INFO - codeparrot_training - Step 39627: {'lr': 0.0004887355693318176, 'samples': 951048, 'steps': 6604, 'loss/train': 2.0234014987945557}
+01/23/2022 08:27:57 - INFO - codeparrot_training - Step 39628: {'lr': 0.0004887355693318176, 'samples': 951072, 'steps': 6604, 'loss/train': 1.902778148651123}
+01/23/2022 08:27:57 - INFO - codeparrot_training - Step 39629: {'lr': 0.0004887355693318176, 'samples': 951096, 'steps': 6604, 'loss/train': 1.0138972997665405}
+01/23/2022 08:27:58 - INFO - codeparrot_training - Step 39630: {'lr': 0.0004887355693318176, 'samples': 951120, 'steps': 6604, 'loss/train': 2.8060996532440186}
+01/23/2022 08:27:58 - INFO - codeparrot_training - Step 39631: {'lr': 0.0004887307125794775, 'samples': 951144, 'steps': 6605, 'loss/train': 3.3973779678344727}
+01/23/2022 08:27:58 - INFO - codeparrot_training - Step 39632: {'lr': 0.0004887307125794775, 'samples': 951168, 'steps': 6605, 'loss/train': 3.085364818572998}
+01/23/2022 08:27:59 - INFO - codeparrot_training - Step 39633: {'lr': 0.0004887307125794775, 'samples': 951192, 'steps': 6605, 'loss/train': 2.3371925354003906}
+01/23/2022 08:27:59 - INFO - codeparrot_training - Step 39634: {'lr': 0.0004887307125794775, 'samples': 951216, 'steps': 6605, 'loss/train': 2.557556390762329}
+01/23/2022 08:27:59 - INFO - codeparrot_training - Step 39635: {'lr': 0.0004887307125794775, 'samples': 951240, 'steps': 6605, 'loss/train': 2.3760151863098145}
+01/23/2022 08:28:00 - INFO - codeparrot_training - Step 39636: {'lr': 0.0004887307125794775, 'samples': 951264, 'steps': 6605, 'loss/train': 3.765437602996826}
+01/23/2022 08:28:00 - INFO - codeparrot_training - Step 39637: {'lr': 0.0004887258548044907, 'samples': 951288, 'steps': 6606, 'loss/train': 3.5777266025543213}
+01/23/2022 08:28:01 - INFO - codeparrot_training - Step 39638: {'lr': 0.0004887258548044907, 'samples': 951312, 'steps': 6606, 'loss/train': 3.3318393230438232}
+01/23/2022 08:28:01 - INFO - codeparrot_training - Step 39639: {'lr': 0.0004887258548044907, 'samples': 951336, 'steps': 6606, 'loss/train': 4.281674861907959}
+01/23/2022 08:28:01 - INFO - codeparrot_training - Step 39640: {'lr': 0.0004887258548044907, 'samples': 951360, 'steps': 6606, 'loss/train': 2.3631505966186523}
+01/23/2022 08:28:02 - INFO - codeparrot_training - Step 39641: {'lr': 0.0004887258548044907, 'samples': 951384, 'steps': 6606, 'loss/train': 2.681581497192383}
+01/23/2022 08:28:02 - INFO - codeparrot_training - Step 39642: {'lr': 0.0004887258548044907, 'samples': 951408, 'steps': 6606, 'loss/train': 1.7600399255752563}
+01/23/2022 08:28:03 - INFO - codeparrot_training - Step 39643: {'lr': 0.0004887209960068782, 'samples': 951432, 'steps': 6607, 'loss/train': 3.8398475646972656}
+01/23/2022 08:28:03 - INFO - codeparrot_training - Step 39644: {'lr': 0.0004887209960068782, 'samples': 951456, 'steps': 6607, 'loss/train': 4.603824138641357}
+01/23/2022 08:28:03 - INFO - codeparrot_training - Step 39645: {'lr': 0.0004887209960068782, 'samples': 951480, 'steps': 6607, 'loss/train': 2.931739568710327}
+01/23/2022 08:28:04 - INFO - codeparrot_training - Step 39646: {'lr': 0.0004887209960068782, 'samples': 951504, 'steps': 6607, 'loss/train': 3.85695481300354}
+01/23/2022 08:28:04 - INFO - codeparrot_training - Step 39647: {'lr': 0.0004887209960068782, 'samples': 951528, 'steps': 6607, 'loss/train': 3.0530128479003906}
+01/23/2022 08:28:04 - INFO - codeparrot_training - Step 39648: {'lr': 0.0004887209960068782, 'samples': 951552, 'steps': 6607, 'loss/train': 2.834651231765747}
+01/23/2022 08:28:05 - INFO - codeparrot_training - Step 39649: {'lr': 0.0004887161361866607, 'samples': 951576, 'steps': 6608, 'loss/train': 2.981818914413452}
+01/23/2022 08:28:05 - INFO - codeparrot_training - Step 39650: {'lr': 0.0004887161361866607, 'samples': 951600, 'steps': 6608, 'loss/train': 2.2917184829711914}
+01/23/2022 08:28:06 - INFO - codeparrot_training - Step 39651: {'lr': 0.0004887161361866607, 'samples': 951624, 'steps': 6608, 'loss/train': 2.799142837524414}
+01/23/2022 08:28:06 - INFO - codeparrot_training - Step 39652: {'lr': 0.0004887161361866607, 'samples': 951648, 'steps': 6608, 'loss/train': 5.215361595153809}
+01/23/2022 08:28:06 - INFO - codeparrot_training - Step 39653: {'lr': 0.0004887161361866607, 'samples': 951672, 'steps': 6608, 'loss/train': 3.030924081802368}
+01/23/2022 08:28:07 - INFO - codeparrot_training - Step 39654: {'lr': 0.0004887161361866607, 'samples': 951696, 'steps': 6608, 'loss/train': 3.294097423553467}
+01/23/2022 08:28:08 - INFO - codeparrot_training - Step 39655: {'lr': 0.0004887112753438592, 'samples': 951720, 'steps': 6609, 'loss/train': 2.8115415573120117}
+01/23/2022 08:28:08 - INFO - codeparrot_training - Step 39656: {'lr': 0.0004887112753438592, 'samples': 951744, 'steps': 6609, 'loss/train': 3.4084322452545166}
+01/23/2022 08:28:09 - INFO - codeparrot_training - Step 39657: {'lr': 0.0004887112753438592, 'samples': 951768, 'steps': 6609, 'loss/train': 2.0213329792022705}
+01/23/2022 08:28:09 - INFO - codeparrot_training - Step 39658: {'lr': 0.0004887112753438592, 'samples': 951792, 'steps': 6609, 'loss/train': 3.032414197921753}
+01/23/2022 08:28:09 - INFO - codeparrot_training - Step 39659: {'lr': 0.0004887112753438592, 'samples': 951816, 'steps': 6609, 'loss/train': 2.9837756156921387}
+01/23/2022 08:28:10 - INFO - codeparrot_training - Step 39660: {'lr': 0.0004887112753438592, 'samples': 951840, 'steps': 6609, 'loss/train': 1.6658883094787598}
+01/23/2022 08:28:10 - INFO - codeparrot_training - Step 39661: {'lr': 0.0004887064134784943, 'samples': 951864, 'steps': 6610, 'loss/train': 3.4511682987213135}
+01/23/2022 08:28:10 - INFO - codeparrot_training - Step 39662: {'lr': 0.0004887064134784943, 'samples': 951888, 'steps': 6610, 'loss/train': 4.321491718292236}
+01/23/2022 08:28:11 - INFO - codeparrot_training - Step 39663: {'lr': 0.0004887064134784943, 'samples': 951912, 'steps': 6610, 'loss/train': 1.6869323253631592}
+01/23/2022 08:28:11 - INFO - codeparrot_training - Step 39664: {'lr': 0.0004887064134784943, 'samples': 951936, 'steps': 6610, 'loss/train': 3.90250563621521}
+01/23/2022 08:28:12 - INFO - codeparrot_training - Step 39665: {'lr': 0.0004887064134784943, 'samples': 951960, 'steps': 6610, 'loss/train': 3.8186960220336914}
+01/23/2022 08:28:12 - INFO - codeparrot_training - Step 39666: {'lr': 0.0004887064134784943, 'samples': 951984, 'steps': 6610, 'loss/train': 3.0159332752227783}
+01/23/2022 08:28:12 - INFO - codeparrot_training - Step 39667: {'lr': 0.0004887015505905869, 'samples': 952008, 'steps': 6611, 'loss/train': 3.523495674133301}
+01/23/2022 08:28:13 - INFO - codeparrot_training - Step 39668: {'lr': 0.0004887015505905869, 'samples': 952032, 'steps': 6611, 'loss/train': 2.928607225418091}
+01/23/2022 08:28:13 - INFO - codeparrot_training - Step 39669: {'lr': 0.0004887015505905869, 'samples': 952056, 'steps': 6611, 'loss/train': 2.5257513523101807}
+01/23/2022 08:28:14 - INFO - codeparrot_training - Step 39670: {'lr': 0.0004887015505905869, 'samples': 952080, 'steps': 6611, 'loss/train': 2.0748000144958496}
+01/23/2022 08:28:14 - INFO - codeparrot_training - Step 39671: {'lr': 0.0004887015505905869, 'samples': 952104, 'steps': 6611, 'loss/train': 2.0762393474578857}
+01/23/2022 08:28:14 - INFO - codeparrot_training - Step 39672: {'lr': 0.0004887015505905869, 'samples': 952128, 'steps': 6611, 'loss/train': 2.969813346862793}
+01/23/2022 08:28:15 - INFO - codeparrot_training - Step 39673: {'lr': 0.0004886966866801579, 'samples': 952152, 'steps': 6612, 'loss/train': 4.1624836921691895}
+01/23/2022 08:28:15 - INFO - codeparrot_training - Step 39674: {'lr': 0.0004886966866801579, 'samples': 952176, 'steps': 6612, 'loss/train': 4.034207820892334}
+01/23/2022 08:28:15 - INFO - codeparrot_training - Step 39675: {'lr': 0.0004886966866801579, 'samples': 952200, 'steps': 6612, 'loss/train': 2.4532032012939453}
+01/23/2022 08:28:16 - INFO - codeparrot_training - Step 39676: {'lr': 0.0004886966866801579, 'samples': 952224, 'steps': 6612, 'loss/train': 2.713003635406494}
+01/23/2022 08:28:16 - INFO - codeparrot_training - Step 39677: {'lr': 0.0004886966866801579, 'samples': 952248, 'steps': 6612, 'loss/train': 3.7252025604248047}
+01/23/2022 08:28:17 - INFO - codeparrot_training - Step 39678: {'lr': 0.0004886966866801579, 'samples': 952272, 'steps': 6612, 'loss/train': 3.3695149421691895}
+01/23/2022 08:28:17 - INFO - codeparrot_training - Step 39679: {'lr': 0.0004886918217472281, 'samples': 952296, 'steps': 6613, 'loss/train': 3.3670287132263184}
+01/23/2022 08:28:17 - INFO - codeparrot_training - Step 39680: {'lr': 0.0004886918217472281, 'samples': 952320, 'steps': 6613, 'loss/train': 2.263645887374878}
+01/23/2022 08:28:18 - INFO - codeparrot_training - Step 39681: {'lr': 0.0004886918217472281, 'samples': 952344, 'steps': 6613, 'loss/train': 3.035715341567993}
+01/23/2022 08:28:18 - INFO - codeparrot_training - Step 39682: {'lr': 0.0004886918217472281, 'samples': 952368, 'steps': 6613, 'loss/train': 2.9020230770111084}
+01/23/2022 08:28:18 - INFO - codeparrot_training - Step 39683: {'lr': 0.0004886918217472281, 'samples': 952392, 'steps': 6613, 'loss/train': 4.027946472167969}
+01/23/2022 08:28:19 - INFO - codeparrot_training - Step 39684: {'lr': 0.0004886918217472281, 'samples': 952416, 'steps': 6613, 'loss/train': 2.30108904838562}
+01/23/2022 08:28:19 - INFO - codeparrot_training - Step 39685: {'lr': 0.0004886869557918183, 'samples': 952440, 'steps': 6614, 'loss/train': 2.699714422225952}
+01/23/2022 08:28:20 - INFO - codeparrot_training - Step 39686: {'lr': 0.0004886869557918183, 'samples': 952464, 'steps': 6614, 'loss/train': 0.7101748585700989}
+01/23/2022 08:28:20 - INFO - codeparrot_training - Step 39687: {'lr': 0.0004886869557918183, 'samples': 952488, 'steps': 6614, 'loss/train': 2.6827268600463867}
+01/23/2022 08:28:20 - INFO - codeparrot_training - Step 39688: {'lr': 0.0004886869557918183, 'samples': 952512, 'steps': 6614, 'loss/train': 3.5946552753448486}
+01/23/2022 08:28:21 - INFO - codeparrot_training - Step 39689: {'lr': 0.0004886869557918183, 'samples': 952536, 'steps': 6614, 'loss/train': 2.6861820220947266}
+01/23/2022 08:28:21 - INFO - codeparrot_training - Step 39690: {'lr': 0.0004886869557918183, 'samples': 952560, 'steps': 6614, 'loss/train': 3.122523069381714}
+01/23/2022 08:28:22 - INFO - codeparrot_training - Step 39691: {'lr': 0.0004886820888139494, 'samples': 952584, 'steps': 6615, 'loss/train': 3.3401145935058594}
+01/23/2022 08:28:22 - INFO - codeparrot_training - Step 39692: {'lr': 0.0004886820888139494, 'samples': 952608, 'steps': 6615, 'loss/train': 2.359253168106079}
+01/23/2022 08:28:22 - INFO - codeparrot_training - Step 39693: {'lr': 0.0004886820888139494, 'samples': 952632, 'steps': 6615, 'loss/train': 2.684231996536255}
+01/23/2022 08:28:25 - INFO - codeparrot_training - Step 39694: {'lr': 0.0004886820888139494, 'samples': 952656, 'steps': 6615, 'loss/train': 3.018141746520996}
+01/23/2022 08:28:25 - INFO - codeparrot_training - Step 39695: {'lr': 0.0004886820888139494, 'samples': 952680, 'steps': 6615, 'loss/train': 2.7553277015686035}
+01/23/2022 08:28:26 - INFO - codeparrot_training - Step 39696: {'lr': 0.0004886820888139494, 'samples': 952704, 'steps': 6615, 'loss/train': 3.1624393463134766}
+01/23/2022 08:28:26 - INFO - codeparrot_training - Step 39697: {'lr': 0.0004886772208136422, 'samples': 952728, 'steps': 6616, 'loss/train': 4.069413185119629}
+01/23/2022 08:28:27 - INFO - codeparrot_training - Step 39698: {'lr': 0.0004886772208136422, 'samples': 952752, 'steps': 6616, 'loss/train': 2.9990532398223877}
+01/23/2022 08:28:27 - INFO - codeparrot_training - Step 39699: {'lr': 0.0004886772208136422, 'samples': 952776, 'steps': 6616, 'loss/train': 3.723998785018921}
+01/23/2022 08:28:27 - INFO - codeparrot_training - Step 39700: {'lr': 0.0004886772208136422, 'samples': 952800, 'steps': 6616, 'loss/train': 4.172040939331055}
+01/23/2022 08:28:28 - INFO - codeparrot_training - Step 39701: {'lr': 0.0004886772208136422, 'samples': 952824, 'steps': 6616, 'loss/train': 3.834052562713623}
+01/23/2022 08:28:28 - INFO - codeparrot_training - Step 39702: {'lr': 0.0004886772208136422, 'samples': 952848, 'steps': 6616, 'loss/train': 2.643099784851074}
+01/23/2022 08:28:29 - INFO - codeparrot_training - Step 39703: {'lr': 0.0004886723517909176, 'samples': 952872, 'steps': 6617, 'loss/train': 3.206719398498535}
+01/23/2022 08:28:29 - INFO - codeparrot_training - Step 39704: {'lr': 0.0004886723517909176, 'samples': 952896, 'steps': 6617, 'loss/train': 3.1201930046081543}
+01/23/2022 08:28:29 - INFO - codeparrot_training - Step 39705: {'lr': 0.0004886723517909176, 'samples': 952920, 'steps': 6617, 'loss/train': 2.5064899921417236}
+01/23/2022 08:28:30 - INFO - codeparrot_training - Step 39706: {'lr': 0.0004886723517909176, 'samples': 952944, 'steps': 6617, 'loss/train': 3.446371078491211}
+01/23/2022 08:28:30 - INFO - codeparrot_training - Step 39707: {'lr': 0.0004886723517909176, 'samples': 952968, 'steps': 6617, 'loss/train': 2.5657005310058594}
+01/23/2022 08:28:30 - INFO - codeparrot_training - Step 39708: {'lr': 0.0004886723517909176, 'samples': 952992, 'steps': 6617, 'loss/train': 3.6336119174957275}
+01/23/2022 08:28:31 - INFO - codeparrot_training - Step 39709: {'lr': 0.0004886674817457964, 'samples': 953016, 'steps': 6618, 'loss/train': 2.6083478927612305}
+01/23/2022 08:28:31 - INFO - codeparrot_training - Step 39710: {'lr': 0.0004886674817457964, 'samples': 953040, 'steps': 6618, 'loss/train': 4.355567932128906}
+01/23/2022 08:28:32 - INFO - codeparrot_training - Step 39711: {'lr': 0.0004886674817457964, 'samples': 953064, 'steps': 6618, 'loss/train': 1.6141122579574585}
+01/23/2022 08:28:32 - INFO - codeparrot_training - Step 39712: {'lr': 0.0004886674817457964, 'samples': 953088, 'steps': 6618, 'loss/train': 3.27876615524292}
+01/23/2022 08:28:32 - INFO - codeparrot_training - Step 39713: {'lr': 0.0004886674817457964, 'samples': 953112, 'steps': 6618, 'loss/train': 3.631563186645508}
+01/23/2022 08:28:33 - INFO - codeparrot_training - Step 39714: {'lr': 0.0004886674817457964, 'samples': 953136, 'steps': 6618, 'loss/train': 4.630978107452393}
+01/23/2022 08:28:33 - INFO - codeparrot_training - Step 39715: {'lr': 0.0004886626106782995, 'samples': 953160, 'steps': 6619, 'loss/train': 2.8116071224212646}
+01/23/2022 08:28:34 - INFO - codeparrot_training - Step 39716: {'lr': 0.0004886626106782995, 'samples': 953184, 'steps': 6619, 'loss/train': 2.9362545013427734}
+01/23/2022 08:28:34 - INFO - codeparrot_training - Step 39717: {'lr': 0.0004886626106782995, 'samples': 953208, 'steps': 6619, 'loss/train': 3.3310537338256836}
+01/23/2022 08:28:34 - INFO - codeparrot_training - Step 39718: {'lr': 0.0004886626106782995, 'samples': 953232, 'steps': 6619, 'loss/train': 3.394538164138794}
+01/23/2022 08:28:35 - INFO - codeparrot_training - Step 39719: {'lr': 0.0004886626106782995, 'samples': 953256, 'steps': 6619, 'loss/train': 2.8919150829315186}
+01/23/2022 08:28:35 - INFO - codeparrot_training - Step 39720: {'lr': 0.0004886626106782995, 'samples': 953280, 'steps': 6619, 'loss/train': 4.02523946762085}
+01/23/2022 08:28:35 - INFO - codeparrot_training - Step 39721: {'lr': 0.0004886577385884478, 'samples': 953304, 'steps': 6620, 'loss/train': 2.4141478538513184}
+01/23/2022 08:28:36 - INFO - codeparrot_training - Step 39722: {'lr': 0.0004886577385884478, 'samples': 953328, 'steps': 6620, 'loss/train': 3.565589427947998}
+01/23/2022 08:28:36 - INFO - codeparrot_training - Step 39723: {'lr': 0.0004886577385884478, 'samples': 953352, 'steps': 6620, 'loss/train': 2.493309259414673}
+01/23/2022 08:28:37 - INFO - codeparrot_training - Step 39724: {'lr': 0.0004886577385884478, 'samples': 953376, 'steps': 6620, 'loss/train': 3.3862175941467285}
+01/23/2022 08:28:37 - INFO - codeparrot_training - Step 39725: {'lr': 0.0004886577385884478, 'samples': 953400, 'steps': 6620, 'loss/train': 3.5813992023468018}
+01/23/2022 08:28:37 - INFO - codeparrot_training - Step 39726: {'lr': 0.0004886577385884478, 'samples': 953424, 'steps': 6620, 'loss/train': 2.6412453651428223}
+01/23/2022 08:28:38 - INFO - codeparrot_training - Step 39727: {'lr': 0.0004886528654762621, 'samples': 953448, 'steps': 6621, 'loss/train': 3.6888575553894043}
+01/23/2022 08:28:38 - INFO - codeparrot_training - Step 39728: {'lr': 0.0004886528654762621, 'samples': 953472, 'steps': 6621, 'loss/train': 4.786043167114258}
+01/23/2022 08:28:39 - INFO - codeparrot_training - Step 39729: {'lr': 0.0004886528654762621, 'samples': 953496, 'steps': 6621, 'loss/train': 3.3301939964294434}
+01/23/2022 08:28:39 - INFO - codeparrot_training - Step 39730: {'lr': 0.0004886528654762621, 'samples': 953520, 'steps': 6621, 'loss/train': 3.4684503078460693}
+01/23/2022 08:28:39 - INFO - codeparrot_training - Step 39731: {'lr': 0.0004886528654762621, 'samples': 953544, 'steps': 6621, 'loss/train': 3.260162115097046}
+01/23/2022 08:28:40 - INFO - codeparrot_training - Step 39732: {'lr': 0.0004886528654762621, 'samples': 953568, 'steps': 6621, 'loss/train': 2.7052433490753174}
+01/23/2022 08:28:41 - INFO - codeparrot_training - Step 39733: {'lr': 0.0004886479913417633, 'samples': 953592, 'steps': 6622, 'loss/train': 4.328808307647705}
+01/23/2022 08:28:41 - INFO - codeparrot_training - Step 39734: {'lr': 0.0004886479913417633, 'samples': 953616, 'steps': 6622, 'loss/train': 3.184997081756592}
+01/23/2022 08:28:41 - INFO - codeparrot_training - Step 39735: {'lr': 0.0004886479913417633, 'samples': 953640, 'steps': 6622, 'loss/train': 3.5130677223205566}
+01/23/2022 08:28:42 - INFO - codeparrot_training - Step 39736: {'lr': 0.0004886479913417633, 'samples': 953664, 'steps': 6622, 'loss/train': 3.109422445297241}
+01/23/2022 08:28:42 - INFO - codeparrot_training - Step 39737: {'lr': 0.0004886479913417633, 'samples': 953688, 'steps': 6622, 'loss/train': 2.702784538269043}
+01/23/2022 08:28:43 - INFO - codeparrot_training - Step 39738: {'lr': 0.0004886479913417633, 'samples': 953712, 'steps': 6622, 'loss/train': 4.173056125640869}
+01/23/2022 08:28:43 - INFO - codeparrot_training - Step 39739: {'lr': 0.0004886431161849722, 'samples': 953736, 'steps': 6623, 'loss/train': 3.7053298950195312}
+01/23/2022 08:28:43 - INFO - codeparrot_training - Step 39740: {'lr': 0.0004886431161849722, 'samples': 953760, 'steps': 6623, 'loss/train': 3.010974884033203}
+01/23/2022 08:28:44 - INFO - codeparrot_training - Step 39741: {'lr': 0.0004886431161849722, 'samples': 953784, 'steps': 6623, 'loss/train': 2.027432680130005}
+01/23/2022 08:28:44 - INFO - codeparrot_training - Step 39742: {'lr': 0.0004886431161849722, 'samples': 953808, 'steps': 6623, 'loss/train': 3.090052604675293}
+01/23/2022 08:28:44 - INFO - codeparrot_training - Step 39743: {'lr': 0.0004886431161849722, 'samples': 953832, 'steps': 6623, 'loss/train': 3.2625107765197754}
+01/23/2022 08:28:45 - INFO - codeparrot_training - Step 39744: {'lr': 0.0004886431161849722, 'samples': 953856, 'steps': 6623, 'loss/train': 3.022860288619995}
+01/23/2022 08:28:45 - INFO - codeparrot_training - Step 39745: {'lr': 0.0004886382400059099, 'samples': 953880, 'steps': 6624, 'loss/train': 3.180166482925415}
+01/23/2022 08:28:46 - INFO - codeparrot_training - Step 39746: {'lr': 0.0004886382400059099, 'samples': 953904, 'steps': 6624, 'loss/train': 2.769979476928711}
+01/23/2022 08:28:46 - INFO - codeparrot_training - Step 39747: {'lr': 0.0004886382400059099, 'samples': 953928, 'steps': 6624, 'loss/train': 2.027430534362793}
+01/23/2022 08:28:46 - INFO - codeparrot_training - Step 39748: {'lr': 0.0004886382400059099, 'samples': 953952, 'steps': 6624, 'loss/train': 3.1798927783966064}
+01/23/2022 08:28:47 - INFO - codeparrot_training - Step 39749: {'lr': 0.0004886382400059099, 'samples': 953976, 'steps': 6624, 'loss/train': 4.100682258605957}
+01/23/2022 08:28:47 - INFO - codeparrot_training - Step 39750: {'lr': 0.0004886382400059099, 'samples': 954000, 'steps': 6624, 'loss/train': 2.0362727642059326}
+01/23/2022 08:28:48 - INFO - codeparrot_training - Step 39751: {'lr': 0.0004886333628045972, 'samples': 954024, 'steps': 6625, 'loss/train': 3.3066422939300537}
+01/23/2022 08:28:48 - INFO - codeparrot_training - Step 39752: {'lr': 0.0004886333628045972, 'samples': 954048, 'steps': 6625, 'loss/train': 2.6357946395874023}
+01/23/2022 08:28:48 - INFO - codeparrot_training - Step 39753: {'lr': 0.0004886333628045972, 'samples': 954072, 'steps': 6625, 'loss/train': 3.653426170349121}
+01/23/2022 08:28:49 - INFO - codeparrot_training - Step 39754: {'lr': 0.0004886333628045972, 'samples': 954096, 'steps': 6625, 'loss/train': 2.1607391834259033}
+01/23/2022 08:28:49 - INFO - codeparrot_training - Step 39755: {'lr': 0.0004886333628045972, 'samples': 954120, 'steps': 6625, 'loss/train': 3.0124433040618896}
+01/23/2022 08:28:49 - INFO - codeparrot_training - Step 39756: {'lr': 0.0004886333628045972, 'samples': 954144, 'steps': 6625, 'loss/train': 3.3708276748657227}
+01/23/2022 08:28:50 - INFO - codeparrot_training - Step 39757: {'lr': 0.0004886284845810548, 'samples': 954168, 'steps': 6626, 'loss/train': 2.02349591255188}
+01/23/2022 08:28:50 - INFO - codeparrot_training - Step 39758: {'lr': 0.0004886284845810548, 'samples': 954192, 'steps': 6626, 'loss/train': 3.0184271335601807}
+01/23/2022 08:28:51 - INFO - codeparrot_training - Step 39759: {'lr': 0.0004886284845810548, 'samples': 954216, 'steps': 6626, 'loss/train': 3.187565326690674}
+01/23/2022 08:28:51 - INFO - codeparrot_training - Step 39760: {'lr': 0.0004886284845810548, 'samples': 954240, 'steps': 6626, 'loss/train': 2.1506645679473877}
+01/23/2022 08:28:51 - INFO - codeparrot_training - Step 39761: {'lr': 0.0004886284845810548, 'samples': 954264, 'steps': 6626, 'loss/train': 2.4071402549743652}
+01/23/2022 08:28:52 - INFO - codeparrot_training - Step 39762: {'lr': 0.0004886284845810548, 'samples': 954288, 'steps': 6626, 'loss/train': 3.6269352436065674}
+01/23/2022 08:28:52 - INFO - codeparrot_training - Step 39763: {'lr': 0.0004886236053353038, 'samples': 954312, 'steps': 6627, 'loss/train': 2.9374659061431885}
+01/23/2022 08:28:53 - INFO - codeparrot_training - Step 39764: {'lr': 0.0004886236053353038, 'samples': 954336, 'steps': 6627, 'loss/train': 3.1129939556121826}
+01/23/2022 08:28:53 - INFO - codeparrot_training - Step 39765: {'lr': 0.0004886236053353038, 'samples': 954360, 'steps': 6627, 'loss/train': 3.7797646522521973}
+01/23/2022 08:28:53 - INFO - codeparrot_training - Step 39766: {'lr': 0.0004886236053353038, 'samples': 954384, 'steps': 6627, 'loss/train': 2.745943784713745}
+01/23/2022 08:28:54 - INFO - codeparrot_training - Step 39767: {'lr': 0.0004886236053353038, 'samples': 954408, 'steps': 6627, 'loss/train': 2.871887445449829}
+01/23/2022 08:28:54 - INFO - codeparrot_training - Step 39768: {'lr': 0.0004886236053353038, 'samples': 954432, 'steps': 6627, 'loss/train': 3.3493001461029053}
+01/23/2022 08:28:54 - INFO - codeparrot_training - Step 39769: {'lr': 0.000488618725067365, 'samples': 954456, 'steps': 6628, 'loss/train': 2.984370231628418}
+01/23/2022 08:28:55 - INFO - codeparrot_training - Step 39770: {'lr': 0.000488618725067365, 'samples': 954480, 'steps': 6628, 'loss/train': 4.166447162628174}
+01/23/2022 08:28:55 - INFO - codeparrot_training - Step 39771: {'lr': 0.000488618725067365, 'samples': 954504, 'steps': 6628, 'loss/train': 3.625199556350708}
+01/23/2022 08:28:56 - INFO - codeparrot_training - Step 39772: {'lr': 0.000488618725067365, 'samples': 954528, 'steps': 6628, 'loss/train': 2.118516206741333}
+01/23/2022 08:28:57 - INFO - codeparrot_training - Step 39773: {'lr': 0.000488618725067365, 'samples': 954552, 'steps': 6628, 'loss/train': 3.5929267406463623}
+01/23/2022 08:28:57 - INFO - codeparrot_training - Step 39774: {'lr': 0.000488618725067365, 'samples': 954576, 'steps': 6628, 'loss/train': 2.9025464057922363}
+01/23/2022 08:28:57 - INFO - codeparrot_training - Step 39775: {'lr': 0.0004886138437772594, 'samples': 954600, 'steps': 6629, 'loss/train': 3.403226375579834}
+01/23/2022 08:28:58 - INFO - codeparrot_training - Step 39776: {'lr': 0.0004886138437772594, 'samples': 954624, 'steps': 6629, 'loss/train': 2.967594623565674}
+01/23/2022 08:28:58 - INFO - codeparrot_training - Step 39777: {'lr': 0.0004886138437772594, 'samples': 954648, 'steps': 6629, 'loss/train': 2.400503396987915}
+01/23/2022 08:28:59 - INFO - codeparrot_training - Step 39778: {'lr': 0.0004886138437772594, 'samples': 954672, 'steps': 6629, 'loss/train': 3.3236186504364014}
+01/23/2022 08:28:59 - INFO - codeparrot_training - Step 39779: {'lr': 0.0004886138437772594, 'samples': 954696, 'steps': 6629, 'loss/train': 3.807459592819214}
+01/23/2022 08:28:59 - INFO - codeparrot_training - Step 39780: {'lr': 0.0004886138437772594, 'samples': 954720, 'steps': 6629, 'loss/train': 3.834021806716919}
+01/23/2022 08:29:00 - INFO - codeparrot_training - Step 39781: {'lr': 0.0004886089614650078, 'samples': 954744, 'steps': 6630, 'loss/train': 3.2713115215301514}
+01/23/2022 08:29:00 - INFO - codeparrot_training - Step 39782: {'lr': 0.0004886089614650078, 'samples': 954768, 'steps': 6630, 'loss/train': 2.7740492820739746}
+01/23/2022 08:29:00 - INFO - codeparrot_training - Step 39783: {'lr': 0.0004886089614650078, 'samples': 954792, 'steps': 6630, 'loss/train': 2.6998703479766846}
+01/23/2022 08:29:01 - INFO - codeparrot_training - Step 39784: {'lr': 0.0004886089614650078, 'samples': 954816, 'steps': 6630, 'loss/train': 3.9995789527893066}
+01/23/2022 08:29:01 - INFO - codeparrot_training - Step 39785: {'lr': 0.0004886089614650078, 'samples': 954840, 'steps': 6630, 'loss/train': 1.41749906539917}
+01/23/2022 08:29:02 - INFO - codeparrot_training - Step 39786: {'lr': 0.0004886089614650078, 'samples': 954864, 'steps': 6630, 'loss/train': 2.569209098815918}
+01/23/2022 08:29:02 - INFO - codeparrot_training - Step 39787: {'lr': 0.0004886040781306313, 'samples': 954888, 'steps': 6631, 'loss/train': 3.80253529548645}
+01/23/2022 08:29:02 - INFO - codeparrot_training - Step 39788: {'lr': 0.0004886040781306313, 'samples': 954912, 'steps': 6631, 'loss/train': 4.367063522338867}
+01/23/2022 08:29:03 - INFO - codeparrot_training - Step 39789: {'lr': 0.0004886040781306313, 'samples': 954936, 'steps': 6631, 'loss/train': 3.3183581829071045}
+01/23/2022 08:29:03 - INFO - codeparrot_training - Step 39790: {'lr': 0.0004886040781306313, 'samples': 954960, 'steps': 6631, 'loss/train': 2.5329484939575195}
+01/23/2022 08:29:04 - INFO - codeparrot_training - Step 39791: {'lr': 0.0004886040781306313, 'samples': 954984, 'steps': 6631, 'loss/train': 4.191631317138672}
+01/23/2022 08:29:04 - INFO - codeparrot_training - Step 39792: {'lr': 0.0004886040781306313, 'samples': 955008, 'steps': 6631, 'loss/train': 3.2131054401397705}
+01/23/2022 08:29:04 - INFO - codeparrot_training - Step 39793: {'lr': 0.0004885991937741506, 'samples': 955032, 'steps': 6632, 'loss/train': 3.575927257537842}
+01/23/2022 08:29:05 - INFO - codeparrot_training - Step 39794: {'lr': 0.0004885991937741506, 'samples': 955056, 'steps': 6632, 'loss/train': 3.045315742492676}
+01/23/2022 08:29:05 - INFO - codeparrot_training - Step 39795: {'lr': 0.0004885991937741506, 'samples': 955080, 'steps': 6632, 'loss/train': 4.060422420501709}
+01/23/2022 08:29:05 - INFO - codeparrot_training - Step 39796: {'lr': 0.0004885991937741506, 'samples': 955104, 'steps': 6632, 'loss/train': 2.6450183391571045}
+01/23/2022 08:29:06 - INFO - codeparrot_training - Step 39797: {'lr': 0.0004885991937741506, 'samples': 955128, 'steps': 6632, 'loss/train': 3.043550968170166}
+01/23/2022 08:29:06 - INFO - codeparrot_training - Step 39798: {'lr': 0.0004885991937741506, 'samples': 955152, 'steps': 6632, 'loss/train': 3.259909152984619}
+01/23/2022 08:29:07 - INFO - codeparrot_training - Step 39799: {'lr': 0.0004885943083955868, 'samples': 955176, 'steps': 6633, 'loss/train': 2.945530891418457}
+01/23/2022 08:29:07 - INFO - codeparrot_training - Step 39800: {'lr': 0.0004885943083955868, 'samples': 955200, 'steps': 6633, 'loss/train': 3.9796769618988037}
+01/23/2022 08:29:07 - INFO - codeparrot_training - Step 39801: {'lr': 0.0004885943083955868, 'samples': 955224, 'steps': 6633, 'loss/train': 4.195952892303467}
+01/23/2022 08:29:08 - INFO - codeparrot_training - Step 39802: {'lr': 0.0004885943083955868, 'samples': 955248, 'steps': 6633, 'loss/train': 3.4458603858947754}
+01/23/2022 08:29:08 - INFO - codeparrot_training - Step 39803: {'lr': 0.0004885943083955868, 'samples': 955272, 'steps': 6633, 'loss/train': 3.8160367012023926}
+01/23/2022 08:29:08 - INFO - codeparrot_training - Step 39804: {'lr': 0.0004885943083955868, 'samples': 955296, 'steps': 6633, 'loss/train': 3.4284820556640625}
+01/23/2022 08:29:09 - INFO - codeparrot_training - Step 39805: {'lr': 0.0004885894219949607, 'samples': 955320, 'steps': 6634, 'loss/train': 3.6932671070098877}
+01/23/2022 08:29:09 - INFO - codeparrot_training - Step 39806: {'lr': 0.0004885894219949607, 'samples': 955344, 'steps': 6634, 'loss/train': 4.3689284324646}
+01/23/2022 08:29:10 - INFO - codeparrot_training - Step 39807: {'lr': 0.0004885894219949607, 'samples': 955368, 'steps': 6634, 'loss/train': 2.1058051586151123}
+01/23/2022 08:29:10 - INFO - codeparrot_training - Step 39808: {'lr': 0.0004885894219949607, 'samples': 955392, 'steps': 6634, 'loss/train': 3.4729204177856445}
+01/23/2022 08:29:10 - INFO - codeparrot_training - Step 39809: {'lr': 0.0004885894219949607, 'samples': 955416, 'steps': 6634, 'loss/train': 3.149629831314087}
+01/23/2022 08:29:11 - INFO - codeparrot_training - Step 39810: {'lr': 0.0004885894219949607, 'samples': 955440, 'steps': 6634, 'loss/train': 1.4435852766036987}
+01/23/2022 08:29:11 - INFO - codeparrot_training - Step 39811: {'lr': 0.0004885845345722932, 'samples': 955464, 'steps': 6635, 'loss/train': 2.522855043411255}
+01/23/2022 08:29:13 - INFO - codeparrot_training - Step 39812: {'lr': 0.0004885845345722932, 'samples': 955488, 'steps': 6635, 'loss/train': 3.1711254119873047}
+01/23/2022 08:29:13 - INFO - codeparrot_training - Step 39813: {'lr': 0.0004885845345722932, 'samples': 955512, 'steps': 6635, 'loss/train': 3.3608157634735107}
+01/23/2022 08:29:13 - INFO - codeparrot_training - Step 39814: {'lr': 0.0004885845345722932, 'samples': 955536, 'steps': 6635, 'loss/train': 3.0416762828826904}
+01/23/2022 08:29:14 - INFO - codeparrot_training - Step 39815: {'lr': 0.0004885845345722932, 'samples': 955560, 'steps': 6635, 'loss/train': 3.283557176589966}
+01/23/2022 08:29:14 - INFO - codeparrot_training - Step 39816: {'lr': 0.0004885845345722932, 'samples': 955584, 'steps': 6635, 'loss/train': 2.9243712425231934}
+01/23/2022 08:29:14 - INFO - codeparrot_training - Step 39817: {'lr': 0.0004885796461276055, 'samples': 955608, 'steps': 6636, 'loss/train': 3.6367290019989014}
+01/23/2022 08:29:15 - INFO - codeparrot_training - Step 39818: {'lr': 0.0004885796461276055, 'samples': 955632, 'steps': 6636, 'loss/train': 2.713850975036621}
+01/23/2022 08:29:15 - INFO - codeparrot_training - Step 39819: {'lr': 0.0004885796461276055, 'samples': 955656, 'steps': 6636, 'loss/train': 3.083235740661621}
+01/23/2022 08:29:16 - INFO - codeparrot_training - Step 39820: {'lr': 0.0004885796461276055, 'samples': 955680, 'steps': 6636, 'loss/train': 2.662104368209839}
+01/23/2022 08:29:16 - INFO - codeparrot_training - Step 39821: {'lr': 0.0004885796461276055, 'samples': 955704, 'steps': 6636, 'loss/train': 2.761251211166382}
+01/23/2022 08:29:16 - INFO - codeparrot_training - Step 39822: {'lr': 0.0004885796461276055, 'samples': 955728, 'steps': 6636, 'loss/train': 3.3743999004364014}
+01/23/2022 08:29:17 - INFO - codeparrot_training - Step 39823: {'lr': 0.0004885747566609182, 'samples': 955752, 'steps': 6637, 'loss/train': 2.2175025939941406}
+01/23/2022 08:29:17 - INFO - codeparrot_training - Step 39824: {'lr': 0.0004885747566609182, 'samples': 955776, 'steps': 6637, 'loss/train': 2.662386417388916}
+01/23/2022 08:29:17 - INFO - codeparrot_training - Step 39825: {'lr': 0.0004885747566609182, 'samples': 955800, 'steps': 6637, 'loss/train': 2.1893649101257324}
+01/23/2022 08:29:18 - INFO - codeparrot_training - Step 39826: {'lr': 0.0004885747566609182, 'samples': 955824, 'steps': 6637, 'loss/train': 2.6791975498199463}
+01/23/2022 08:29:18 - INFO - codeparrot_training - Step 39827: {'lr': 0.0004885747566609182, 'samples': 955848, 'steps': 6637, 'loss/train': 2.922384023666382}
+01/23/2022 08:29:19 - INFO - codeparrot_training - Step 39828: {'lr': 0.0004885747566609182, 'samples': 955872, 'steps': 6637, 'loss/train': 2.8408634662628174}
+01/23/2022 08:29:19 - INFO - codeparrot_training - Step 39829: {'lr': 0.0004885698661722524, 'samples': 955896, 'steps': 6638, 'loss/train': 2.590585470199585}
+01/23/2022 08:29:19 - INFO - codeparrot_training - Step 39830: {'lr': 0.0004885698661722524, 'samples': 955920, 'steps': 6638, 'loss/train': 3.581235647201538}
+01/23/2022 08:29:20 - INFO - codeparrot_training - Step 39831: {'lr': 0.0004885698661722524, 'samples': 955944, 'steps': 6638, 'loss/train': 3.1158182621002197}
+01/23/2022 08:29:20 - INFO - codeparrot_training - Step 39832: {'lr': 0.0004885698661722524, 'samples': 955968, 'steps': 6638, 'loss/train': 3.819568157196045}
+01/23/2022 08:29:20 - INFO - codeparrot_training - Step 39833: {'lr': 0.0004885698661722524, 'samples': 955992, 'steps': 6638, 'loss/train': 4.542995452880859}
+01/23/2022 08:29:21 - INFO - codeparrot_training - Step 39834: {'lr': 0.0004885698661722524, 'samples': 956016, 'steps': 6638, 'loss/train': 4.221615314483643}
+01/23/2022 08:29:21 - INFO - codeparrot_training - Step 39835: {'lr': 0.0004885649746616291, 'samples': 956040, 'steps': 6639, 'loss/train': 3.048348903656006}
+01/23/2022 08:29:22 - INFO - codeparrot_training - Step 39836: {'lr': 0.0004885649746616291, 'samples': 956064, 'steps': 6639, 'loss/train': 3.056980848312378}
+01/23/2022 08:29:22 - INFO - codeparrot_training - Step 39837: {'lr': 0.0004885649746616291, 'samples': 956088, 'steps': 6639, 'loss/train': 3.1610851287841797}
+01/23/2022 08:29:22 - INFO - codeparrot_training - Step 39838: {'lr': 0.0004885649746616291, 'samples': 956112, 'steps': 6639, 'loss/train': 1.819608449935913}
+01/23/2022 08:29:23 - INFO - codeparrot_training - Step 39839: {'lr': 0.0004885649746616291, 'samples': 956136, 'steps': 6639, 'loss/train': 2.2544469833374023}
+01/23/2022 08:29:23 - INFO - codeparrot_training - Step 39840: {'lr': 0.0004885649746616291, 'samples': 956160, 'steps': 6639, 'loss/train': 2.982330560684204}
+01/23/2022 08:29:24 - INFO - codeparrot_training - Step 39841: {'lr': 0.0004885600821290692, 'samples': 956184, 'steps': 6640, 'loss/train': 3.409172773361206}
+01/23/2022 08:29:24 - INFO - codeparrot_training - Step 39842: {'lr': 0.0004885600821290692, 'samples': 956208, 'steps': 6640, 'loss/train': 2.2998411655426025}
+01/23/2022 08:29:24 - INFO - codeparrot_training - Step 39843: {'lr': 0.0004885600821290692, 'samples': 956232, 'steps': 6640, 'loss/train': 2.53373122215271}
+01/23/2022 08:29:25 - INFO - codeparrot_training - Step 39844: {'lr': 0.0004885600821290692, 'samples': 956256, 'steps': 6640, 'loss/train': 3.5081567764282227}
+01/23/2022 08:29:25 - INFO - codeparrot_training - Step 39845: {'lr': 0.0004885600821290692, 'samples': 956280, 'steps': 6640, 'loss/train': 3.3911802768707275}
+01/23/2022 08:29:25 - INFO - codeparrot_training - Step 39846: {'lr': 0.0004885600821290692, 'samples': 956304, 'steps': 6640, 'loss/train': 3.513537645339966}
+01/23/2022 08:29:26 - INFO - codeparrot_training - Step 39847: {'lr': 0.0004885551885745937, 'samples': 956328, 'steps': 6641, 'loss/train': 3.2590630054473877}
+01/23/2022 08:29:26 - INFO - codeparrot_training - Step 39848: {'lr': 0.0004885551885745937, 'samples': 956352, 'steps': 6641, 'loss/train': 3.5807740688323975}
+01/23/2022 08:29:27 - INFO - codeparrot_training - Step 39849: {'lr': 0.0004885551885745937, 'samples': 956376, 'steps': 6641, 'loss/train': 2.770341634750366}
+01/23/2022 08:29:27 - INFO - codeparrot_training - Step 39850: {'lr': 0.0004885551885745937, 'samples': 956400, 'steps': 6641, 'loss/train': 3.400106430053711}
+01/23/2022 08:29:27 - INFO - codeparrot_training - Step 39851: {'lr': 0.0004885551885745937, 'samples': 956424, 'steps': 6641, 'loss/train': 4.496646881103516}
+01/23/2022 08:29:28 - INFO - codeparrot_training - Step 39852: {'lr': 0.0004885551885745937, 'samples': 956448, 'steps': 6641, 'loss/train': 3.446043014526367}
+01/23/2022 08:29:28 - INFO - codeparrot_training - Step 39853: {'lr': 0.0004885502939982235, 'samples': 956472, 'steps': 6642, 'loss/train': 4.089207649230957}
+01/23/2022 08:29:29 - INFO - codeparrot_training - Step 39854: {'lr': 0.0004885502939982235, 'samples': 956496, 'steps': 6642, 'loss/train': 3.452099323272705}
+01/23/2022 08:29:30 - INFO - codeparrot_training - Step 39855: {'lr': 0.0004885502939982235, 'samples': 956520, 'steps': 6642, 'loss/train': 2.869022846221924}
+01/23/2022 08:29:30 - INFO - codeparrot_training - Step 39856: {'lr': 0.0004885502939982235, 'samples': 956544, 'steps': 6642, 'loss/train': 3.4155802726745605}
+01/23/2022 08:29:30 - INFO - codeparrot_training - Step 39857: {'lr': 0.0004885502939982235, 'samples': 956568, 'steps': 6642, 'loss/train': 3.8459722995758057}
+01/23/2022 08:29:31 - INFO - codeparrot_training - Step 39858: {'lr': 0.0004885502939982235, 'samples': 956592, 'steps': 6642, 'loss/train': 3.001433849334717}
+01/23/2022 08:29:31 - INFO - codeparrot_training - Step 39859: {'lr': 0.0004885453983999795, 'samples': 956616, 'steps': 6643, 'loss/train': 3.3134593963623047}
+01/23/2022 08:29:32 - INFO - codeparrot_training - Step 39860: {'lr': 0.0004885453983999795, 'samples': 956640, 'steps': 6643, 'loss/train': 3.269183874130249}
+01/23/2022 08:29:32 - INFO - codeparrot_training - Step 39861: {'lr': 0.0004885453983999795, 'samples': 956664, 'steps': 6643, 'loss/train': 3.6981899738311768}
+01/23/2022 08:29:32 - INFO - codeparrot_training - Step 39862: {'lr': 0.0004885453983999795, 'samples': 956688, 'steps': 6643, 'loss/train': 4.101418495178223}
+01/23/2022 08:29:33 - INFO - codeparrot_training - Step 39863: {'lr': 0.0004885453983999795, 'samples': 956712, 'steps': 6643, 'loss/train': 3.088718891143799}
+01/23/2022 08:29:33 - INFO - codeparrot_training - Step 39864: {'lr': 0.0004885453983999795, 'samples': 956736, 'steps': 6643, 'loss/train': 2.6822032928466797}
+01/23/2022 08:29:34 - INFO - codeparrot_training - Step 39865: {'lr': 0.0004885405017798828, 'samples': 956760, 'steps': 6644, 'loss/train': 2.9116995334625244}
+01/23/2022 08:29:34 - INFO - codeparrot_training - Step 39866: {'lr': 0.0004885405017798828, 'samples': 956784, 'steps': 6644, 'loss/train': 3.2902159690856934}
+01/23/2022 08:29:34 - INFO - codeparrot_training - Step 39867: {'lr': 0.0004885405017798828, 'samples': 956808, 'steps': 6644, 'loss/train': 3.2901036739349365}
+01/23/2022 08:29:35 - INFO - codeparrot_training - Step 39868: {'lr': 0.0004885405017798828, 'samples': 956832, 'steps': 6644, 'loss/train': 3.1714375019073486}
+01/23/2022 08:29:35 - INFO - codeparrot_training - Step 39869: {'lr': 0.0004885405017798828, 'samples': 956856, 'steps': 6644, 'loss/train': 2.9770548343658447}
+01/23/2022 08:29:35 - INFO - codeparrot_training - Step 39870: {'lr': 0.0004885405017798828, 'samples': 956880, 'steps': 6644, 'loss/train': 5.494380950927734}
+01/23/2022 08:29:36 - INFO - codeparrot_training - Step 39871: {'lr': 0.0004885356041379544, 'samples': 956904, 'steps': 6645, 'loss/train': 2.590120553970337}
+01/23/2022 08:29:36 - INFO - codeparrot_training - Step 39872: {'lr': 0.0004885356041379544, 'samples': 956928, 'steps': 6645, 'loss/train': 3.3389928340911865}
+01/23/2022 08:29:37 - INFO - codeparrot_training - Step 39873: {'lr': 0.0004885356041379544, 'samples': 956952, 'steps': 6645, 'loss/train': 2.1112148761749268}
+01/23/2022 08:29:37 - INFO - codeparrot_training - Step 39874: {'lr': 0.0004885356041379544, 'samples': 956976, 'steps': 6645, 'loss/train': 3.1791787147521973}
+01/23/2022 08:29:37 - INFO - codeparrot_training - Step 39875: {'lr': 0.0004885356041379544, 'samples': 957000, 'steps': 6645, 'loss/train': 2.027475357055664}
+01/23/2022 08:29:38 - INFO - codeparrot_training - Step 39876: {'lr': 0.0004885356041379544, 'samples': 957024, 'steps': 6645, 'loss/train': 3.418513774871826}
+01/23/2022 08:29:38 - INFO - codeparrot_training - Step 39877: {'lr': 0.0004885307054742151, 'samples': 957048, 'steps': 6646, 'loss/train': 2.9661426544189453}
+01/23/2022 08:29:38 - INFO - codeparrot_training - Step 39878: {'lr': 0.0004885307054742151, 'samples': 957072, 'steps': 6646, 'loss/train': 2.2386481761932373}
+01/23/2022 08:29:39 - INFO - codeparrot_training - Step 39879: {'lr': 0.0004885307054742151, 'samples': 957096, 'steps': 6646, 'loss/train': 0.5203653573989868}
+01/23/2022 08:29:39 - INFO - codeparrot_training - Step 39880: {'lr': 0.0004885307054742151, 'samples': 957120, 'steps': 6646, 'loss/train': 3.1756415367126465}
+01/23/2022 08:29:40 - INFO - codeparrot_training - Step 39881: {'lr': 0.0004885307054742151, 'samples': 957144, 'steps': 6646, 'loss/train': 3.2371983528137207}
+01/23/2022 08:29:40 - INFO - codeparrot_training - Step 39882: {'lr': 0.0004885307054742151, 'samples': 957168, 'steps': 6646, 'loss/train': 2.9927401542663574}
+01/23/2022 08:29:40 - INFO - codeparrot_training - Step 39883: {'lr': 0.0004885258057886861, 'samples': 957192, 'steps': 6647, 'loss/train': 2.91530704498291}
+01/23/2022 08:29:41 - INFO - codeparrot_training - Step 39884: {'lr': 0.0004885258057886861, 'samples': 957216, 'steps': 6647, 'loss/train': 3.3105757236480713}
+01/23/2022 08:29:41 - INFO - codeparrot_training - Step 39885: {'lr': 0.0004885258057886861, 'samples': 957240, 'steps': 6647, 'loss/train': 3.223963499069214}
+01/23/2022 08:29:42 - INFO - codeparrot_training - Step 39886: {'lr': 0.0004885258057886861, 'samples': 957264, 'steps': 6647, 'loss/train': 1.2863751649856567}
+01/23/2022 08:29:42 - INFO - codeparrot_training - Step 39887: {'lr': 0.0004885258057886861, 'samples': 957288, 'steps': 6647, 'loss/train': 3.888192653656006}
+01/23/2022 08:29:42 - INFO - codeparrot_training - Step 39888: {'lr': 0.0004885258057886861, 'samples': 957312, 'steps': 6647, 'loss/train': 3.485015869140625}
+01/23/2022 08:29:43 - INFO - codeparrot_training - Step 39889: {'lr': 0.0004885209050813882, 'samples': 957336, 'steps': 6648, 'loss/train': 2.9341423511505127}
+01/23/2022 08:29:43 - INFO - codeparrot_training - Step 39890: {'lr': 0.0004885209050813882, 'samples': 957360, 'steps': 6648, 'loss/train': 4.055736064910889}
+01/23/2022 08:29:43 - INFO - codeparrot_training - Step 39891: {'lr': 0.0004885209050813882, 'samples': 957384, 'steps': 6648, 'loss/train': 3.517564535140991}
+01/23/2022 08:29:44 - INFO - codeparrot_training - Step 39892: {'lr': 0.0004885209050813882, 'samples': 957408, 'steps': 6648, 'loss/train': 3.396437644958496}
+01/23/2022 08:29:44 - INFO - codeparrot_training - Step 39893: {'lr': 0.0004885209050813882, 'samples': 957432, 'steps': 6648, 'loss/train': 1.649721622467041}
+01/23/2022 08:29:46 - INFO - codeparrot_training - Step 39894: {'lr': 0.0004885209050813882, 'samples': 957456, 'steps': 6648, 'loss/train': 1.8731651306152344}
+01/23/2022 08:29:46 - INFO - codeparrot_training - Step 39895: {'lr': 0.0004885160033523426, 'samples': 957480, 'steps': 6649, 'loss/train': 3.2513067722320557}
+01/23/2022 08:29:47 - INFO - codeparrot_training - Step 39896: {'lr': 0.0004885160033523426, 'samples': 957504, 'steps': 6649, 'loss/train': 3.538451910018921}
+01/23/2022 08:29:47 - INFO - codeparrot_training - Step 39897: {'lr': 0.0004885160033523426, 'samples': 957528, 'steps': 6649, 'loss/train': 3.929492235183716}
+01/23/2022 08:29:47 - INFO - codeparrot_training - Step 39898: {'lr': 0.0004885160033523426, 'samples': 957552, 'steps': 6649, 'loss/train': 3.3803365230560303}
+01/23/2022 08:29:48 - INFO - codeparrot_training - Step 39899: {'lr': 0.0004885160033523426, 'samples': 957576, 'steps': 6649, 'loss/train': 4.467901229858398}
+01/23/2022 08:29:48 - INFO - codeparrot_training - Step 39900: {'lr': 0.0004885160033523426, 'samples': 957600, 'steps': 6649, 'loss/train': 2.768684148788452}
+01/23/2022 08:29:49 - INFO - codeparrot_training - Step 39901: {'lr': 0.0004885111006015701, 'samples': 957624, 'steps': 6650, 'loss/train': 3.2786989212036133}
+01/23/2022 08:29:49 - INFO - codeparrot_training - Step 39902: {'lr': 0.0004885111006015701, 'samples': 957648, 'steps': 6650, 'loss/train': 3.276458263397217}
+01/23/2022 08:29:49 - INFO - codeparrot_training - Step 39903: {'lr': 0.0004885111006015701, 'samples': 957672, 'steps': 6650, 'loss/train': 3.26737642288208}
+01/23/2022 08:29:50 - INFO - codeparrot_training - Step 39904: {'lr': 0.0004885111006015701, 'samples': 957696, 'steps': 6650, 'loss/train': 1.518012523651123}
+01/23/2022 08:29:50 - INFO - codeparrot_training - Step 39905: {'lr': 0.0004885111006015701, 'samples': 957720, 'steps': 6650, 'loss/train': 2.3423309326171875}
+01/23/2022 08:29:50 - INFO - codeparrot_training - Step 39906: {'lr': 0.0004885111006015701, 'samples': 957744, 'steps': 6650, 'loss/train': 2.7226200103759766}
+01/23/2022 08:29:51 - INFO - codeparrot_training - Step 39907: {'lr': 0.0004885061968290919, 'samples': 957768, 'steps': 6651, 'loss/train': 1.627081036567688}
+01/23/2022 08:29:51 - INFO - codeparrot_training - Step 39908: {'lr': 0.0004885061968290919, 'samples': 957792, 'steps': 6651, 'loss/train': 4.238592624664307}
+01/23/2022 08:29:52 - INFO - codeparrot_training - Step 39909: {'lr': 0.0004885061968290919, 'samples': 957816, 'steps': 6651, 'loss/train': 3.1236557960510254}
+01/23/2022 08:29:52 - INFO - codeparrot_training - Step 39910: {'lr': 0.0004885061968290919, 'samples': 957840, 'steps': 6651, 'loss/train': 1.35121488571167}
+01/23/2022 08:29:52 - INFO - codeparrot_training - Step 39911: {'lr': 0.0004885061968290919, 'samples': 957864, 'steps': 6651, 'loss/train': 0.28902384638786316}
+01/23/2022 08:29:53 - INFO - codeparrot_training - Step 39912: {'lr': 0.0004885061968290919, 'samples': 957888, 'steps': 6651, 'loss/train': 2.614633560180664}
+01/23/2022 08:29:53 - INFO - codeparrot_training - Step 39913: {'lr': 0.0004885012920349287, 'samples': 957912, 'steps': 6652, 'loss/train': 4.164730072021484}
+01/23/2022 08:29:53 - INFO - codeparrot_training - Step 39914: {'lr': 0.0004885012920349287, 'samples': 957936, 'steps': 6652, 'loss/train': 4.035120964050293}
+01/23/2022 08:29:54 - INFO - codeparrot_training - Step 39915: {'lr': 0.0004885012920349287, 'samples': 957960, 'steps': 6652, 'loss/train': 3.973980188369751}
+01/23/2022 08:29:54 - INFO - codeparrot_training - Step 39916: {'lr': 0.0004885012920349287, 'samples': 957984, 'steps': 6652, 'loss/train': 5.49979829788208}
+01/23/2022 08:29:55 - INFO - codeparrot_training - Step 39917: {'lr': 0.0004885012920349287, 'samples': 958008, 'steps': 6652, 'loss/train': 4.136147975921631}
+01/23/2022 08:29:55 - INFO - codeparrot_training - Step 39918: {'lr': 0.0004885012920349287, 'samples': 958032, 'steps': 6652, 'loss/train': 4.04859733581543}
+01/23/2022 08:29:55 - INFO - codeparrot_training - Step 39919: {'lr': 0.0004884963862191018, 'samples': 958056, 'steps': 6653, 'loss/train': 4.262698650360107}
+01/23/2022 08:29:56 - INFO - codeparrot_training - Step 39920: {'lr': 0.0004884963862191018, 'samples': 958080, 'steps': 6653, 'loss/train': 5.1488165855407715}
+01/23/2022 08:29:56 - INFO - codeparrot_training - Step 39921: {'lr': 0.0004884963862191018, 'samples': 958104, 'steps': 6653, 'loss/train': 4.4063262939453125}
+01/23/2022 08:29:57 - INFO - codeparrot_training - Step 39922: {'lr': 0.0004884963862191018, 'samples': 958128, 'steps': 6653, 'loss/train': 4.5016961097717285}
+01/23/2022 08:29:57 - INFO - codeparrot_training - Step 39923: {'lr': 0.0004884963862191018, 'samples': 958152, 'steps': 6653, 'loss/train': 5.2277679443359375}
+01/23/2022 08:29:57 - INFO - codeparrot_training - Step 39924: {'lr': 0.0004884963862191018, 'samples': 958176, 'steps': 6653, 'loss/train': 4.282363414764404}
+01/23/2022 08:29:58 - INFO - codeparrot_training - Step 39925: {'lr': 0.0004884914793816321, 'samples': 958200, 'steps': 6654, 'loss/train': 4.179851531982422}
+01/23/2022 08:29:58 - INFO - codeparrot_training - Step 39926: {'lr': 0.0004884914793816321, 'samples': 958224, 'steps': 6654, 'loss/train': 4.129681587219238}
+01/23/2022 08:29:58 - INFO - codeparrot_training - Step 39927: {'lr': 0.0004884914793816321, 'samples': 958248, 'steps': 6654, 'loss/train': 3.082751750946045}
+01/23/2022 08:29:59 - INFO - codeparrot_training - Step 39928: {'lr': 0.0004884914793816321, 'samples': 958272, 'steps': 6654, 'loss/train': 2.447608232498169}
+01/23/2022 08:29:59 - INFO - codeparrot_training - Step 39929: {'lr': 0.0004884914793816321, 'samples': 958296, 'steps': 6654, 'loss/train': 3.4752938747406006}
+01/23/2022 08:30:00 - INFO - codeparrot_training - Step 39930: {'lr': 0.0004884914793816321, 'samples': 958320, 'steps': 6654, 'loss/train': 2.25624942779541}
+01/23/2022 08:30:00 - INFO - codeparrot_training - Step 39931: {'lr': 0.0004884865715225407, 'samples': 958344, 'steps': 6655, 'loss/train': 3.2024900913238525}
+01/23/2022 08:30:00 - INFO - codeparrot_training - Step 39932: {'lr': 0.0004884865715225407, 'samples': 958368, 'steps': 6655, 'loss/train': 6.2524094581604}
+01/23/2022 08:30:01 - INFO - codeparrot_training - Step 39933: {'lr': 0.0004884865715225407, 'samples': 958392, 'steps': 6655, 'loss/train': 1.5896258354187012}
+01/23/2022 08:30:01 - INFO - codeparrot_training - Step 39934: {'lr': 0.0004884865715225407, 'samples': 958416, 'steps': 6655, 'loss/train': 3.6058921813964844}
+01/23/2022 08:30:01 - INFO - codeparrot_training - Step 39935: {'lr': 0.0004884865715225407, 'samples': 958440, 'steps': 6655, 'loss/train': 3.55676531791687}
+01/23/2022 08:30:02 - INFO - codeparrot_training - Step 39936: {'lr': 0.0004884865715225407, 'samples': 958464, 'steps': 6655, 'loss/train': 3.236042022705078}
+01/23/2022 08:30:02 - INFO - codeparrot_training - Step 39937: {'lr': 0.0004884816626418484, 'samples': 958488, 'steps': 6656, 'loss/train': 2.6322202682495117}
+01/23/2022 08:30:03 - INFO - codeparrot_training - Step 39938: {'lr': 0.0004884816626418484, 'samples': 958512, 'steps': 6656, 'loss/train': 3.089087724685669}
+01/23/2022 08:30:04 - INFO - codeparrot_training - Step 39939: {'lr': 0.0004884816626418484, 'samples': 958536, 'steps': 6656, 'loss/train': 4.48184061050415}
+01/23/2022 08:30:04 - INFO - codeparrot_training - Step 39940: {'lr': 0.0004884816626418484, 'samples': 958560, 'steps': 6656, 'loss/train': 3.16591215133667}
+01/23/2022 08:30:04 - INFO - codeparrot_training - Step 39941: {'lr': 0.0004884816626418484, 'samples': 958584, 'steps': 6656, 'loss/train': 2.313385009765625}
+01/23/2022 08:30:05 - INFO - codeparrot_training - Step 39942: {'lr': 0.0004884816626418484, 'samples': 958608, 'steps': 6656, 'loss/train': 3.636476516723633}
+01/23/2022 08:30:05 - INFO - codeparrot_training - Step 39943: {'lr': 0.0004884767527395765, 'samples': 958632, 'steps': 6657, 'loss/train': 3.8470137119293213}
+01/23/2022 08:30:06 - INFO - codeparrot_training - Step 39944: {'lr': 0.0004884767527395765, 'samples': 958656, 'steps': 6657, 'loss/train': 3.3568716049194336}
+01/23/2022 08:30:06 - INFO - codeparrot_training - Step 39945: {'lr': 0.0004884767527395765, 'samples': 958680, 'steps': 6657, 'loss/train': 3.7812230587005615}
+01/23/2022 08:30:06 - INFO - codeparrot_training - Step 39946: {'lr': 0.0004884767527395765, 'samples': 958704, 'steps': 6657, 'loss/train': 2.2676970958709717}
+01/23/2022 08:30:07 - INFO - codeparrot_training - Step 39947: {'lr': 0.0004884767527395765, 'samples': 958728, 'steps': 6657, 'loss/train': 3.740279197692871}
+01/23/2022 08:30:07 - INFO - codeparrot_training - Step 39948: {'lr': 0.0004884767527395765, 'samples': 958752, 'steps': 6657, 'loss/train': 3.1229186058044434}
+01/23/2022 08:30:08 - INFO - codeparrot_training - Step 39949: {'lr': 0.0004884718418157459, 'samples': 958776, 'steps': 6658, 'loss/train': 3.1710000038146973}
+01/23/2022 08:30:08 - INFO - codeparrot_training - Step 39950: {'lr': 0.0004884718418157459, 'samples': 958800, 'steps': 6658, 'loss/train': 4.361971378326416}
+01/23/2022 08:30:08 - INFO - codeparrot_training - Step 39951: {'lr': 0.0004884718418157459, 'samples': 958824, 'steps': 6658, 'loss/train': 3.5005061626434326}
+01/23/2022 08:30:09 - INFO - codeparrot_training - Step 39952: {'lr': 0.0004884718418157459, 'samples': 958848, 'steps': 6658, 'loss/train': 3.8436758518218994}
+01/23/2022 08:30:09 - INFO - codeparrot_training - Step 39953: {'lr': 0.0004884718418157459, 'samples': 958872, 'steps': 6658, 'loss/train': 3.2606756687164307}
+01/23/2022 08:30:09 - INFO - codeparrot_training - Step 39954: {'lr': 0.0004884718418157459, 'samples': 958896, 'steps': 6658, 'loss/train': 4.555755138397217}
+01/23/2022 08:30:10 - INFO - codeparrot_training - Step 39955: {'lr': 0.0004884669298703775, 'samples': 958920, 'steps': 6659, 'loss/train': 3.4722156524658203}
+01/23/2022 08:30:10 - INFO - codeparrot_training - Step 39956: {'lr': 0.0004884669298703775, 'samples': 958944, 'steps': 6659, 'loss/train': 3.8061954975128174}
+01/23/2022 08:30:11 - INFO - codeparrot_training - Step 39957: {'lr': 0.0004884669298703775, 'samples': 958968, 'steps': 6659, 'loss/train': 2.7510790824890137}
+01/23/2022 08:30:11 - INFO - codeparrot_training - Step 39958: {'lr': 0.0004884669298703775, 'samples': 958992, 'steps': 6659, 'loss/train': 3.3413379192352295}
+01/23/2022 08:30:11 - INFO - codeparrot_training - Step 39959: {'lr': 0.0004884669298703775, 'samples': 959016, 'steps': 6659, 'loss/train': 2.6616463661193848}
+01/23/2022 08:30:12 - INFO - codeparrot_training - Step 39960: {'lr': 0.0004884669298703775, 'samples': 959040, 'steps': 6659, 'loss/train': 2.90570330619812}
+01/23/2022 08:30:12 - INFO - codeparrot_training - Step 39961: {'lr': 0.0004884620169034927, 'samples': 959064, 'steps': 6660, 'loss/train': 2.0795249938964844}
+01/23/2022 08:30:13 - INFO - codeparrot_training - Step 39962: {'lr': 0.0004884620169034927, 'samples': 959088, 'steps': 6660, 'loss/train': 1.352937936782837}
+01/23/2022 08:30:13 - INFO - codeparrot_training - Step 39963: {'lr': 0.0004884620169034927, 'samples': 959112, 'steps': 6660, 'loss/train': 2.9954187870025635}
+01/23/2022 08:30:13 - INFO - codeparrot_training - Step 39964: {'lr': 0.0004884620169034927, 'samples': 959136, 'steps': 6660, 'loss/train': 1.479219675064087}
+01/23/2022 08:30:14 - INFO - codeparrot_training - Step 39965: {'lr': 0.0004884620169034927, 'samples': 959160, 'steps': 6660, 'loss/train': 3.185002088546753}
+01/23/2022 08:30:14 - INFO - codeparrot_training - Step 39966: {'lr': 0.0004884620169034927, 'samples': 959184, 'steps': 6660, 'loss/train': 3.5054240226745605}
+01/23/2022 08:30:14 - INFO - codeparrot_training - Step 39967: {'lr': 0.0004884571029151123, 'samples': 959208, 'steps': 6661, 'loss/train': 2.84767746925354}
+01/23/2022 08:30:15 - INFO - codeparrot_training - Step 39968: {'lr': 0.0004884571029151123, 'samples': 959232, 'steps': 6661, 'loss/train': 4.004669666290283}
+01/23/2022 08:30:15 - INFO - codeparrot_training - Step 39969: {'lr': 0.0004884571029151123, 'samples': 959256, 'steps': 6661, 'loss/train': 3.277655601501465}
+01/23/2022 08:30:16 - INFO - codeparrot_training - Step 39970: {'lr': 0.0004884571029151123, 'samples': 959280, 'steps': 6661, 'loss/train': 3.101933479309082}
+01/23/2022 08:30:16 - INFO - codeparrot_training - Step 39971: {'lr': 0.0004884571029151123, 'samples': 959304, 'steps': 6661, 'loss/train': 3.928311824798584}
+01/23/2022 08:30:16 - INFO - codeparrot_training - Step 39972: {'lr': 0.0004884571029151123, 'samples': 959328, 'steps': 6661, 'loss/train': 3.798422336578369}
+01/23/2022 08:30:17 - INFO - codeparrot_training - Step 39973: {'lr': 0.0004884521879052573, 'samples': 959352, 'steps': 6662, 'loss/train': 3.1176626682281494}
+01/23/2022 08:30:17 - INFO - codeparrot_training - Step 39974: {'lr': 0.0004884521879052573, 'samples': 959376, 'steps': 6662, 'loss/train': 2.654510259628296}
+01/23/2022 08:30:17 - INFO - codeparrot_training - Step 39975: {'lr': 0.0004884521879052573, 'samples': 959400, 'steps': 6662, 'loss/train': 2.766543388366699}
+01/23/2022 08:30:18 - INFO - codeparrot_training - Step 39976: {'lr': 0.0004884521879052573, 'samples': 959424, 'steps': 6662, 'loss/train': 3.171295166015625}
+01/23/2022 08:30:18 - INFO - codeparrot_training - Step 39977: {'lr': 0.0004884521879052573, 'samples': 959448, 'steps': 6662, 'loss/train': 0.4512699246406555}
+01/23/2022 08:30:19 - INFO - codeparrot_training - Step 39978: {'lr': 0.0004884521879052573, 'samples': 959472, 'steps': 6662, 'loss/train': 2.9288363456726074}
+01/23/2022 08:30:19 - INFO - codeparrot_training - Step 39979: {'lr': 0.000488447271873949, 'samples': 959496, 'steps': 6663, 'loss/train': 3.1982343196868896}
+01/23/2022 08:30:19 - INFO - codeparrot_training - Step 39980: {'lr': 0.000488447271873949, 'samples': 959520, 'steps': 6663, 'loss/train': 2.804089307785034}
+01/23/2022 08:30:22 - INFO - codeparrot_training - Step 39981: {'lr': 0.000488447271873949, 'samples': 959544, 'steps': 6663, 'loss/train': 2.6990373134613037}
+01/23/2022 08:30:23 - INFO - codeparrot_training - Step 39982: {'lr': 0.000488447271873949, 'samples': 959568, 'steps': 6663, 'loss/train': 4.412657260894775}
+01/23/2022 08:30:23 - INFO - codeparrot_training - Step 39983: {'lr': 0.000488447271873949, 'samples': 959592, 'steps': 6663, 'loss/train': 3.632756471633911}
+01/23/2022 08:30:23 - INFO - codeparrot_training - Step 39984: {'lr': 0.000488447271873949, 'samples': 959616, 'steps': 6663, 'loss/train': 4.059080123901367}
+01/23/2022 08:30:24 - INFO - codeparrot_training - Step 39985: {'lr': 0.0004884423548212082, 'samples': 959640, 'steps': 6664, 'loss/train': 3.10762095451355}
+01/23/2022 08:30:24 - INFO - codeparrot_training - Step 39986: {'lr': 0.0004884423548212082, 'samples': 959664, 'steps': 6664, 'loss/train': 3.3630433082580566}
+01/23/2022 08:30:25 - INFO - codeparrot_training - Step 39987: {'lr': 0.0004884423548212082, 'samples': 959688, 'steps': 6664, 'loss/train': 3.03529953956604}
+01/23/2022 08:30:25 - INFO - codeparrot_training - Step 39988: {'lr': 0.0004884423548212082, 'samples': 959712, 'steps': 6664, 'loss/train': 4.152653217315674}
+01/23/2022 08:30:25 - INFO - codeparrot_training - Step 39989: {'lr': 0.0004884423548212082, 'samples': 959736, 'steps': 6664, 'loss/train': 3.928889036178589}
+01/23/2022 08:30:26 - INFO - codeparrot_training - Step 39990: {'lr': 0.0004884423548212082, 'samples': 959760, 'steps': 6664, 'loss/train': 3.6399424076080322}
+01/23/2022 08:30:26 - INFO - codeparrot_training - Step 39991: {'lr': 0.000488437436747056, 'samples': 959784, 'steps': 6665, 'loss/train': 3.409499168395996}
+01/23/2022 08:30:27 - INFO - codeparrot_training - Step 39992: {'lr': 0.000488437436747056, 'samples': 959808, 'steps': 6665, 'loss/train': 2.5945727825164795}
+01/23/2022 08:30:27 - INFO - codeparrot_training - Step 39993: {'lr': 0.000488437436747056, 'samples': 959832, 'steps': 6665, 'loss/train': 1.3407015800476074}
+01/23/2022 08:30:27 - INFO - codeparrot_training - Step 39994: {'lr': 0.000488437436747056, 'samples': 959856, 'steps': 6665, 'loss/train': 4.079473972320557}
+01/23/2022 08:30:28 - INFO - codeparrot_training - Step 39995: {'lr': 0.000488437436747056, 'samples': 959880, 'steps': 6665, 'loss/train': 3.096470594406128}
+01/23/2022 08:30:28 - INFO - codeparrot_training - Step 39996: {'lr': 0.000488437436747056, 'samples': 959904, 'steps': 6665, 'loss/train': 2.0152385234832764}
+01/23/2022 08:30:29 - INFO - codeparrot_training - Step 39997: {'lr': 0.0004884325176515137, 'samples': 959928, 'steps': 6666, 'loss/train': 4.447699546813965}
+01/23/2022 08:30:29 - INFO - codeparrot_training - Step 39998: {'lr': 0.0004884325176515137, 'samples': 959952, 'steps': 6666, 'loss/train': 2.3649871349334717}
+01/23/2022 08:30:29 - INFO - codeparrot_training - Step 39999: {'lr': 0.0004884325176515137, 'samples': 959976, 'steps': 6666, 'loss/train': 1.8183807134628296}
+01/23/2022 08:30:30 - INFO - codeparrot_training - Step 40000: {'lr': 0.0004884325176515137, 'samples': 960000, 'steps': 6666, 'loss/train': 1.8346413373947144}
+01/23/2022 08:30:30 - INFO - codeparrot_training - Evaluating and saving model checkpoint
+01/23/2022 08:32:05 - INFO - codeparrot_training - Evaluation loss: 2.9963340759277344 perplexity: 20.012039184570312
+01/23/2022 08:32:05 - INFO - codeparrot_training - Step 40000: {'loss/eval': 2.9963340759277344, 'perplexity': 20.012039184570312}