diff --git "a/log/debug_0.log" "b/log/debug_0.log"
--- "a/log/debug_0.log"
+++ "b/log/debug_0.log"
@@ -47584,3 +47584,1009 @@ Use FP16 precision: False
 02/26/2022 14:26:46 - INFO - codeparrot_training - Step 46998: {'lr': 4.810066315719175e-06, 'samples': 24063488, 'steps': 46998, 'loss/train': 1.6281605958938599}
 02/26/2022 14:26:52 - INFO - codeparrot_training - Step 46999: {'lr': 4.806872582291038e-06, 'samples': 24064000, 'steps': 46999, 'loss/train': 1.2316577434539795}
 02/26/2022 14:26:52 - INFO - codeparrot_training - Evaluating and saving model checkpoint
+02/26/2022 14:27:10 - WARNING - huggingface_hub.repository - Several commits (47) will be pushed upstream.
+02/26/2022 14:27:10 - WARNING - huggingface_hub.repository - The progress bars may be unreliable.
+02/26/2022 14:27:44 - WARNING - huggingface_hub.repository - To https://huggingface.co/ncoop57/multi-code-clippy
+   572893f..2e5c310  floral-grass-11 -> floral-grass-11
+
+02/26/2022 14:27:48 - INFO - codeparrot_training - Step 47000: {'lr': 4.803679899192393e-06, 'samples': 24064512, 'steps': 47000, 'loss/train': 1.5690922737121582}
+02/26/2022 14:27:56 - INFO - codeparrot_training - Step 47001: {'lr': 4.8004882664369195e-06, 'samples': 24065024, 'steps': 47001, 'loss/train': 2.004244804382324}
+02/26/2022 14:27:59 - INFO - codeparrot_training - Step 47002: {'lr': 4.797297684038304e-06, 'samples': 24065536, 'steps': 47002, 'loss/train': 2.0392816066741943}
+02/26/2022 14:28:03 - INFO - codeparrot_training - Step 47003: {'lr': 4.794108152010201e-06, 'samples': 24066048, 'steps': 47003, 'loss/train': 2.2683329582214355}
+02/26/2022 14:28:08 - INFO - codeparrot_training - Step 47004: {'lr': 4.790919670366295e-06, 'samples': 24066560, 'steps': 47004, 'loss/train': 2.059736490249634}
+02/26/2022 14:28:12 - INFO - codeparrot_training - Step 47005: {'lr': 4.787732239120185e-06, 'samples': 24067072, 'steps': 47005, 'loss/train': 0.6227854490280151}
+02/26/2022 14:28:17 - INFO - codeparrot_training - Step 47006: {'lr': 4.784545858285611e-06, 'samples': 24067584, 'steps': 47006, 'loss/train': 2.4907712936401367}
+02/26/2022 14:28:21 - INFO - codeparrot_training - Step 47007: {'lr': 4.781360527876144e-06, 'samples': 24068096, 'steps': 47007, 'loss/train': 1.6988836526870728}
+02/26/2022 14:28:27 - INFO - codeparrot_training - Step 47008: {'lr': 4.778176247905497e-06, 'samples': 24068608, 'steps': 47008, 'loss/train': 1.662015676498413}
+02/26/2022 14:28:30 - INFO - codeparrot_training - Step 47009: {'lr': 4.774993018387241e-06, 'samples': 24069120, 'steps': 47009, 'loss/train': 2.578217029571533}
+02/26/2022 14:28:35 - INFO - codeparrot_training - Step 47010: {'lr': 4.7718108393350615e-06, 'samples': 24069632, 'steps': 47010, 'loss/train': 2.2534897327423096}
+02/26/2022 14:28:39 - INFO - codeparrot_training - Step 47011: {'lr': 4.7686297107625846e-06, 'samples': 24070144, 'steps': 47011, 'loss/train': 0.869590699672699}
+02/26/2022 14:28:46 - INFO - codeparrot_training - Step 47012: {'lr': 4.7654496326834105e-06, 'samples': 24070656, 'steps': 47012, 'loss/train': 1.3880791664123535}
+02/26/2022 14:28:50 - INFO - codeparrot_training - Step 47013: {'lr': 4.762270605111169e-06, 'samples': 24071168, 'steps': 47013, 'loss/train': 1.8887298107147217}
+02/26/2022 14:28:55 - INFO - codeparrot_training - Step 47014: {'lr': 4.759092628059514e-06, 'samples': 24071680, 'steps': 47014, 'loss/train': 0.9282025098800659}
+02/26/2022 14:28:58 - INFO - codeparrot_training - Step 47015: {'lr': 4.755915701542019e-06, 'samples': 24072192, 'steps': 47015, 'loss/train': 2.086386203765869}
+02/26/2022 14:29:04 - INFO - codeparrot_training - Step 47016: {'lr': 4.752739825572339e-06, 'samples': 24072704, 'steps': 47016, 'loss/train': 1.3414241075515747}
+02/26/2022 14:29:07 - INFO - codeparrot_training - Step 47017: {'lr': 4.749565000164019e-06, 'samples': 24073216, 'steps': 47017, 'loss/train': 1.908897042274475}
+02/26/2022 14:29:13 - INFO - codeparrot_training - Step 47018: {'lr': 4.746391225330687e-06, 'samples': 24073728, 'steps': 47018, 'loss/train': 0.7843557596206665}
+02/26/2022 14:29:17 - INFO - codeparrot_training - Step 47019: {'lr': 4.743218501085916e-06, 'samples': 24074240, 'steps': 47019, 'loss/train': 1.6694109439849854}
+02/26/2022 14:29:22 - INFO - codeparrot_training - Step 47020: {'lr': 4.74004682744339e-06, 'samples': 24074752, 'steps': 47020, 'loss/train': 3.3496742248535156}
+02/26/2022 14:29:26 - INFO - codeparrot_training - Step 47021: {'lr': 4.73687620441654e-06, 'samples': 24075264, 'steps': 47021, 'loss/train': 1.4146969318389893}
+02/26/2022 14:29:33 - INFO - codeparrot_training - Step 47022: {'lr': 4.73370663201908e-06, 'samples': 24075776, 'steps': 47022, 'loss/train': 0.36437684297561646}
+02/26/2022 14:29:36 - INFO - codeparrot_training - Step 47023: {'lr': 4.730538110264526e-06, 'samples': 24076288, 'steps': 47023, 'loss/train': 2.077320098876953}
+02/26/2022 14:29:42 - INFO - codeparrot_training - Step 47024: {'lr': 4.727370639166506e-06, 'samples': 24076800, 'steps': 47024, 'loss/train': 1.8466856479644775}
+02/26/2022 14:29:45 - INFO - codeparrot_training - Step 47025: {'lr': 4.724204218738482e-06, 'samples': 24077312, 'steps': 47025, 'loss/train': 1.3524516820907593}
+02/26/2022 14:29:51 - INFO - codeparrot_training - Step 47026: {'lr': 4.721038848994136e-06, 'samples': 24077824, 'steps': 47026, 'loss/train': 0.6863139271736145}
+02/26/2022 14:29:57 - INFO - codeparrot_training - Step 47027: {'lr': 4.717874529946931e-06, 'samples': 24078336, 'steps': 47027, 'loss/train': 1.8413203954696655}
+02/26/2022 14:30:00 - INFO - codeparrot_training - Step 47028: {'lr': 4.714711261610494e-06, 'samples': 24078848, 'steps': 47028, 'loss/train': 0.9946373105049133}
+02/26/2022 14:30:05 - INFO - codeparrot_training - Step 47029: {'lr': 4.7115490439983425e-06, 'samples': 24079360, 'steps': 47029, 'loss/train': 2.519493818283081}
+02/26/2022 14:30:09 - INFO - codeparrot_training - Step 47030: {'lr': 4.7083878771240205e-06, 'samples': 24079872, 'steps': 47030, 'loss/train': 1.7761424779891968}
+02/26/2022 14:30:14 - INFO - codeparrot_training - Step 47031: {'lr': 4.705227761001074e-06, 'samples': 24080384, 'steps': 47031, 'loss/train': 2.5566325187683105}
+02/26/2022 14:30:18 - INFO - codeparrot_training - Step 47032: {'lr': 4.702068695643047e-06, 'samples': 24080896, 'steps': 47032, 'loss/train': 2.4438281059265137}
+02/26/2022 14:30:24 - INFO - codeparrot_training - Step 47033: {'lr': 4.698910681063484e-06, 'samples': 24081408, 'steps': 47033, 'loss/train': 1.466216802597046}
+02/26/2022 14:30:27 - INFO - codeparrot_training - Step 47034: {'lr': 4.695753717275875e-06, 'samples': 24081920, 'steps': 47034, 'loss/train': 2.448518991470337}
+02/26/2022 14:30:33 - INFO - codeparrot_training - Step 47035: {'lr': 4.6925978042937636e-06, 'samples': 24082432, 'steps': 47035, 'loss/train': 2.0654125213623047}
+02/26/2022 14:30:36 - INFO - codeparrot_training - Step 47036: {'lr': 4.689442942130667e-06, 'samples': 24082944, 'steps': 47036, 'loss/train': 0.5978533029556274}
+02/26/2022 14:30:40 - INFO - codeparrot_training - Step 47037: {'lr': 4.686289130800131e-06, 'samples': 24083456, 'steps': 47037, 'loss/train': 0.4005730450153351}
+02/26/2022 14:30:47 - INFO - codeparrot_training - Step 47038: {'lr': 4.683136370315616e-06, 'samples': 24083968, 'steps': 47038, 'loss/train': 2.3211472034454346}
+02/26/2022 14:30:52 - INFO - codeparrot_training - Step 47039: {'lr': 4.679984660690639e-06, 'samples': 24084480, 'steps': 47039, 'loss/train': 1.9712399244308472}
+02/26/2022 14:30:56 - INFO - codeparrot_training - Step 47040: {'lr': 4.676834001938718e-06, 'samples': 24084992, 'steps': 47040, 'loss/train': 1.4720009565353394}
+02/26/2022 14:31:01 - INFO - codeparrot_training - Step 47041: {'lr': 4.673684394073341e-06, 'samples': 24085504, 'steps': 47041, 'loss/train': 2.356607675552368}
+02/26/2022 14:31:05 - INFO - codeparrot_training - Step 47042: {'lr': 4.6705358371079974e-06, 'samples': 24086016, 'steps': 47042, 'loss/train': 1.11809241771698}
+02/26/2022 14:31:10 - INFO - codeparrot_training - Step 47043: {'lr': 4.6673883310561775e-06, 'samples': 24086528, 'steps': 47043, 'loss/train': 2.2559280395507812}
+02/26/2022 14:31:14 - INFO - codeparrot_training - Step 47044: {'lr': 4.664241875931368e-06, 'samples': 24087040, 'steps': 47044, 'loss/train': 0.714023232460022}
+02/26/2022 14:31:19 - INFO - codeparrot_training - Step 47045: {'lr': 4.661096471747034e-06, 'samples': 24087552, 'steps': 47045, 'loss/train': 1.7764296531677246}
+02/26/2022 14:31:23 - INFO - codeparrot_training - Step 47046: {'lr': 4.6579521185166895e-06, 'samples': 24088064, 'steps': 47046, 'loss/train': 0.6463498473167419}
+02/26/2022 14:31:29 - INFO - codeparrot_training - Step 47047: {'lr': 4.654808816253742e-06, 'samples': 24088576, 'steps': 47047, 'loss/train': 1.051945447921753}
+02/26/2022 14:31:32 - INFO - codeparrot_training - Step 47048: {'lr': 4.651666564971679e-06, 'samples': 24089088, 'steps': 47048, 'loss/train': 1.0134605169296265}
+02/26/2022 14:31:35 - INFO - codeparrot_training - Step 47049: {'lr': 4.648525364683992e-06, 'samples': 24089600, 'steps': 47049, 'loss/train': 2.4695560932159424}
+02/26/2022 14:31:42 - INFO - codeparrot_training - Step 47050: {'lr': 4.645385215404113e-06, 'samples': 24090112, 'steps': 47050, 'loss/train': 1.4363354444503784}
+02/26/2022 14:31:46 - INFO - codeparrot_training - Step 47051: {'lr': 4.642246117145505e-06, 'samples': 24090624, 'steps': 47051, 'loss/train': 2.5005080699920654}
+02/26/2022 14:31:51 - INFO - codeparrot_training - Step 47052: {'lr': 4.6391080699215716e-06, 'samples': 24091136, 'steps': 47052, 'loss/train': 2.089142322540283}
+02/26/2022 14:31:55 - INFO - codeparrot_training - Step 47053: {'lr': 4.635971073745832e-06, 'samples': 24091648, 'steps': 47053, 'loss/train': 1.2328163385391235}
+02/26/2022 14:32:00 - INFO - codeparrot_training - Step 47054: {'lr': 4.632835128631635e-06, 'samples': 24092160, 'steps': 47054, 'loss/train': 2.4630727767944336}
+02/26/2022 14:32:04 - INFO - codeparrot_training - Step 47055: {'lr': 4.629700234592498e-06, 'samples': 24092672, 'steps': 47055, 'loss/train': 1.7775522470474243}
+02/26/2022 14:32:09 - INFO - codeparrot_training - Step 47056: {'lr': 4.626566391641774e-06, 'samples': 24093184, 'steps': 47056, 'loss/train': 1.4107571840286255}
+02/26/2022 14:32:13 - INFO - codeparrot_training - Step 47057: {'lr': 4.6234335997929474e-06, 'samples': 24093696, 'steps': 47057, 'loss/train': 0.07298669964075089}
+02/26/2022 14:32:18 - INFO - codeparrot_training - Step 47058: {'lr': 4.6203018590594e-06, 'samples': 24094208, 'steps': 47058, 'loss/train': 0.7180238366127014}
+02/26/2022 14:32:22 - INFO - codeparrot_training - Step 47059: {'lr': 4.617171169454621e-06, 'samples': 24094720, 'steps': 47059, 'loss/train': 2.3873889446258545}
+02/26/2022 14:32:28 - INFO - codeparrot_training - Step 47060: {'lr': 4.6140415309919026e-06, 'samples': 24095232, 'steps': 47060, 'loss/train': 2.0663018226623535}
+02/26/2022 14:32:31 - INFO - codeparrot_training - Step 47061: {'lr': 4.610912943684709e-06, 'samples': 24095744, 'steps': 47061, 'loss/train': 2.0417423248291016}
+02/26/2022 14:32:37 - INFO - codeparrot_training - Step 47062: {'lr': 4.607785407546444e-06, 'samples': 24096256, 'steps': 47062, 'loss/train': 0.7667291164398193}
+02/26/2022 14:32:40 - INFO - codeparrot_training - Step 47063: {'lr': 4.604658922590543e-06, 'samples': 24096768, 'steps': 47063, 'loss/train': 2.317803144454956}
+02/26/2022 14:32:46 - INFO - codeparrot_training - Step 47064: {'lr': 4.601533488830356e-06, 'samples': 24097280, 'steps': 47064, 'loss/train': 0.20103920996189117}
+02/26/2022 14:32:49 - INFO - codeparrot_training - Step 47065: {'lr': 4.598409106279261e-06, 'samples': 24097792, 'steps': 47065, 'loss/train': 1.5555360317230225}
+02/26/2022 14:32:55 - INFO - codeparrot_training - Step 47066: {'lr': 4.595285774950636e-06, 'samples': 24098304, 'steps': 47066, 'loss/train': 2.5949933528900146}
+02/26/2022 14:32:58 - INFO - codeparrot_training - Step 47067: {'lr': 4.5921634948579145e-06, 'samples': 24098816, 'steps': 47067, 'loss/train': 1.6819952726364136}
+02/26/2022 14:33:04 - INFO - codeparrot_training - Step 47068: {'lr': 4.589042266014448e-06, 'samples': 24099328, 'steps': 47068, 'loss/train': 1.6068577766418457}
+02/26/2022 14:33:07 - INFO - codeparrot_training - Step 47069: {'lr': 4.585922088433558e-06, 'samples': 24099840, 'steps': 47069, 'loss/train': 1.8582770824432373}
+02/26/2022 14:33:14 - INFO - codeparrot_training - Step 47070: {'lr': 4.5828029621286795e-06, 'samples': 24100352, 'steps': 47070, 'loss/train': 1.2109534740447998}
+02/26/2022 14:33:17 - INFO - codeparrot_training - Step 47071: {'lr': 4.579684887113133e-06, 'samples': 24100864, 'steps': 47071, 'loss/train': 1.5589314699172974}
+02/26/2022 14:33:23 - INFO - codeparrot_training - Step 47072: {'lr': 4.5765678634003e-06, 'samples': 24101376, 'steps': 47072, 'loss/train': 2.1774210929870605}
+02/26/2022 14:33:26 - INFO - codeparrot_training - Step 47073: {'lr': 4.5734518910034995e-06, 'samples': 24101888, 'steps': 47073, 'loss/train': 2.6302433013916016}
+02/26/2022 14:33:32 - INFO - codeparrot_training - Step 47074: {'lr': 4.570336969936112e-06, 'samples': 24102400, 'steps': 47074, 'loss/train': 2.142443895339966}
+02/26/2022 14:33:35 - INFO - codeparrot_training - Step 47075: {'lr': 4.56722310021146e-06, 'samples': 24102912, 'steps': 47075, 'loss/train': 2.084932804107666}
+02/26/2022 14:33:41 - INFO - codeparrot_training - Step 47076: {'lr': 4.564110281842892e-06, 'samples': 24103424, 'steps': 47076, 'loss/train': 1.6592644453048706}
+02/26/2022 14:33:44 - INFO - codeparrot_training - Step 47077: {'lr': 4.560998514843761e-06, 'samples': 24103936, 'steps': 47077, 'loss/train': 0.32203760743141174}
+02/26/2022 14:33:50 - INFO - codeparrot_training - Step 47078: {'lr': 4.557887799227334e-06, 'samples': 24104448, 'steps': 47078, 'loss/train': 0.6809161305427551}
+02/26/2022 14:33:53 - INFO - codeparrot_training - Step 47079: {'lr': 4.554778135007015e-06, 'samples': 24104960, 'steps': 47079, 'loss/train': 1.2344796657562256}
+02/26/2022 14:34:00 - INFO - codeparrot_training - Step 47080: {'lr': 4.551669522196072e-06, 'samples': 24105472, 'steps': 47080, 'loss/train': 1.651402473449707}
+02/26/2022 14:34:04 - INFO - codeparrot_training - Step 47081: {'lr': 4.548561960807857e-06, 'samples': 24105984, 'steps': 47081, 'loss/train': 2.0178136825561523}
+02/26/2022 14:34:09 - INFO - codeparrot_training - Step 47082: {'lr': 4.5454554508556345e-06, 'samples': 24106496, 'steps': 47082, 'loss/train': 1.548479437828064}
+02/26/2022 14:34:12 - INFO - codeparrot_training - Step 47083: {'lr': 4.542349992352785e-06, 'samples': 24107008, 'steps': 47083, 'loss/train': 1.00944983959198}
+02/26/2022 14:34:18 - INFO - codeparrot_training - Step 47084: {'lr': 4.539245585312546e-06, 'samples': 24107520, 'steps': 47084, 'loss/train': 1.1888713836669922}
+02/26/2022 14:34:21 - INFO - codeparrot_training - Step 47085: {'lr': 4.536142229748213e-06, 'samples': 24108032, 'steps': 47085, 'loss/train': 1.8347615003585815}
+02/26/2022 14:34:27 - INFO - codeparrot_training - Step 47086: {'lr': 4.53303992567311e-06, 'samples': 24108544, 'steps': 47086, 'loss/train': 2.052255392074585}
+02/26/2022 14:34:30 - INFO - codeparrot_training - Step 47087: {'lr': 4.529938673100531e-06, 'samples': 24109056, 'steps': 47087, 'loss/train': 2.5584685802459717}
+02/26/2022 14:34:36 - INFO - codeparrot_training - Step 47088: {'lr': 4.526838472043771e-06, 'samples': 24109568, 'steps': 47088, 'loss/train': 4.138465404510498}
+02/26/2022 14:34:39 - INFO - codeparrot_training - Step 47089: {'lr': 4.523739322516041e-06, 'samples': 24110080, 'steps': 47089, 'loss/train': 1.7930083274841309}
+02/26/2022 14:34:45 - INFO - codeparrot_training - Step 47090: {'lr': 4.5206412245306916e-06, 'samples': 24110592, 'steps': 47090, 'loss/train': 2.3545634746551514}
+02/26/2022 14:34:48 - INFO - codeparrot_training - Step 47091: {'lr': 4.517544178100963e-06, 'samples': 24111104, 'steps': 47091, 'loss/train': 1.5640872716903687}
+02/26/2022 14:34:54 - INFO - codeparrot_training - Step 47092: {'lr': 4.5144481832401215e-06, 'samples': 24111616, 'steps': 47092, 'loss/train': 1.0407648086547852}
+02/26/2022 14:34:57 - INFO - codeparrot_training - Step 47093: {'lr': 4.511353239961436e-06, 'samples': 24112128, 'steps': 47093, 'loss/train': 1.559370517730713}
+02/26/2022 14:35:03 - INFO - codeparrot_training - Step 47094: {'lr': 4.508259348278171e-06, 'samples': 24112640, 'steps': 47094, 'loss/train': 1.5072500705718994}
+02/26/2022 14:35:06 - INFO - codeparrot_training - Step 47095: {'lr': 4.505166508203567e-06, 'samples': 24113152, 'steps': 47095, 'loss/train': 2.1189537048339844}
+02/26/2022 14:35:13 - INFO - codeparrot_training - Step 47096: {'lr': 4.502074719750865e-06, 'samples': 24113664, 'steps': 47096, 'loss/train': 1.1228779554367065}
+02/26/2022 14:35:16 - INFO - codeparrot_training - Step 47097: {'lr': 4.4989839829333305e-06, 'samples': 24114176, 'steps': 47097, 'loss/train': 2.733769178390503}
+02/26/2022 14:35:22 - INFO - codeparrot_training - Step 47098: {'lr': 4.495894297764175e-06, 'samples': 24114688, 'steps': 47098, 'loss/train': 1.8858383893966675}
+02/26/2022 14:35:25 - INFO - codeparrot_training - Step 47099: {'lr': 4.492805664256666e-06, 'samples': 24115200, 'steps': 47099, 'loss/train': 1.477362871170044}
+02/26/2022 14:35:31 - INFO - codeparrot_training - Step 47100: {'lr': 4.489718082424044e-06, 'samples': 24115712, 'steps': 47100, 'loss/train': 1.7046114206314087}
+02/26/2022 14:35:34 - INFO - codeparrot_training - Step 47101: {'lr': 4.486631552279463e-06, 'samples': 24116224, 'steps': 47101, 'loss/train': 1.1065185070037842}
+02/26/2022 14:35:40 - INFO - codeparrot_training - Step 47102: {'lr': 4.48354607383622e-06, 'samples': 24116736, 'steps': 47102, 'loss/train': 2.017695426940918}
+02/26/2022 14:35:43 - INFO - codeparrot_training - Step 47103: {'lr': 4.480461647107526e-06, 'samples': 24117248, 'steps': 47103, 'loss/train': 2.986078977584839}
+02/26/2022 14:35:49 - INFO - codeparrot_training - Step 47104: {'lr': 4.477378272106564e-06, 'samples': 24117760, 'steps': 47104, 'loss/train': 1.5272072553634644}
+02/26/2022 14:35:52 - INFO - codeparrot_training - Step 47105: {'lr': 4.474295948846519e-06, 'samples': 24118272, 'steps': 47105, 'loss/train': 1.681247591972351}
+02/26/2022 14:35:58 - INFO - codeparrot_training - Step 47106: {'lr': 4.4712146773406295e-06, 'samples': 24118784, 'steps': 47106, 'loss/train': 1.3817273378372192}
+02/26/2022 14:36:04 - INFO - codeparrot_training - Step 47107: {'lr': 4.468134457602136e-06, 'samples': 24119296, 'steps': 47107, 'loss/train': 1.3703384399414062}
+02/26/2022 14:36:07 - INFO - codeparrot_training - Step 47108: {'lr': 4.465055289644166e-06, 'samples': 24119808, 'steps': 47108, 'loss/train': 1.4463871717453003}
+02/26/2022 14:36:11 - INFO - codeparrot_training - Step 47109: {'lr': 4.461977173479903e-06, 'samples': 24120320, 'steps': 47109, 'loss/train': 1.2517337799072266}
+02/26/2022 14:36:17 - INFO - codeparrot_training - Step 47110: {'lr': 4.458900109122588e-06, 'samples': 24120832, 'steps': 47110, 'loss/train': 0.4688619077205658}
+02/26/2022 14:36:20 - INFO - codeparrot_training - Step 47111: {'lr': 4.455824096585376e-06, 'samples': 24121344, 'steps': 47111, 'loss/train': 1.6020241975784302}
+02/26/2022 14:36:26 - INFO - codeparrot_training - Step 47112: {'lr': 4.452749135881479e-06, 'samples': 24121856, 'steps': 47112, 'loss/train': 2.4513211250305176}
+02/26/2022 14:36:29 - INFO - codeparrot_training - Step 47113: {'lr': 4.449675227023969e-06, 'samples': 24122368, 'steps': 47113, 'loss/train': 1.5038057565689087}
+02/26/2022 14:36:35 - INFO - codeparrot_training - Step 47114: {'lr': 4.446602370026115e-06, 'samples': 24122880, 'steps': 47114, 'loss/train': 1.8735201358795166}
+02/26/2022 14:36:38 - INFO - codeparrot_training - Step 47115: {'lr': 4.443530564901043e-06, 'samples': 24123392, 'steps': 47115, 'loss/train': 2.096792221069336}
+02/26/2022 14:36:45 - INFO - codeparrot_training - Step 47116: {'lr': 4.44045981166194e-06, 'samples': 24123904, 'steps': 47116, 'loss/train': 1.113234043121338}
+02/26/2022 14:36:48 - INFO - codeparrot_training - Step 47117: {'lr': 4.437390110321876e-06, 'samples': 24124416, 'steps': 47117, 'loss/train': 2.3915045261383057}
+02/26/2022 14:36:54 - INFO - codeparrot_training - Step 47118: {'lr': 4.434321460894092e-06, 'samples': 24124928, 'steps': 47118, 'loss/train': 1.185437560081482}
+02/26/2022 14:36:57 - INFO - codeparrot_training - Step 47119: {'lr': 4.431253863391688e-06, 'samples': 24125440, 'steps': 47119, 'loss/train': 1.9552744626998901}
+02/26/2022 14:37:03 - INFO - codeparrot_training - Step 47120: {'lr': 4.4281873178278475e-06, 'samples': 24125952, 'steps': 47120, 'loss/train': 0.7424376010894775}
+02/26/2022 14:37:06 - INFO - codeparrot_training - Step 47121: {'lr': 4.425121824215644e-06, 'samples': 24126464, 'steps': 47121, 'loss/train': 1.4990954399108887}
+02/26/2022 14:37:12 - INFO - codeparrot_training - Step 47122: {'lr': 4.422057382568234e-06, 'samples': 24126976, 'steps': 47122, 'loss/train': 1.594918131828308}
+02/26/2022 14:37:15 - INFO - codeparrot_training - Step 47123: {'lr': 4.418993992898746e-06, 'samples': 24127488, 'steps': 47123, 'loss/train': 2.3720953464508057}
+02/26/2022 14:37:21 - INFO - codeparrot_training - Step 47124: {'lr': 4.415931655220334e-06, 'samples': 24128000, 'steps': 47124, 'loss/train': 2.011507749557495}
+02/26/2022 14:37:24 - INFO - codeparrot_training - Step 47125: {'lr': 4.412870369546074e-06, 'samples': 24128512, 'steps': 47125, 'loss/train': 1.8809527158737183}
+02/26/2022 14:37:31 - INFO - codeparrot_training - Step 47126: {'lr': 4.40981013588912e-06, 'samples': 24129024, 'steps': 47126, 'loss/train': 0.7757999897003174}
+02/26/2022 14:37:34 - INFO - codeparrot_training - Step 47127: {'lr': 4.406750954262517e-06, 'samples': 24129536, 'steps': 47127, 'loss/train': 2.9754371643066406}
+02/26/2022 14:37:39 - INFO - codeparrot_training - Step 47128: {'lr': 4.403692824679423e-06, 'samples': 24130048, 'steps': 47128, 'loss/train': 0.17659859359264374}
+02/26/2022 14:37:43 - INFO - codeparrot_training - Step 47129: {'lr': 4.400635747152937e-06, 'samples': 24130560, 'steps': 47129, 'loss/train': 1.472274899482727}
+02/26/2022 14:37:48 - INFO - codeparrot_training - Step 47130: {'lr': 4.397579721696132e-06, 'samples': 24131072, 'steps': 47130, 'loss/train': 1.6382604837417603}
+02/26/2022 14:37:52 - INFO - codeparrot_training - Step 47131: {'lr': 4.394524748322109e-06, 'samples': 24131584, 'steps': 47131, 'loss/train': 0.8831396102905273}
+02/26/2022 14:37:57 - INFO - codeparrot_training - Step 47132: {'lr': 4.391470827043942e-06, 'samples': 24132096, 'steps': 47132, 'loss/train': 1.23939049243927}
+02/26/2022 14:38:01 - INFO - codeparrot_training - Step 47133: {'lr': 4.388417957874729e-06, 'samples': 24132608, 'steps': 47133, 'loss/train': 2.3350114822387695}
+02/26/2022 14:38:06 - INFO - codeparrot_training - Step 47134: {'lr': 4.385366140827546e-06, 'samples': 24133120, 'steps': 47134, 'loss/train': 1.7037105560302734}
+02/26/2022 14:38:10 - INFO - codeparrot_training - Step 47135: {'lr': 4.382315375915463e-06, 'samples': 24133632, 'steps': 47135, 'loss/train': 1.7490791082382202}
+02/26/2022 14:38:15 - INFO - codeparrot_training - Step 47136: {'lr': 4.379265663151527e-06, 'samples': 24134144, 'steps': 47136, 'loss/train': 0.4786655902862549}
+02/26/2022 14:38:19 - INFO - codeparrot_training - Step 47137: {'lr': 4.376217002548866e-06, 'samples': 24134656, 'steps': 47137, 'loss/train': 2.0586559772491455}
+02/26/2022 14:38:24 - INFO - codeparrot_training - Step 47138: {'lr': 4.373169394120469e-06, 'samples': 24135168, 'steps': 47138, 'loss/train': 1.3500629663467407}
+02/26/2022 14:38:28 - INFO - codeparrot_training - Step 47139: {'lr': 4.370122837879437e-06, 'samples': 24135680, 'steps': 47139, 'loss/train': 0.6475369334220886}
+02/26/2022 14:38:33 - INFO - codeparrot_training - Step 47140: {'lr': 4.367077333838787e-06, 'samples': 24136192, 'steps': 47140, 'loss/train': 2.9690399169921875}
+02/26/2022 14:38:37 - INFO - codeparrot_training - Step 47141: {'lr': 4.364032882011593e-06, 'samples': 24136704, 'steps': 47141, 'loss/train': 1.6853876113891602}
+02/26/2022 14:38:43 - INFO - codeparrot_training - Step 47142: {'lr': 4.360989482410871e-06, 'samples': 24137216, 'steps': 47142, 'loss/train': 1.6282105445861816}
+02/26/2022 14:38:46 - INFO - codeparrot_training - Step 47143: {'lr': 4.357947135049667e-06, 'samples': 24137728, 'steps': 47143, 'loss/train': 2.1060256958007812}
+02/26/2022 14:38:52 - INFO - codeparrot_training - Step 47144: {'lr': 4.354905839941026e-06, 'samples': 24138240, 'steps': 47144, 'loss/train': 1.1118172407150269}
+02/26/2022 14:38:55 - INFO - codeparrot_training - Step 47145: {'lr': 4.351865597097965e-06, 'samples': 24138752, 'steps': 47145, 'loss/train': 1.6369283199310303}
+02/26/2022 14:39:01 - INFO - codeparrot_training - Step 47146: {'lr': 4.348826406533501e-06, 'samples': 24139264, 'steps': 47146, 'loss/train': 1.6946613788604736}
+02/26/2022 14:39:04 - INFO - codeparrot_training - Step 47147: {'lr': 4.3457882682607085e-06, 'samples': 24139776, 'steps': 47147, 'loss/train': 1.0469201803207397}
+02/26/2022 14:39:10 - INFO - codeparrot_training - Step 47148: {'lr': 4.342751182292493e-06, 'samples': 24140288, 'steps': 47148, 'loss/train': 1.8560876846313477}
+02/26/2022 14:39:13 - INFO - codeparrot_training - Step 47149: {'lr': 4.339715148641954e-06, 'samples': 24140800, 'steps': 47149, 'loss/train': 3.023561477661133}
+02/26/2022 14:39:19 - INFO - codeparrot_training - Step 47150: {'lr': 4.336680167322055e-06, 'samples': 24141312, 'steps': 47150, 'loss/train': 1.7399812936782837}
+02/26/2022 14:39:22 - INFO - codeparrot_training - Step 47151: {'lr': 4.333646238345868e-06, 'samples': 24141824, 'steps': 47151, 'loss/train': 1.856559157371521}
+02/26/2022 14:39:29 - INFO - codeparrot_training - Step 47152: {'lr': 4.330613361726271e-06, 'samples': 24142336, 'steps': 47152, 'loss/train': 1.5507713556289673}
+02/26/2022 14:39:32 - INFO - codeparrot_training - Step 47153: {'lr': 4.327581537476338e-06, 'samples': 24142848, 'steps': 47153, 'loss/train': 1.2864854335784912}
+02/26/2022 14:39:38 - INFO - codeparrot_training - Step 47154: {'lr': 4.324550765609003e-06, 'samples': 24143360, 'steps': 47154, 'loss/train': 2.471898078918457}
+02/26/2022 14:39:41 - INFO - codeparrot_training - Step 47155: {'lr': 4.321521046137339e-06, 'samples': 24143872, 'steps': 47155, 'loss/train': 2.5953712463378906}
+02/26/2022 14:39:47 - INFO - codeparrot_training - Step 47156: {'lr': 4.318492379074224e-06, 'samples': 24144384, 'steps': 47156, 'loss/train': 1.9046624898910522}
+02/26/2022 14:39:50 - INFO - codeparrot_training - Step 47157: {'lr': 4.315464764432703e-06, 'samples': 24144896, 'steps': 47157, 'loss/train': 1.9269174337387085}
+02/26/2022 14:39:56 - INFO - codeparrot_training - Step 47158: {'lr': 4.3124382022256825e-06, 'samples': 24145408, 'steps': 47158, 'loss/train': 1.6814074516296387}
+02/26/2022 14:39:59 - INFO - codeparrot_training - Step 47159: {'lr': 4.309412692466208e-06, 'samples': 24145920, 'steps': 47159, 'loss/train': 1.686313271522522}
+02/26/2022 14:40:05 - INFO - codeparrot_training - Step 47160: {'lr': 4.306388235167158e-06, 'samples': 24146432, 'steps': 47160, 'loss/train': 1.508376955986023}
+02/26/2022 14:40:09 - INFO - codeparrot_training - Step 47161: {'lr': 4.30336483034155e-06, 'samples': 24146944, 'steps': 47161, 'loss/train': 1.4307645559310913}
+02/26/2022 14:40:12 - INFO - codeparrot_training - Step 47162: {'lr': 4.300342478002289e-06, 'samples': 24147456, 'steps': 47162, 'loss/train': 1.6573303937911987}
+02/26/2022 14:40:18 - INFO - codeparrot_training - Step 47163: {'lr': 4.297321178162339e-06, 'samples': 24147968, 'steps': 47163, 'loss/train': 2.3991005420684814}
+02/26/2022 14:40:22 - INFO - codeparrot_training - Step 47164: {'lr': 4.294300930834688e-06, 'samples': 24148480, 'steps': 47164, 'loss/train': 1.4439725875854492}
+02/26/2022 14:40:27 - INFO - codeparrot_training - Step 47165: {'lr': 4.291281736032188e-06, 'samples': 24148992, 'steps': 47165, 'loss/train': 2.271268367767334}
+02/26/2022 14:40:31 - INFO - codeparrot_training - Step 47166: {'lr': 4.288263593767827e-06, 'samples': 24149504, 'steps': 47166, 'loss/train': 2.306211233139038}
+02/26/2022 14:40:36 - INFO - codeparrot_training - Step 47167: {'lr': 4.285246504054513e-06, 'samples': 24150016, 'steps': 47167, 'loss/train': 1.6509721279144287}
+02/26/2022 14:40:40 - INFO - codeparrot_training - Step 47168: {'lr': 4.282230466905207e-06, 'samples': 24150528, 'steps': 47168, 'loss/train': 1.5485646724700928}
+02/26/2022 14:40:45 - INFO - codeparrot_training - Step 47169: {'lr': 4.279215482332788e-06, 'samples': 24151040, 'steps': 47169, 'loss/train': 1.28037428855896}
+02/26/2022 14:40:49 - INFO - codeparrot_training - Step 47170: {'lr': 4.2762015503501614e-06, 'samples': 24151552, 'steps': 47170, 'loss/train': 1.7375378608703613}
+02/26/2022 14:40:54 - INFO - codeparrot_training - Step 47171: {'lr': 4.2731886709702904e-06, 'samples': 24152064, 'steps': 47171, 'loss/train': 2.593419313430786}
+02/26/2022 14:40:58 - INFO - codeparrot_training - Step 47172: {'lr': 4.270176844206053e-06, 'samples': 24152576, 'steps': 47172, 'loss/train': 0.46203529834747314}
+02/26/2022 14:41:04 - INFO - codeparrot_training - Step 47173: {'lr': 4.267166070070355e-06, 'samples': 24153088, 'steps': 47173, 'loss/train': 1.961051344871521}
+02/26/2022 14:41:07 - INFO - codeparrot_training - Step 47174: {'lr': 4.264156348576048e-06, 'samples': 24153600, 'steps': 47174, 'loss/train': 1.4205868244171143}
+02/26/2022 14:41:13 - INFO - codeparrot_training - Step 47175: {'lr': 4.261147679736094e-06, 'samples': 24154112, 'steps': 47175, 'loss/train': 2.190095901489258}
+02/26/2022 14:41:16 - INFO - codeparrot_training - Step 47176: {'lr': 4.258140063563343e-06, 'samples': 24154624, 'steps': 47176, 'loss/train': 1.9713971614837646}
+02/26/2022 14:41:22 - INFO - codeparrot_training - Step 47177: {'lr': 4.255133500070701e-06, 'samples': 24155136, 'steps': 47177, 'loss/train': 3.0334203243255615}
+02/26/2022 14:41:25 - INFO - codeparrot_training - Step 47178: {'lr': 4.252127989271048e-06, 'samples': 24155648, 'steps': 47178, 'loss/train': 0.36063963174819946}
+02/26/2022 14:41:31 - INFO - codeparrot_training - Step 47179: {'lr': 4.249123531177207e-06, 'samples': 24156160, 'steps': 47179, 'loss/train': 1.4529038667678833}
+02/26/2022 14:41:35 - INFO - codeparrot_training - Step 47180: {'lr': 4.246120125802111e-06, 'samples': 24156672, 'steps': 47180, 'loss/train': 0.04210694879293442}
+02/26/2022 14:41:40 - INFO - codeparrot_training - Step 47181: {'lr': 4.243117773158611e-06, 'samples': 24157184, 'steps': 47181, 'loss/train': 0.3954634368419647}
+02/26/2022 14:41:44 - INFO - codeparrot_training - Step 47182: {'lr': 4.240116473259531e-06, 'samples': 24157696, 'steps': 47182, 'loss/train': 1.2419888973236084}
+02/26/2022 14:41:49 - INFO - codeparrot_training - Step 47183: {'lr': 4.2371162261177484e-06, 'samples': 24158208, 'steps': 47183, 'loss/train': 1.7005199193954468}
+02/26/2022 14:41:53 - INFO - codeparrot_training - Step 47184: {'lr': 4.234117031746143e-06, 'samples': 24158720, 'steps': 47184, 'loss/train': 0.935300350189209}
+02/26/2022 14:41:58 - INFO - codeparrot_training - Step 47185: {'lr': 4.231118890157509e-06, 'samples': 24159232, 'steps': 47185, 'loss/train': 3.2207820415496826}
+02/26/2022 14:42:02 - INFO - codeparrot_training - Step 47186: {'lr': 4.2281218013647815e-06, 'samples': 24159744, 'steps': 47186, 'loss/train': 1.2835636138916016}
+02/26/2022 14:42:07 - INFO - codeparrot_training - Step 47187: {'lr': 4.2251257653806996e-06, 'samples': 24160256, 'steps': 47187, 'loss/train': 1.5822453498840332}
+02/26/2022 14:42:11 - INFO - codeparrot_training - Step 47188: {'lr': 4.222130782218142e-06, 'samples': 24160768, 'steps': 47188, 'loss/train': 3.468564510345459}
+02/26/2022 14:42:17 - INFO - codeparrot_training - Step 47189: {'lr': 4.219136851889905e-06, 'samples': 24161280, 'steps': 47189, 'loss/train': 2.1898415088653564}
+02/26/2022 14:42:20 - INFO - codeparrot_training - Step 47190: {'lr': 4.216143974408892e-06, 'samples': 24161792, 'steps': 47190, 'loss/train': 0.965121865272522}
+02/26/2022 14:42:26 - INFO - codeparrot_training - Step 47191: {'lr': 4.213152149787847e-06, 'samples': 24162304, 'steps': 47191, 'loss/train': 2.138892650604248}
+02/26/2022 14:42:29 - INFO - codeparrot_training - Step 47192: {'lr': 4.210161378039618e-06, 'samples': 24162816, 'steps': 47192, 'loss/train': 1.96351957321167}
+02/26/2022 14:42:35 - INFO - codeparrot_training - Step 47193: {'lr': 4.207171659177e-06, 'samples': 24163328, 'steps': 47193, 'loss/train': 2.7734086513519287}
+02/26/2022 14:42:38 - INFO - codeparrot_training - Step 47194: {'lr': 4.204182993212819e-06, 'samples': 24163840, 'steps': 47194, 'loss/train': 1.9226479530334473}
+02/26/2022 14:42:44 - INFO - codeparrot_training - Step 47195: {'lr': 4.201195380159867e-06, 'samples': 24164352, 'steps': 47195, 'loss/train': 1.737646460533142}
+02/26/2022 14:42:47 - INFO - codeparrot_training - Step 47196: {'lr': 4.1982088200309685e-06, 'samples': 24164864, 'steps': 47196, 'loss/train': 1.0406370162963867}
+02/26/2022 14:42:53 - INFO - codeparrot_training - Step 47197: {'lr': 4.195223312838836e-06, 'samples': 24165376, 'steps': 47197, 'loss/train': 1.7770664691925049}
+02/26/2022 14:42:59 - INFO - codeparrot_training - Step 47198: {'lr': 4.192238858596375e-06, 'samples': 24165888, 'steps': 47198, 'loss/train': 1.632149338722229}
+02/26/2022 14:43:03 - INFO - codeparrot_training - Step 47199: {'lr': 4.1892554573162715e-06, 'samples': 24166400, 'steps': 47199, 'loss/train': 1.4625093936920166}
+02/26/2022 14:43:08 - INFO - codeparrot_training - Step 47200: {'lr': 4.186273109011374e-06, 'samples': 24166912, 'steps': 47200, 'loss/train': 2.194828987121582}
+02/26/2022 14:43:11 - INFO - codeparrot_training - Step 47201: {'lr': 4.183291813694395e-06, 'samples': 24167424, 'steps': 47201, 'loss/train': 1.5172010660171509}
+02/26/2022 14:43:17 - INFO - codeparrot_training - Step 47202: {'lr': 4.180311571378159e-06, 'samples': 24167936, 'steps': 47202, 'loss/train': 1.6183710098266602}
+02/26/2022 14:43:21 - INFO - codeparrot_training - Step 47203: {'lr': 4.177332382075405e-06, 'samples': 24168448, 'steps': 47203, 'loss/train': 1.1115070581436157}
+02/26/2022 14:43:24 - INFO - codeparrot_training - Step 47204: {'lr': 4.1743542457989005e-06, 'samples': 24168960, 'steps': 47204, 'loss/train': 1.5006262063980103}
+02/26/2022 14:43:30 - INFO - codeparrot_training - Step 47205: {'lr': 4.171377162561385e-06, 'samples': 24169472, 'steps': 47205, 'loss/train': 2.4143946170806885}
+02/26/2022 14:43:33 - INFO - codeparrot_training - Step 47206: {'lr': 4.168401132375654e-06, 'samples': 24169984, 'steps': 47206, 'loss/train': 0.43767157196998596}
+02/26/2022 14:43:39 - INFO - codeparrot_training - Step 47207: {'lr': 4.16542615525442e-06, 'samples': 24170496, 'steps': 47207, 'loss/train': 1.577985167503357}
+02/26/2022 14:43:42 - INFO - codeparrot_training - Step 47208: {'lr': 4.162452231210479e-06, 'samples': 24171008, 'steps': 47208, 'loss/train': 1.830161690711975}
+02/26/2022 14:43:49 - INFO - codeparrot_training - Step 47209: {'lr': 4.159479360256485e-06, 'samples': 24171520, 'steps': 47209, 'loss/train': 0.8036623597145081}
+02/26/2022 14:43:53 - INFO - codeparrot_training - Step 47210: {'lr': 4.156507542405235e-06, 'samples': 24172032, 'steps': 47210, 'loss/train': 0.3942053020000458}
+02/26/2022 14:43:58 - INFO - codeparrot_training - Step 47211: {'lr': 4.153536777669442e-06, 'samples': 24172544, 'steps': 47211, 'loss/train': 1.0266683101654053}
+02/26/2022 14:44:02 - INFO - codeparrot_training - Step 47212: {'lr': 4.150567066061845e-06, 'samples': 24173056, 'steps': 47212, 'loss/train': 0.4471285045146942}
+02/26/2022 14:44:07 - INFO - codeparrot_training - Step 47213: {'lr': 4.147598407595127e-06, 'samples': 24173568, 'steps': 47213, 'loss/train': 2.4111576080322266}
+02/26/2022 14:44:11 - INFO - codeparrot_training - Step 47214: {'lr': 4.144630802282057e-06, 'samples': 24174080, 'steps': 47214, 'loss/train': 1.5770294666290283}
+02/26/2022 14:44:16 - INFO - codeparrot_training - Step 47215: {'lr': 4.141664250135291e-06, 'samples': 24174592, 'steps': 47215, 'loss/train': 2.1797947883605957}
+02/26/2022 14:44:20 - INFO - codeparrot_training - Step 47216: {'lr': 4.138698751167597e-06, 'samples': 24175104, 'steps': 47216, 'loss/train': 1.363838791847229}
+02/26/2022 14:44:25 - INFO - codeparrot_training - Step 47217: {'lr': 4.135734305391603e-06, 'samples': 24175616, 'steps': 47217, 'loss/train': 1.8961713314056396}
+02/26/2022 14:44:29 - INFO - codeparrot_training - Step 47218: {'lr': 4.132770912820105e-06, 'samples': 24176128, 'steps': 47218, 'loss/train': 1.6389878988265991}
+02/26/2022 14:44:34 - INFO - codeparrot_training - Step 47219: {'lr': 4.129808573465732e-06, 'samples': 24176640, 'steps': 47219, 'loss/train': 1.65084969997406}
+02/26/2022 14:44:38 - INFO - codeparrot_training - Step 47220: {'lr': 4.12684728734114e-06, 'samples': 24177152, 'steps': 47220, 'loss/train': 1.2980422973632812}
+02/26/2022 14:44:44 - INFO - codeparrot_training - Step 47221: {'lr': 4.123887054459124e-06, 'samples': 24177664, 'steps': 47221, 'loss/train': 1.327880859375}
+02/26/2022 14:44:47 - INFO - codeparrot_training - Step 47222: {'lr': 4.120927874832259e-06, 'samples': 24178176, 'steps': 47222, 'loss/train': 0.8891405463218689}
+02/26/2022 14:44:53 - INFO - codeparrot_training - Step 47223: {'lr': 4.117969748473282e-06, 'samples': 24178688, 'steps': 47223, 'loss/train': 1.5660713911056519}
+02/26/2022 14:44:58 - INFO - codeparrot_training - Step 47224: {'lr': 4.115012675394825e-06, 'samples': 24179200, 'steps': 47224, 'loss/train': 1.7935417890548706}
+02/26/2022 14:45:02 - INFO - codeparrot_training - Step 47225: {'lr': 4.112056655609625e-06, 'samples': 24179712, 'steps': 47225, 'loss/train': 2.003937244415283}
+02/26/2022 14:45:07 - INFO - codeparrot_training - Step 47226: {'lr': 4.1091016891302566e-06, 'samples': 24180224, 'steps': 47226, 'loss/train': 0.07567253708839417}
+02/26/2022 14:45:11 - INFO - codeparrot_training - Step 47227: {'lr': 4.106147775969432e-06, 'samples': 24180736, 'steps': 47227, 'loss/train': 1.8219544887542725}
+02/26/2022 14:45:17 - INFO - codeparrot_training - Step 47228: {'lr': 4.10319491613978e-06, 'samples': 24181248, 'steps': 47228, 'loss/train': 0.9790166616439819}
+02/26/2022 14:45:20 - INFO - codeparrot_training - Step 47229: {'lr': 4.100243109653984e-06, 'samples': 24181760, 'steps': 47229, 'loss/train': 1.9543795585632324}
+02/26/2022 14:45:26 - INFO - codeparrot_training - Step 47230: {'lr': 4.097292356524618e-06, 'samples': 24182272, 'steps': 47230, 'loss/train': 1.9397372007369995}
+02/26/2022 14:45:29 - INFO - codeparrot_training - Step 47231: {'lr': 4.0943426567644215e-06, 'samples': 24182784, 'steps': 47231, 'loss/train': 2.0798254013061523}
+02/26/2022 14:45:35 - INFO - codeparrot_training - Step 47232: {'lr': 4.091394010385913e-06, 'samples': 24183296, 'steps': 47232, 'loss/train': 2.189432144165039}
+02/26/2022 14:45:38 - INFO - codeparrot_training - Step 47233: {'lr': 4.088446417401831e-06, 'samples': 24183808, 'steps': 47233, 'loss/train': 2.1390795707702637}
+02/26/2022 14:45:44 - INFO - codeparrot_training - Step 47234: {'lr': 4.0854998778247776e-06, 'samples': 24184320, 'steps': 47234, 'loss/train': 2.073695421218872}
+02/26/2022 14:45:48 - INFO - codeparrot_training - Step 47235: {'lr': 4.0825543916673255e-06, 'samples': 24184832, 'steps': 47235, 'loss/train': 1.6629016399383545}
+02/26/2022 14:45:53 - INFO - codeparrot_training - Step 47236: {'lr': 4.079609958942132e-06, 'samples': 24185344, 'steps': 47236, 'loss/train': 2.354749917984009}
+02/26/2022 14:45:57 - INFO - codeparrot_training - Step 47237: {'lr': 4.076666579661797e-06, 'samples': 24185856, 'steps': 47237, 'loss/train': 1.5385057926177979}
+02/26/2022 14:46:02 - INFO - codeparrot_training - Step 47238: {'lr': 4.073724253838923e-06, 'samples': 24186368, 'steps': 47238, 'loss/train': 0.4567966163158417}
+02/26/2022 14:46:06 - INFO - codeparrot_training - Step 47239: {'lr': 4.070782981486166e-06, 'samples': 24186880, 'steps': 47239, 'loss/train': 2.5535647869110107}
+02/26/2022 14:46:11 - INFO - codeparrot_training - Step 47240: {'lr': 4.067842762616014e-06, 'samples': 24187392, 'steps': 47240, 'loss/train': 2.1779215335845947}
+02/26/2022 14:46:15 - INFO - codeparrot_training - Step 47241: {'lr': 4.064903597241182e-06, 'samples': 24187904, 'steps': 47241, 'loss/train': 1.413515567779541}
+02/26/2022 14:46:20 - INFO - codeparrot_training - Step 47242: {'lr': 4.061965485374186e-06, 'samples': 24188416, 'steps': 47242, 'loss/train': 2.044708728790283}
+02/26/2022 14:46:24 - INFO - codeparrot_training - Step 47243: {'lr': 4.059028427027683e-06, 'samples': 24188928, 'steps': 47243, 'loss/train': 1.5489630699157715}
+02/26/2022 14:46:27 - INFO - codeparrot_training - Step 47244: {'lr': 4.056092422214164e-06, 'samples': 24189440, 'steps': 47244, 'loss/train': 1.5935860872268677}
+02/26/2022 14:46:33 - INFO - codeparrot_training - Step 47245: {'lr': 4.053157470946256e-06, 'samples': 24189952, 'steps': 47245, 'loss/train': 3.0423436164855957}
+02/26/2022 14:46:36 - INFO - codeparrot_training - Step 47246: {'lr': 4.050223573236506e-06, 'samples': 24190464, 'steps': 47246, 'loss/train': 2.2079644203186035}
+02/26/2022 14:46:42 - INFO - codeparrot_training - Step 47247: {'lr': 4.047290729097542e-06, 'samples': 24190976, 'steps': 47247, 'loss/train': 0.03209612891077995}
+02/26/2022 14:46:46 - INFO - codeparrot_training - Step 47248: {'lr': 4.0443589385418534e-06, 'samples': 24191488, 'steps': 47248, 'loss/train': 2.941185235977173}
+02/26/2022 14:46:51 - INFO - codeparrot_training - Step 47249: {'lr': 4.041428201582042e-06, 'samples': 24192000, 'steps': 47249, 'loss/train': 1.6676666736602783}
+02/26/2022 14:46:55 - INFO - codeparrot_training - Step 47250: {'lr': 4.038498518230627e-06, 'samples': 24192512, 'steps': 47250, 'loss/train': 2.11346173286438}
+02/26/2022 14:47:01 - INFO - codeparrot_training - Step 47251: {'lr': 4.035569888500235e-06, 'samples': 24193024, 'steps': 47251, 'loss/train': 1.7145932912826538}
+02/26/2022 14:47:06 - INFO - codeparrot_training - Step 47252: {'lr': 4.032642312403329e-06, 'samples': 24193536, 'steps': 47252, 'loss/train': 1.607780933380127}
+02/26/2022 14:47:09 - INFO - codeparrot_training - Step 47253: {'lr': 4.029715789952482e-06, 'samples': 24194048, 'steps': 47253, 'loss/train': 1.9386937618255615}
+02/26/2022 14:47:15 - INFO - codeparrot_training - Step 47254: {'lr': 4.026790321160212e-06, 'samples': 24194560, 'steps': 47254, 'loss/train': 1.0495073795318604}
+02/26/2022 14:47:18 - INFO - codeparrot_training - Step 47255: {'lr': 4.023865906039093e-06, 'samples': 24195072, 'steps': 47255, 'loss/train': 1.0093233585357666}
+02/26/2022 14:47:25 - INFO - codeparrot_training - Step 47256: {'lr': 4.020942544601614e-06, 'samples': 24195584, 'steps': 47256, 'loss/train': 1.9473412036895752}
+02/26/2022 14:47:28 - INFO - codeparrot_training - Step 47257: {'lr': 4.01802023686032e-06, 'samples': 24196096, 'steps': 47257, 'loss/train': 1.7149568796157837}
+02/26/2022 14:47:34 - INFO - codeparrot_training - Step 47258: {'lr': 4.0150989828277306e-06, 'samples': 24196608, 'steps': 47258, 'loss/train': 1.9411085844039917}
+02/26/2022 14:47:37 - INFO - codeparrot_training - Step 47259: {'lr': 4.012178782516307e-06, 'samples': 24197120, 'steps': 47259, 'loss/train': 2.141460657119751}
+02/26/2022 14:47:43 - INFO - codeparrot_training - Step 47260: {'lr': 4.009259635938623e-06, 'samples': 24197632, 'steps': 47260, 'loss/train': 2.2508716583251953}
+02/26/2022 14:47:46 - INFO - codeparrot_training - Step 47261: {'lr': 4.006341543107167e-06, 'samples': 24198144, 'steps': 47261, 'loss/train': 2.378004789352417}
+02/26/2022 14:47:52 - INFO - codeparrot_training - Step 47262: {'lr': 4.003424504034431e-06, 'samples': 24198656, 'steps': 47262, 'loss/train': 1.7187739610671997}
+02/26/2022 14:47:55 - INFO - codeparrot_training - Step 47263: {'lr': 4.000508518732876e-06, 'samples': 24199168, 'steps': 47263, 'loss/train': 1.439517855644226}
+02/26/2022 14:48:01 - INFO - codeparrot_training - Step 47264: {'lr': 3.997593587215076e-06, 'samples': 24199680, 'steps': 47264, 'loss/train': 1.0263742208480835}
+02/26/2022 14:48:04 - INFO - codeparrot_training - Step 47265: {'lr': 3.994679709493437e-06, 'samples': 24200192, 'steps': 47265, 'loss/train': 1.2490408420562744}
+02/26/2022 14:48:10 - INFO - codeparrot_training - Step 47266: {'lr': 3.991766885580505e-06, 'samples': 24200704, 'steps': 47266, 'loss/train': 1.748345136642456}
+02/26/2022 14:48:13 - INFO - codeparrot_training - Step 47267: {'lr': 3.988855115488688e-06, 'samples': 24201216, 'steps': 47267, 'loss/train': 1.529651165008545}
+02/26/2022 14:48:19 - INFO - codeparrot_training - Step 47268: {'lr': 3.9859443992305285e-06, 'samples': 24201728, 'steps': 47268, 'loss/train': 2.4793283939361572}
+02/26/2022 14:48:22 - INFO - codeparrot_training - Step 47269: {'lr': 3.983034736818464e-06, 'samples': 24202240, 'steps': 47269, 'loss/train': 1.4896708726882935}
+02/26/2022 14:48:28 - INFO - codeparrot_training - Step 47270: {'lr': 3.980126128264955e-06, 'samples': 24202752, 'steps': 47270, 'loss/train': 2.1946334838867188}
+02/26/2022 14:48:32 - INFO - codeparrot_training - Step 47271: {'lr': 3.977218573582464e-06, 'samples': 24203264, 'steps': 47271, 'loss/train': 2.0485382080078125}
+02/26/2022 14:48:37 - INFO - codeparrot_training - Step 47272: {'lr': 3.974312072783454e-06, 'samples': 24203776, 'steps': 47272, 'loss/train': 2.103181838989258}
+02/26/2022 14:48:41 - INFO - codeparrot_training - Step 47273: {'lr': 3.971406625880358e-06, 'samples': 24204288, 'steps': 47273, 'loss/train': 2.0182337760925293}
+02/26/2022 14:48:46 - INFO - codeparrot_training - Step 47274: {'lr': 3.9685022328856405e-06, 'samples': 24204800, 'steps': 47274, 'loss/train': 1.8431813716888428}
+02/26/2022 14:48:50 - INFO - codeparrot_training - Step 47275: {'lr': 3.965598893811734e-06, 'samples': 24205312, 'steps': 47275, 'loss/train': 1.9366247653961182}
+02/26/2022 14:48:55 - INFO - codeparrot_training - Step 47276: {'lr': 3.9626966086710735e-06, 'samples': 24205824, 'steps': 47276, 'loss/train': 2.3885579109191895}
+02/26/2022 14:48:59 - INFO - codeparrot_training - Step 47277: {'lr': 3.959795377476095e-06, 'samples': 24206336, 'steps': 47277, 'loss/train': 1.496716022491455}
+02/26/2022 14:49:04 - INFO - codeparrot_training - Step 47278: {'lr': 3.956895200239258e-06, 'samples': 24206848, 'steps': 47278, 'loss/train': 2.5304102897644043}
+02/26/2022 14:49:08 - INFO - codeparrot_training - Step 47279: {'lr': 3.953996076972888e-06, 'samples': 24207360, 'steps': 47279, 'loss/train': 1.0460617542266846}
+02/26/2022 14:49:13 - INFO - codeparrot_training - Step 47280: {'lr': 3.95109800768953e-06, 'samples': 24207872, 'steps': 47280, 'loss/train': 2.0970051288604736}
+02/26/2022 14:49:17 - INFO - codeparrot_training - Step 47281: {'lr': 3.948200992401507e-06, 'samples': 24208384, 'steps': 47281, 'loss/train': 1.5946303606033325}
+02/26/2022 14:49:23 - INFO - codeparrot_training - Step 47282: {'lr': 3.945305031121283e-06, 'samples': 24208896, 'steps': 47282, 'loss/train': 2.0065109729766846}
+02/26/2022 14:49:26 - INFO - codeparrot_training - Step 47283: {'lr': 3.9424101238612065e-06, 'samples': 24209408, 'steps': 47283, 'loss/train': 0.2687365710735321}
+02/26/2022 14:49:32 - INFO - codeparrot_training - Step 47284: {'lr': 3.939516270633742e-06, 'samples': 24209920, 'steps': 47284, 'loss/train': 2.01710844039917}
+02/26/2022 14:49:35 - INFO - codeparrot_training - Step 47285: {'lr': 3.93662347145124e-06, 'samples': 24210432, 'steps': 47285, 'loss/train': 2.6557397842407227}
+02/26/2022 14:49:41 - INFO - codeparrot_training - Step 47286: {'lr': 3.933731726326162e-06, 'samples': 24210944, 'steps': 47286, 'loss/train': 2.4320626258850098}
+02/26/2022 14:49:44 - INFO - codeparrot_training - Step 47287: {'lr': 3.9308410352707765e-06, 'samples': 24211456, 'steps': 47287, 'loss/train': 1.5946294069290161}
+02/26/2022 14:49:50 - INFO - codeparrot_training - Step 47288: {'lr': 3.927951398297547e-06, 'samples': 24211968, 'steps': 47288, 'loss/train': 1.4650859832763672}
+02/26/2022 14:49:53 - INFO - codeparrot_training - Step 47289: {'lr': 3.92506281541885e-06, 'samples': 24212480, 'steps': 47289, 'loss/train': 1.8791393041610718}
+02/26/2022 14:49:59 - INFO - codeparrot_training - Step 47290: {'lr': 3.922175286647067e-06, 'samples': 24212992, 'steps': 47290, 'loss/train': 2.2669014930725098}
+02/26/2022 14:50:02 - INFO - codeparrot_training - Step 47291: {'lr': 3.919288811994492e-06, 'samples': 24213504, 'steps': 47291, 'loss/train': 2.48101806640625}
+02/26/2022 14:50:09 - INFO - codeparrot_training - Step 47292: {'lr': 3.916403391473588e-06, 'samples': 24214016, 'steps': 47292, 'loss/train': 1.8488614559173584}
+02/26/2022 14:50:12 - INFO - codeparrot_training - Step 47293: {'lr': 3.913519025096651e-06, 'samples': 24214528, 'steps': 47293, 'loss/train': 1.4907691478729248}
+02/26/2022 14:50:17 - INFO - codeparrot_training - Step 47294: {'lr': 3.910635712876059e-06, 'samples': 24215040, 'steps': 47294, 'loss/train': 0.8014967441558838}
+02/26/2022 14:50:21 - INFO - codeparrot_training - Step 47295: {'lr': 3.907753454824165e-06, 'samples': 24215552, 'steps': 47295, 'loss/train': 1.6446328163146973}
+02/26/2022 14:50:26 - INFO - codeparrot_training - Step 47296: {'lr': 3.904872250953317e-06, 'samples': 24216064, 'steps': 47296, 'loss/train': 1.9052540063858032}
+02/26/2022 14:50:30 - INFO - codeparrot_training - Step 47297: {'lr': 3.901992101275842e-06, 'samples': 24216576, 'steps': 47297, 'loss/train': 2.2618050575256348}
+02/26/2022 14:50:36 - INFO - codeparrot_training - Step 47298: {'lr': 3.899113005804089e-06, 'samples': 24217088, 'steps': 47298, 'loss/train': 2.1732382774353027}
+02/26/2022 14:50:39 - INFO - codeparrot_training - Step 47299: {'lr': 3.89623496455041e-06, 'samples': 24217600, 'steps': 47299, 'loss/train': 2.699371814727783}
+02/26/2022 14:50:45 - INFO - codeparrot_training - Step 47300: {'lr': 3.893357977527101e-06, 'samples': 24218112, 'steps': 47300, 'loss/train': 1.1672178506851196}
+02/26/2022 14:50:48 - INFO - codeparrot_training - Step 47301: {'lr': 3.890482044746485e-06, 'samples': 24218624, 'steps': 47301, 'loss/train': 1.9867523908615112}
+02/26/2022 14:50:54 - INFO - codeparrot_training - Step 47302: {'lr': 3.887607166220914e-06, 'samples': 24219136, 'steps': 47302, 'loss/train': 2.301151752471924}
+02/26/2022 14:50:58 - INFO - codeparrot_training - Step 47303: {'lr': 3.884733341962682e-06, 'samples': 24219648, 'steps': 47303, 'loss/train': 1.9328258037567139}
+02/26/2022 14:51:03 - INFO - codeparrot_training - Step 47304: {'lr': 3.881860571984086e-06, 'samples': 24220160, 'steps': 47304, 'loss/train': 2.714582681655884}
+02/26/2022 14:51:06 - INFO - codeparrot_training - Step 47305: {'lr': 3.87898885629745e-06, 'samples': 24220672, 'steps': 47305, 'loss/train': 1.4150865077972412}
+02/26/2022 14:51:12 - INFO - codeparrot_training - Step 47306: {'lr': 3.876118194915096e-06, 'samples': 24221184, 'steps': 47306, 'loss/train': 1.7300724983215332}
+02/26/2022 14:51:15 - INFO - codeparrot_training - Step 47307: {'lr': 3.873248587849293e-06, 'samples': 24221696, 'steps': 47307, 'loss/train': 1.9061883687973022}
+02/26/2022 14:51:21 - INFO - codeparrot_training - Step 47308: {'lr': 3.870380035112336e-06, 'samples': 24222208, 'steps': 47308, 'loss/train': 1.6285836696624756}
+02/26/2022 14:51:24 - INFO - codeparrot_training - Step 47309: {'lr': 3.867512536716522e-06, 'samples': 24222720, 'steps': 47309, 'loss/train': 2.1681153774261475}
+02/26/2022 14:51:30 - INFO - codeparrot_training - Step 47310: {'lr': 3.86464609267409e-06, 'samples': 24223232, 'steps': 47310, 'loss/train': 2.4143123626708984}
+02/26/2022 14:51:33 - INFO - codeparrot_training - Step 47311: {'lr': 3.861780702997392e-06, 'samples': 24223744, 'steps': 47311, 'loss/train': 0.46697837114334106}
+02/26/2022 14:51:40 - INFO - codeparrot_training - Step 47312: {'lr': 3.858916367698667e-06, 'samples': 24224256, 'steps': 47312, 'loss/train': 2.148773193359375}
+02/26/2022 14:51:43 - INFO - codeparrot_training - Step 47313: {'lr': 3.856053086790184e-06, 'samples': 24224768, 'steps': 47313, 'loss/train': 1.2401666641235352}
+02/26/2022 14:51:49 - INFO - codeparrot_training - Step 47314: {'lr': 3.853190860284184e-06, 'samples': 24225280, 'steps': 47314, 'loss/train': 1.8470346927642822}
+02/26/2022 14:51:52 - INFO - codeparrot_training - Step 47315: {'lr': 3.850329688192988e-06, 'samples': 24225792, 'steps': 47315, 'loss/train': 1.945963740348816}
+02/26/2022 14:51:58 - INFO - codeparrot_training - Step 47316: {'lr': 3.8474695705287836e-06, 'samples': 24226304, 'steps': 47316, 'loss/train': 1.5298091173171997}
+02/26/2022 14:52:01 - INFO - codeparrot_training - Step 47317: {'lr': 3.844610507303919e-06, 'samples': 24226816, 'steps': 47317, 'loss/train': 1.9373265504837036}
+02/26/2022 14:52:07 - INFO - codeparrot_training - Step 47318: {'lr': 3.841752498530526e-06, 'samples': 24227328, 'steps': 47318, 'loss/train': 1.685444712638855}
+02/26/2022 14:52:10 - INFO - codeparrot_training - Step 47319: {'lr': 3.838895544220927e-06, 'samples': 24227840, 'steps': 47319, 'loss/train': 1.9345566034317017}
+02/26/2022 14:52:16 - INFO - codeparrot_training - Step 47320: {'lr': 3.836039644387307e-06, 'samples': 24228352, 'steps': 47320, 'loss/train': 1.5173789262771606}
+02/26/2022 14:52:19 - INFO - codeparrot_training - Step 47321: {'lr': 3.833184799041961e-06, 'samples': 24228864, 'steps': 47321, 'loss/train': 0.48507097363471985}
+02/26/2022 14:52:25 - INFO - codeparrot_training - Step 47322: {'lr': 3.830331008197046e-06, 'samples': 24229376, 'steps': 47322, 'loss/train': 1.4980988502502441}
+02/26/2022 14:52:28 - INFO - codeparrot_training - Step 47323: {'lr': 3.8274782718648594e-06, 'samples': 24229888, 'steps': 47323, 'loss/train': 0.8690078258514404}
+02/26/2022 14:52:34 - INFO - codeparrot_training - Step 47324: {'lr': 3.824626590057556e-06, 'samples': 24230400, 'steps': 47324, 'loss/train': 3.355264186859131}
+02/26/2022 14:52:37 - INFO - codeparrot_training - Step 47325: {'lr': 3.8217759627874326e-06, 'samples': 24230912, 'steps': 47325, 'loss/train': 2.0121114253997803}
+02/26/2022 14:52:43 - INFO - codeparrot_training - Step 47326: {'lr': 3.8189263900665905e-06, 'samples': 24231424, 'steps': 47326, 'loss/train': 1.4708033800125122}
+02/26/2022 14:52:46 - INFO - codeparrot_training - Step 47327: {'lr': 3.816077871907325e-06, 'samples': 24231936, 'steps': 47327, 'loss/train': 2.1761159896850586}
+02/26/2022 14:52:53 - INFO - codeparrot_training - Step 47328: {'lr': 3.8132304083217937e-06, 'samples': 24232448, 'steps': 47328, 'loss/train': 1.5536339282989502}
+02/26/2022 14:52:56 - INFO - codeparrot_training - Step 47329: {'lr': 3.810383999322237e-06, 'samples': 24232960, 'steps': 47329, 'loss/train': 1.9731495380401611}
+02/26/2022 14:53:02 - INFO - codeparrot_training - Step 47330: {'lr': 3.8075386449208103e-06, 'samples': 24233472, 'steps': 47330, 'loss/train': 1.7803375720977783}
+02/26/2022 14:53:05 - INFO - codeparrot_training - Step 47331: {'lr': 3.8046943451297276e-06, 'samples': 24233984, 'steps': 47331, 'loss/train': 1.6086125373840332}
+02/26/2022 14:53:11 - INFO - codeparrot_training - Step 47332: {'lr': 3.8018510999611454e-06, 'samples': 24234496, 'steps': 47332, 'loss/train': 2.4230523109436035}
+02/26/2022 14:53:14 - INFO - codeparrot_training - Step 47333: {'lr': 3.7990089094272485e-06, 'samples': 24235008, 'steps': 47333, 'loss/train': 0.918489396572113}
+02/26/2022 14:53:20 - INFO - codeparrot_training - Step 47334: {'lr': 3.796167773540221e-06, 'samples': 24235520, 'steps': 47334, 'loss/train': 1.5616077184677124}
+02/26/2022 14:53:23 - INFO - codeparrot_training - Step 47335: {'lr': 3.7933276923122484e-06, 'samples': 24236032, 'steps': 47335, 'loss/train': 1.6945126056671143}
+02/26/2022 14:53:29 - INFO - codeparrot_training - Step 47336: {'lr': 3.790488665755459e-06, 'samples': 24236544, 'steps': 47336, 'loss/train': 0.9781691431999207}
+02/26/2022 14:53:32 - INFO - codeparrot_training - Step 47337: {'lr': 3.7876506938820386e-06, 'samples': 24237056, 'steps': 47337, 'loss/train': 2.12276291847229}
+02/26/2022 14:53:39 - INFO - codeparrot_training - Step 47338: {'lr': 3.7848137767041713e-06, 'samples': 24237568, 'steps': 47338, 'loss/train': 1.6119569540023804}
+02/26/2022 14:53:42 - INFO - codeparrot_training - Step 47339: {'lr': 3.7819779142339582e-06, 'samples': 24238080, 'steps': 47339, 'loss/train': 1.9657509326934814}
+02/26/2022 14:53:48 - INFO - codeparrot_training - Step 47340: {'lr': 3.779143106483557e-06, 'samples': 24238592, 'steps': 47340, 'loss/train': 1.524492621421814}
+02/26/2022 14:53:51 - INFO - codeparrot_training - Step 47341: {'lr': 3.7763093534651237e-06, 'samples': 24239104, 'steps': 47341, 'loss/train': 1.890255331993103}
+02/26/2022 14:53:57 - INFO - codeparrot_training - Step 47342: {'lr': 3.7734766551908163e-06, 'samples': 24239616, 'steps': 47342, 'loss/train': 0.9133675694465637}
+02/26/2022 14:54:00 - INFO - codeparrot_training - Step 47343: {'lr': 3.770645011672763e-06, 'samples': 24240128, 'steps': 47343, 'loss/train': 1.5039483308792114}
+02/26/2022 14:54:06 - INFO - codeparrot_training - Step 47344: {'lr': 3.7678144229230382e-06, 'samples': 24240640, 'steps': 47344, 'loss/train': 2.1652932167053223}
+02/26/2022 14:54:09 - INFO - codeparrot_training - Step 47345: {'lr': 3.764984888953826e-06, 'samples': 24241152, 'steps': 47345, 'loss/train': 1.8980461359024048}
+02/26/2022 14:54:15 - INFO - codeparrot_training - Step 47346: {'lr': 3.762156409777229e-06, 'samples': 24241664, 'steps': 47346, 'loss/train': 0.4810251295566559}
+02/26/2022 14:54:18 - INFO - codeparrot_training - Step 47347: {'lr': 3.759328985405375e-06, 'samples': 24242176, 'steps': 47347, 'loss/train': 2.0912744998931885}
+02/26/2022 14:54:24 - INFO - codeparrot_training - Step 47348: {'lr': 3.756502615850338e-06, 'samples': 24242688, 'steps': 47348, 'loss/train': 1.7837584018707275}
+02/26/2022 14:54:27 - INFO - codeparrot_training - Step 47349: {'lr': 3.7536773011242753e-06, 'samples': 24243200, 'steps': 47349, 'loss/train': 2.5813305377960205}
+02/26/2022 14:54:34 - INFO - codeparrot_training - Step 47350: {'lr': 3.7508530412392883e-06, 'samples': 24243712, 'steps': 47350, 'loss/train': 1.704620361328125}
+02/26/2022 14:54:39 - INFO - codeparrot_training - Step 47351: {'lr': 3.748029836207395e-06, 'samples': 24244224, 'steps': 47351, 'loss/train': 0.703568696975708}
+02/26/2022 14:54:43 - INFO - codeparrot_training - Step 47352: {'lr': 3.745207686040808e-06, 'samples': 24244736, 'steps': 47352, 'loss/train': 0.9513370990753174}
+02/26/2022 14:54:48 - INFO - codeparrot_training - Step 47353: {'lr': 3.7423865907515175e-06, 'samples': 24245248, 'steps': 47353, 'loss/train': 2.3073689937591553}
+02/26/2022 14:54:52 - INFO - codeparrot_training - Step 47354: {'lr': 3.7395665503516806e-06, 'samples': 24245760, 'steps': 47354, 'loss/train': 1.8946318626403809}
+02/26/2022 14:54:57 - INFO - codeparrot_training - Step 47355: {'lr': 3.7367475648533157e-06, 'samples': 24246272, 'steps': 47355, 'loss/train': 1.4713902473449707}
+02/26/2022 14:55:01 - INFO - codeparrot_training - Step 47356: {'lr': 3.7339296342685515e-06, 'samples': 24246784, 'steps': 47356, 'loss/train': 1.4497612714767456}
+02/26/2022 14:55:06 - INFO - codeparrot_training - Step 47357: {'lr': 3.7311127586094064e-06, 'samples': 24247296, 'steps': 47357, 'loss/train': 1.3966649770736694}
+02/26/2022 14:55:10 - INFO - codeparrot_training - Step 47358: {'lr': 3.7282969378879816e-06, 'samples': 24247808, 'steps': 47358, 'loss/train': 0.34197771549224854}
+02/26/2022 14:55:15 - INFO - codeparrot_training - Step 47359: {'lr': 3.7254821721163233e-06, 'samples': 24248320, 'steps': 47359, 'loss/train': 0.8078081011772156}
+02/26/2022 14:55:19 - INFO - codeparrot_training - Step 47360: {'lr': 3.722668461306533e-06, 'samples': 24248832, 'steps': 47360, 'loss/train': 1.8701469898223877}
+02/26/2022 14:55:24 - INFO - codeparrot_training - Step 47361: {'lr': 3.719855805470601e-06, 'samples': 24249344, 'steps': 47361, 'loss/train': 2.415700912475586}
+02/26/2022 14:55:28 - INFO - codeparrot_training - Step 47362: {'lr': 3.7170442046206287e-06, 'samples': 24249856, 'steps': 47362, 'loss/train': 1.931935429573059}
+02/26/2022 14:55:34 - INFO - codeparrot_training - Step 47363: {'lr': 3.7142336587686064e-06, 'samples': 24250368, 'steps': 47363, 'loss/train': 1.5630022287368774}
+02/26/2022 14:55:37 - INFO - codeparrot_training - Step 47364: {'lr': 3.711424167926608e-06, 'samples': 24250880, 'steps': 47364, 'loss/train': 2.204237937927246}
+02/26/2022 14:55:43 - INFO - codeparrot_training - Step 47365: {'lr': 3.7086157321066794e-06, 'samples': 24251392, 'steps': 47365, 'loss/train': 3.933102607727051}
+02/26/2022 14:55:46 - INFO - codeparrot_training - Step 47366: {'lr': 3.7058083513208384e-06, 'samples': 24251904, 'steps': 47366, 'loss/train': 1.891339659690857}
+02/26/2022 14:55:52 - INFO - codeparrot_training - Step 47367: {'lr': 3.703002025581076e-06, 'samples': 24252416, 'steps': 47367, 'loss/train': 2.610166072845459}
+02/26/2022 14:55:55 - INFO - codeparrot_training - Step 47368: {'lr': 3.700196754899465e-06, 'samples': 24252928, 'steps': 47368, 'loss/train': 1.6646748781204224}
+02/26/2022 14:56:01 - INFO - codeparrot_training - Step 47369: {'lr': 3.697392539287997e-06, 'samples': 24253440, 'steps': 47369, 'loss/train': 2.169846296310425}
+02/26/2022 14:56:04 - INFO - codeparrot_training - Step 47370: {'lr': 3.694589378758689e-06, 'samples': 24253952, 'steps': 47370, 'loss/train': 1.9926788806915283}
+02/26/2022 14:56:10 - INFO - codeparrot_training - Step 47371: {'lr': 3.69178727332356e-06, 'samples': 24254464, 'steps': 47371, 'loss/train': 0.7172079086303711}
+02/26/2022 14:56:13 - INFO - codeparrot_training - Step 47372: {'lr': 3.6889862229946004e-06, 'samples': 24254976, 'steps': 47372, 'loss/train': 1.9406155347824097}
+02/26/2022 14:56:19 - INFO - codeparrot_training - Step 47373: {'lr': 3.6861862277838e-06, 'samples': 24255488, 'steps': 47373, 'loss/train': 2.2167246341705322}
+02/26/2022 14:56:23 - INFO - codeparrot_training - Step 47374: {'lr': 3.683387287703177e-06, 'samples': 24256000, 'steps': 47374, 'loss/train': 2.2148501873016357}
+02/26/2022 14:56:28 - INFO - codeparrot_training - Step 47375: {'lr': 3.680589402764695e-06, 'samples': 24256512, 'steps': 47375, 'loss/train': 0.8398675322532654}
+02/26/2022 14:56:32 - INFO - codeparrot_training - Step 47376: {'lr': 3.6777925729803707e-06, 'samples': 24257024, 'steps': 47376, 'loss/train': 2.2023701667785645}
+02/26/2022 14:56:37 - INFO - codeparrot_training - Step 47377: {'lr': 3.67499679836214e-06, 'samples': 24257536, 'steps': 47377, 'loss/train': 1.3533493280410767}
+02/26/2022 14:56:41 - INFO - codeparrot_training - Step 47378: {'lr': 3.672202078922049e-06, 'samples': 24258048, 'steps': 47378, 'loss/train': 1.2798556089401245}
+02/26/2022 14:56:46 - INFO - codeparrot_training - Step 47379: {'lr': 3.6694084146719764e-06, 'samples': 24258560, 'steps': 47379, 'loss/train': 1.0382750034332275}
+02/26/2022 14:56:50 - INFO - codeparrot_training - Step 47380: {'lr': 3.6666158056239683e-06, 'samples': 24259072, 'steps': 47380, 'loss/train': 1.899681806564331}
+02/26/2022 14:56:55 - INFO - codeparrot_training - Step 47381: {'lr': 3.6638242517899323e-06, 'samples': 24259584, 'steps': 47381, 'loss/train': 1.4494026899337769}
+02/26/2022 14:56:59 - INFO - codeparrot_training - Step 47382: {'lr': 3.661033753181886e-06, 'samples': 24260096, 'steps': 47382, 'loss/train': 1.8838310241699219}
+02/26/2022 14:57:05 - INFO - codeparrot_training - Step 47383: {'lr': 3.6582443098117367e-06, 'samples': 24260608, 'steps': 47383, 'loss/train': 2.3559370040893555}
+02/26/2022 14:57:08 - INFO - codeparrot_training - Step 47384: {'lr': 3.6554559216914475e-06, 'samples': 24261120, 'steps': 47384, 'loss/train': 1.1949645280838013}
+02/26/2022 14:57:14 - INFO - codeparrot_training - Step 47385: {'lr': 3.6526685888329525e-06, 'samples': 24261632, 'steps': 47385, 'loss/train': 0.6307389736175537}
+02/26/2022 14:57:17 - INFO - codeparrot_training - Step 47386: {'lr': 3.6498823112482428e-06, 'samples': 24262144, 'steps': 47386, 'loss/train': 1.106641411781311}
+02/26/2022 14:57:23 - INFO - codeparrot_training - Step 47387: {'lr': 3.6470970889491417e-06, 'samples': 24262656, 'steps': 47387, 'loss/train': 1.0352836847305298}
+02/26/2022 14:57:26 - INFO - codeparrot_training - Step 47388: {'lr': 3.6443129219476958e-06, 'samples': 24263168, 'steps': 47388, 'loss/train': 1.0688190460205078}
+02/26/2022 14:57:32 - INFO - codeparrot_training - Step 47389: {'lr': 3.6415298102557838e-06, 'samples': 24263680, 'steps': 47389, 'loss/train': 1.5037307739257812}
+02/26/2022 14:57:35 - INFO - codeparrot_training - Step 47390: {'lr': 3.6387477538853132e-06, 'samples': 24264192, 'steps': 47390, 'loss/train': 1.4985733032226562}
+02/26/2022 14:57:41 - INFO - codeparrot_training - Step 47391: {'lr': 3.6359667528482464e-06, 'samples': 24264704, 'steps': 47391, 'loss/train': 1.855826735496521}
+02/26/2022 14:57:44 - INFO - codeparrot_training - Step 47392: {'lr': 3.6331868071564634e-06, 'samples': 24265216, 'steps': 47392, 'loss/train': 1.3368381261825562}
+02/26/2022 14:57:51 - INFO - codeparrot_training - Step 47393: {'lr': 3.6304079168218705e-06, 'samples': 24265728, 'steps': 47393, 'loss/train': 0.9868101477622986}
+02/26/2022 14:57:54 - INFO - codeparrot_training - Step 47394: {'lr': 3.6276300818563477e-06, 'samples': 24266240, 'steps': 47394, 'loss/train': 2.121750831604004}
+02/26/2022 14:57:59 - INFO - codeparrot_training - Step 47395: {'lr': 3.624853302271858e-06, 'samples': 24266752, 'steps': 47395, 'loss/train': 0.9207903146743774}
+02/26/2022 14:58:03 - INFO - codeparrot_training - Step 47396: {'lr': 3.6220775780802794e-06, 'samples': 24267264, 'steps': 47396, 'loss/train': 1.5492602586746216}
+02/26/2022 14:58:08 - INFO - codeparrot_training - Step 47397: {'lr': 3.619302909293465e-06, 'samples': 24267776, 'steps': 47397, 'loss/train': 1.608044981956482}
+02/26/2022 14:58:12 - INFO - codeparrot_training - Step 47398: {'lr': 3.616529295923321e-06, 'samples': 24268288, 'steps': 47398, 'loss/train': 1.4074846506118774}
+02/26/2022 14:58:18 - INFO - codeparrot_training - Step 47399: {'lr': 3.613756737981727e-06, 'samples': 24268800, 'steps': 47399, 'loss/train': 1.6231176853179932}
+02/26/2022 14:58:21 - INFO - codeparrot_training - Step 47400: {'lr': 3.6109852354805626e-06, 'samples': 24269312, 'steps': 47400, 'loss/train': 1.6907190084457397}
+02/26/2022 14:58:27 - INFO - codeparrot_training - Step 47401: {'lr': 3.608214788431735e-06, 'samples': 24269824, 'steps': 47401, 'loss/train': 1.3162966966629028}
+02/26/2022 14:58:30 - INFO - codeparrot_training - Step 47402: {'lr': 3.6054453968470124e-06, 'samples': 24270336, 'steps': 47402, 'loss/train': 0.8982712626457214}
+02/26/2022 14:58:36 - INFO - codeparrot_training - Step 47403: {'lr': 3.6026770607383853e-06, 'samples': 24270848, 'steps': 47403, 'loss/train': 1.4896438121795654}
+02/26/2022 14:58:39 - INFO - codeparrot_training - Step 47404: {'lr': 3.599909780117622e-06, 'samples': 24271360, 'steps': 47404, 'loss/train': 1.3685739040374756}
+02/26/2022 14:58:45 - INFO - codeparrot_training - Step 47405: {'lr': 3.5971435549966012e-06, 'samples': 24271872, 'steps': 47405, 'loss/train': 1.7780133485794067}
+02/26/2022 14:58:48 - INFO - codeparrot_training - Step 47406: {'lr': 3.594378385387176e-06, 'samples': 24272384, 'steps': 47406, 'loss/train': 2.2715048789978027}
+02/26/2022 14:58:54 - INFO - codeparrot_training - Step 47407: {'lr': 3.5916142713011967e-06, 'samples': 24272896, 'steps': 47407, 'loss/train': 2.9677579402923584}
+02/26/2022 14:58:57 - INFO - codeparrot_training - Step 47408: {'lr': 3.588851212750488e-06, 'samples': 24273408, 'steps': 47408, 'loss/train': 2.1170194149017334}
+02/26/2022 14:59:03 - INFO - codeparrot_training - Step 47409: {'lr': 3.586089209746929e-06, 'samples': 24273920, 'steps': 47409, 'loss/train': 1.920351505279541}
+02/26/2022 14:59:07 - INFO - codeparrot_training - Step 47410: {'lr': 3.5833282623022877e-06, 'samples': 24274432, 'steps': 47410, 'loss/train': 2.446288585662842}
+02/26/2022 14:59:12 - INFO - codeparrot_training - Step 47411: {'lr': 3.5805683704284165e-06, 'samples': 24274944, 'steps': 47411, 'loss/train': 1.3646293878555298}
+02/26/2022 14:59:16 - INFO - codeparrot_training - Step 47412: {'lr': 3.577809534137139e-06, 'samples': 24275456, 'steps': 47412, 'loss/train': 2.390080690383911}
+02/26/2022 14:59:21 - INFO - codeparrot_training - Step 47413: {'lr': 3.5750517534403336e-06, 'samples': 24275968, 'steps': 47413, 'loss/train': 0.7405255436897278}
+02/26/2022 14:59:25 - INFO - codeparrot_training - Step 47414: {'lr': 3.572295028349687e-06, 'samples': 24276480, 'steps': 47414, 'loss/train': 1.668779969215393}
+02/26/2022 14:59:30 - INFO - codeparrot_training - Step 47415: {'lr': 3.569539358877133e-06, 'samples': 24276992, 'steps': 47415, 'loss/train': 2.4658994674682617}
+02/26/2022 14:59:34 - INFO - codeparrot_training - Step 47416: {'lr': 3.566784745034385e-06, 'samples': 24277504, 'steps': 47416, 'loss/train': 1.9553532600402832}
+02/26/2022 14:59:39 - INFO - codeparrot_training - Step 47417: {'lr': 3.564031186833322e-06, 'samples': 24278016, 'steps': 47417, 'loss/train': 1.6932862997055054}
+02/26/2022 14:59:43 - INFO - codeparrot_training - Step 47418: {'lr': 3.561278684285657e-06, 'samples': 24278528, 'steps': 47418, 'loss/train': 1.7158762216567993}
+02/26/2022 14:59:49 - INFO - codeparrot_training - Step 47419: {'lr': 3.558527237403242e-06, 'samples': 24279040, 'steps': 47419, 'loss/train': 1.5551339387893677}
+02/26/2022 14:59:52 - INFO - codeparrot_training - Step 47420: {'lr': 3.555776846197817e-06, 'samples': 24279552, 'steps': 47420, 'loss/train': 1.2387826442718506}
+02/26/2022 14:59:58 - INFO - codeparrot_training - Step 47421: {'lr': 3.553027510681206e-06, 'samples': 24280064, 'steps': 47421, 'loss/train': 1.6238585710525513}
+02/26/2022 15:00:02 - INFO - codeparrot_training - Step 47422: {'lr': 3.5502792308651776e-06, 'samples': 24280576, 'steps': 47422, 'loss/train': 1.372475266456604}
+02/26/2022 15:00:07 - INFO - codeparrot_training - Step 47423: {'lr': 3.5475320067614726e-06, 'samples': 24281088, 'steps': 47423, 'loss/train': 1.4243037700653076}
+02/26/2022 15:00:11 - INFO - codeparrot_training - Step 47424: {'lr': 3.5447858383818865e-06, 'samples': 24281600, 'steps': 47424, 'loss/train': 2.4198484420776367}
+02/26/2022 15:00:16 - INFO - codeparrot_training - Step 47425: {'lr': 3.542040725738216e-06, 'samples': 24282112, 'steps': 47425, 'loss/train': 0.3040185868740082}
+02/26/2022 15:00:20 - INFO - codeparrot_training - Step 47426: {'lr': 3.539296668842146e-06, 'samples': 24282624, 'steps': 47426, 'loss/train': 1.7414060831069946}
+02/26/2022 15:00:25 - INFO - codeparrot_training - Step 47427: {'lr': 3.5365536677055e-06, 'samples': 24283136, 'steps': 47427, 'loss/train': 2.551719903945923}
+02/26/2022 15:00:29 - INFO - codeparrot_training - Step 47428: {'lr': 3.5338117223399634e-06, 'samples': 24283648, 'steps': 47428, 'loss/train': 1.7943593263626099}
+02/26/2022 15:00:35 - INFO - codeparrot_training - Step 47429: {'lr': 3.5310708327573326e-06, 'samples': 24284160, 'steps': 47429, 'loss/train': 1.3163498640060425}
+02/26/2022 15:00:38 - INFO - codeparrot_training - Step 47430: {'lr': 3.528330998969348e-06, 'samples': 24284672, 'steps': 47430, 'loss/train': 3.230424404144287}
+02/26/2022 15:00:45 - INFO - codeparrot_training - Step 47431: {'lr': 3.525592220987722e-06, 'samples': 24285184, 'steps': 47431, 'loss/train': 1.8927180767059326}
+02/26/2022 15:00:48 - INFO - codeparrot_training - Step 47432: {'lr': 3.5228544988241682e-06, 'samples': 24285696, 'steps': 47432, 'loss/train': 1.9084442853927612}
+02/26/2022 15:00:53 - INFO - codeparrot_training - Step 47433: {'lr': 3.520117832490455e-06, 'samples': 24286208, 'steps': 47433, 'loss/train': 1.9714888334274292}
+02/26/2022 15:00:57 - INFO - codeparrot_training - Step 47434: {'lr': 3.517382221998294e-06, 'samples': 24286720, 'steps': 47434, 'loss/train': 1.7211190462112427}
+02/26/2022 15:01:03 - INFO - codeparrot_training - Step 47435: {'lr': 3.5146476673593995e-06, 'samples': 24287232, 'steps': 47435, 'loss/train': 1.8167824745178223}
+02/26/2022 15:01:06 - INFO - codeparrot_training - Step 47436: {'lr': 3.511914168585484e-06, 'samples': 24287744, 'steps': 47436, 'loss/train': 0.7277877926826477}
+02/26/2022 15:01:11 - INFO - codeparrot_training - Step 47437: {'lr': 3.5091817256882595e-06, 'samples': 24288256, 'steps': 47437, 'loss/train': 1.646148920059204}
+02/26/2022 15:01:15 - INFO - codeparrot_training - Step 47438: {'lr': 3.50645033867944e-06, 'samples': 24288768, 'steps': 47438, 'loss/train': 1.8899078369140625}
+02/26/2022 15:01:20 - INFO - codeparrot_training - Step 47439: {'lr': 3.50372000757071e-06, 'samples': 24289280, 'steps': 47439, 'loss/train': 0.9157168865203857}
+02/26/2022 15:01:24 - INFO - codeparrot_training - Step 47440: {'lr': 3.5009907323737824e-06, 'samples': 24289792, 'steps': 47440, 'loss/train': 1.067230224609375}
+02/26/2022 15:01:30 - INFO - codeparrot_training - Step 47441: {'lr': 3.4982625131003144e-06, 'samples': 24290304, 'steps': 47441, 'loss/train': 2.1384260654449463}
+02/26/2022 15:01:34 - INFO - codeparrot_training - Step 47442: {'lr': 3.4955353497620468e-06, 'samples': 24290816, 'steps': 47442, 'loss/train': 2.2401299476623535}
+02/26/2022 15:01:39 - INFO - codeparrot_training - Step 47443: {'lr': 3.492809242370609e-06, 'samples': 24291328, 'steps': 47443, 'loss/train': 1.7456283569335938}
+02/26/2022 15:01:43 - INFO - codeparrot_training - Step 47444: {'lr': 3.4900841909377145e-06, 'samples': 24291840, 'steps': 47444, 'loss/train': 1.1179039478302002}
+02/26/2022 15:01:48 - INFO - codeparrot_training - Step 47445: {'lr': 3.4873601954749923e-06, 'samples': 24292352, 'steps': 47445, 'loss/train': 1.4707119464874268}
+02/26/2022 15:01:52 - INFO - codeparrot_training - Step 47446: {'lr': 3.484637255994183e-06, 'samples': 24292864, 'steps': 47446, 'loss/train': 1.4192888736724854}
+02/26/2022 15:01:57 - INFO - codeparrot_training - Step 47447: {'lr': 3.4819153725068886e-06, 'samples': 24293376, 'steps': 47447, 'loss/train': 1.6675523519515991}
+02/26/2022 15:02:01 - INFO - codeparrot_training - Step 47448: {'lr': 3.479194545024822e-06, 'samples': 24293888, 'steps': 47448, 'loss/train': 1.9217102527618408}
+02/26/2022 15:02:06 - INFO - codeparrot_training - Step 47449: {'lr': 3.476474773559557e-06, 'samples': 24294400, 'steps': 47449, 'loss/train': 0.9532285332679749}
+02/26/2022 15:02:10 - INFO - codeparrot_training - Step 47450: {'lr': 3.4737560581228343e-06, 'samples': 24294912, 'steps': 47450, 'loss/train': 1.9057599306106567}
+02/26/2022 15:02:16 - INFO - codeparrot_training - Step 47451: {'lr': 3.4710383987262283e-06, 'samples': 24295424, 'steps': 47451, 'loss/train': 2.1189582347869873}
+02/26/2022 15:02:19 - INFO - codeparrot_training - Step 47452: {'lr': 3.468321795381452e-06, 'samples': 24295936, 'steps': 47452, 'loss/train': 1.730580449104309}
+02/26/2022 15:02:25 - INFO - codeparrot_training - Step 47453: {'lr': 3.4656062481000504e-06, 'samples': 24296448, 'steps': 47453, 'loss/train': 1.8071969747543335}
+02/26/2022 15:02:28 - INFO - codeparrot_training - Step 47454: {'lr': 3.462891756893738e-06, 'samples': 24296960, 'steps': 47454, 'loss/train': 1.624642252922058}
+02/26/2022 15:02:34 - INFO - codeparrot_training - Step 47455: {'lr': 3.4601783217740877e-06, 'samples': 24297472, 'steps': 47455, 'loss/train': 1.4693812131881714}
+02/26/2022 15:02:37 - INFO - codeparrot_training - Step 47456: {'lr': 3.457465942752813e-06, 'samples': 24297984, 'steps': 47456, 'loss/train': 1.4773683547973633}
+02/26/2022 15:02:43 - INFO - codeparrot_training - Step 47457: {'lr': 3.454754619841405e-06, 'samples': 24298496, 'steps': 47457, 'loss/train': 2.012878179550171}
+02/26/2022 15:02:46 - INFO - codeparrot_training - Step 47458: {'lr': 3.4520443530515754e-06, 'samples': 24299008, 'steps': 47458, 'loss/train': 1.4629181623458862}
+02/26/2022 15:02:52 - INFO - codeparrot_training - Step 47459: {'lr': 3.4493351423948715e-06, 'samples': 24299520, 'steps': 47459, 'loss/train': 1.534316062927246}
+02/26/2022 15:02:55 - INFO - codeparrot_training - Step 47460: {'lr': 3.4466269878829504e-06, 'samples': 24300032, 'steps': 47460, 'loss/train': 1.194318413734436}
+02/26/2022 15:03:01 - INFO - codeparrot_training - Step 47461: {'lr': 3.443919889527386e-06, 'samples': 24300544, 'steps': 47461, 'loss/train': 1.2803740501403809}
+02/26/2022 15:03:04 - INFO - codeparrot_training - Step 47462: {'lr': 3.44121384733978e-06, 'samples': 24301056, 'steps': 47462, 'loss/train': 2.1351022720336914}
+02/26/2022 15:03:10 - INFO - codeparrot_training - Step 47463: {'lr': 3.4385088613317075e-06, 'samples': 24301568, 'steps': 47463, 'loss/train': 1.7526729106903076}
+02/26/2022 15:03:13 - INFO - codeparrot_training - Step 47464: {'lr': 3.435804931514769e-06, 'samples': 24302080, 'steps': 47464, 'loss/train': 1.1786935329437256}
+02/26/2022 15:03:19 - INFO - codeparrot_training - Step 47465: {'lr': 3.4331020579005666e-06, 'samples': 24302592, 'steps': 47465, 'loss/train': 2.1805975437164307}
+02/26/2022 15:03:22 - INFO - codeparrot_training - Step 47466: {'lr': 3.430400240500675e-06, 'samples': 24303104, 'steps': 47466, 'loss/train': 2.414189577102661}
+02/26/2022 15:03:29 - INFO - codeparrot_training - Step 47467: {'lr': 3.4276994793266127e-06, 'samples': 24303616, 'steps': 47467, 'loss/train': 1.058998703956604}
+02/26/2022 15:03:32 - INFO - codeparrot_training - Step 47468: {'lr': 3.4249997743900083e-06, 'samples': 24304128, 'steps': 47468, 'loss/train': 3.5504348278045654}
+02/26/2022 15:03:38 - INFO - codeparrot_training - Step 47469: {'lr': 3.4223011257024097e-06, 'samples': 24304640, 'steps': 47469, 'loss/train': 2.891859769821167}
+02/26/2022 15:03:41 - INFO - codeparrot_training - Step 47470: {'lr': 3.4196035332753893e-06, 'samples': 24305152, 'steps': 47470, 'loss/train': 1.0789157152175903}
+02/26/2022 15:03:47 - INFO - codeparrot_training - Step 47471: {'lr': 3.416906997120467e-06, 'samples': 24305664, 'steps': 47471, 'loss/train': 2.1937248706817627}
+02/26/2022 15:03:50 - INFO - codeparrot_training - Step 47472: {'lr': 3.4142115172492437e-06, 'samples': 24306176, 'steps': 47472, 'loss/train': 1.5575580596923828}
+02/26/2022 15:03:56 - INFO - codeparrot_training - Step 47473: {'lr': 3.411517093673211e-06, 'samples': 24306688, 'steps': 47473, 'loss/train': 1.9336330890655518}
+02/26/2022 15:03:59 - INFO - codeparrot_training - Step 47474: {'lr': 3.40882372640397e-06, 'samples': 24307200, 'steps': 47474, 'loss/train': 2.114950180053711}
+02/26/2022 15:04:05 - INFO - codeparrot_training - Step 47475: {'lr': 3.4061314154529843e-06, 'samples': 24307712, 'steps': 47475, 'loss/train': 1.6089240312576294}
+02/26/2022 15:04:08 - INFO - codeparrot_training - Step 47476: {'lr': 3.4034401608318557e-06, 'samples': 24308224, 'steps': 47476, 'loss/train': 0.06756830215454102}
+02/26/2022 15:04:15 - INFO - codeparrot_training - Step 47477: {'lr': 3.4007499625520746e-06, 'samples': 24308736, 'steps': 47477, 'loss/train': 2.412686586380005}
+02/26/2022 15:04:18 - INFO - codeparrot_training - Step 47478: {'lr': 3.3980608206251873e-06, 'samples': 24309248, 'steps': 47478, 'loss/train': 2.7915639877319336}
+02/26/2022 15:04:24 - INFO - codeparrot_training - Step 47479: {'lr': 3.3953727350627126e-06, 'samples': 24309760, 'steps': 47479, 'loss/train': 2.4462890625}
+02/26/2022 15:04:27 - INFO - codeparrot_training - Step 47480: {'lr': 3.392685705876142e-06, 'samples': 24310272, 'steps': 47480, 'loss/train': 0.7830101251602173}
+02/26/2022 15:04:33 - INFO - codeparrot_training - Step 47481: {'lr': 3.3899997330769927e-06, 'samples': 24310784, 'steps': 47481, 'loss/train': 1.7793724536895752}
+02/26/2022 15:04:36 - INFO - codeparrot_training - Step 47482: {'lr': 3.387314816676784e-06, 'samples': 24311296, 'steps': 47482, 'loss/train': 1.7306865453720093}
+02/26/2022 15:04:42 - INFO - codeparrot_training - Step 47483: {'lr': 3.384630956687007e-06, 'samples': 24311808, 'steps': 47483, 'loss/train': 1.7700717449188232}
+02/26/2022 15:04:45 - INFO - codeparrot_training - Step 47484: {'lr': 3.381948153119152e-06, 'samples': 24312320, 'steps': 47484, 'loss/train': 2.3642008304595947}
+02/26/2022 15:04:51 - INFO - codeparrot_training - Step 47485: {'lr': 3.3792664059847376e-06, 'samples': 24312832, 'steps': 47485, 'loss/train': 1.9936814308166504}
+02/26/2022 15:04:54 - INFO - codeparrot_training - Step 47486: {'lr': 3.3765857152952274e-06, 'samples': 24313344, 'steps': 47486, 'loss/train': 1.6109830141067505}
+02/26/2022 15:05:01 - INFO - codeparrot_training - Step 47487: {'lr': 3.373906081062139e-06, 'samples': 24313856, 'steps': 47487, 'loss/train': 1.959956407546997}
+02/26/2022 15:05:04 - INFO - codeparrot_training - Step 47488: {'lr': 3.3712275032968807e-06, 'samples': 24314368, 'steps': 47488, 'loss/train': 0.8398148417472839}
+02/26/2022 15:05:10 - INFO - codeparrot_training - Step 47489: {'lr': 3.368549982010999e-06, 'samples': 24314880, 'steps': 47489, 'loss/train': 1.0779094696044922}
+02/26/2022 15:05:13 - INFO - codeparrot_training - Step 47490: {'lr': 3.3658735172159284e-06, 'samples': 24315392, 'steps': 47490, 'loss/train': 2.073573350906372}
+02/26/2022 15:05:19 - INFO - codeparrot_training - Step 47491: {'lr': 3.36319810892316e-06, 'samples': 24315904, 'steps': 47491, 'loss/train': 0.5230470895767212}
+02/26/2022 15:05:22 - INFO - codeparrot_training - Step 47492: {'lr': 3.360523757144102e-06, 'samples': 24316416, 'steps': 47492, 'loss/train': 1.7574613094329834}
+02/26/2022 15:05:28 - INFO - codeparrot_training - Step 47493: {'lr': 3.3578504618902437e-06, 'samples': 24316928, 'steps': 47493, 'loss/train': 0.9887236952781677}
+02/26/2022 15:05:31 - INFO - codeparrot_training - Step 47494: {'lr': 3.35517822317305e-06, 'samples': 24317440, 'steps': 47494, 'loss/train': 1.171176791191101}
+02/26/2022 15:05:37 - INFO - codeparrot_training - Step 47495: {'lr': 3.352507041003955e-06, 'samples': 24317952, 'steps': 47495, 'loss/train': 1.6361877918243408}
+02/26/2022 15:05:40 - INFO - codeparrot_training - Step 47496: {'lr': 3.349836915394394e-06, 'samples': 24318464, 'steps': 47496, 'loss/train': 1.488653540611267}
+02/26/2022 15:05:46 - INFO - codeparrot_training - Step 47497: {'lr': 3.347167846355803e-06, 'samples': 24318976, 'steps': 47497, 'loss/train': 2.1035194396972656}
+02/26/2022 15:05:49 - INFO - codeparrot_training - Step 47498: {'lr': 3.3444998338996448e-06, 'samples': 24319488, 'steps': 47498, 'loss/train': 2.71098256111145}
+02/26/2022 15:05:55 - INFO - codeparrot_training - Step 47499: {'lr': 3.3418328780372986e-06, 'samples': 24320000, 'steps': 47499, 'loss/train': 0.03847160562872887}
+02/26/2022 15:05:58 - INFO - codeparrot_training - Step 47500: {'lr': 3.339166978780256e-06, 'samples': 24320512, 'steps': 47500, 'loss/train': 1.979993224143982}
+02/26/2022 15:06:04 - INFO - codeparrot_training - Step 47501: {'lr': 3.3365021361398962e-06, 'samples': 24321024, 'steps': 47501, 'loss/train': 1.8054109811782837}
+02/26/2022 15:06:07 - INFO - codeparrot_training - Step 47502: {'lr': 3.3338383501275993e-06, 'samples': 24321536, 'steps': 47502, 'loss/train': 0.2787412106990814}
+02/26/2022 15:06:14 - INFO - codeparrot_training - Step 47503: {'lr': 3.3311756207548284e-06, 'samples': 24322048, 'steps': 47503, 'loss/train': 2.277825355529785}
+02/26/2022 15:06:17 - INFO - codeparrot_training - Step 47504: {'lr': 3.328513948032991e-06, 'samples': 24322560, 'steps': 47504, 'loss/train': 1.191369891166687}
+02/26/2022 15:06:23 - INFO - codeparrot_training - Step 47505: {'lr': 3.325853331973466e-06, 'samples': 24323072, 'steps': 47505, 'loss/train': 1.222564697265625}
+02/26/2022 15:06:26 - INFO - codeparrot_training - Step 47506: {'lr': 3.3231937725876347e-06, 'samples': 24323584, 'steps': 47506, 'loss/train': 0.8408288359642029}
+02/26/2022 15:06:32 - INFO - codeparrot_training - Step 47507: {'lr': 3.3205352698869317e-06, 'samples': 24324096, 'steps': 47507, 'loss/train': 0.8858546018600464}
+02/26/2022 15:06:35 - INFO - codeparrot_training - Step 47508: {'lr': 3.317877823882737e-06, 'samples': 24324608, 'steps': 47508, 'loss/train': 1.0575590133666992}
+02/26/2022 15:06:41 - INFO - codeparrot_training - Step 47509: {'lr': 3.3152214345864294e-06, 'samples': 24325120, 'steps': 47509, 'loss/train': 1.9508386850357056}
+02/26/2022 15:06:46 - INFO - codeparrot_training - Step 47510: {'lr': 3.3125661020093346e-06, 'samples': 24325632, 'steps': 47510, 'loss/train': 2.042740821838379}
+02/26/2022 15:06:50 - INFO - codeparrot_training - Step 47511: {'lr': 3.3099118261629147e-06, 'samples': 24326144, 'steps': 47511, 'loss/train': 1.6537288427352905}
+02/26/2022 15:06:57 - INFO - codeparrot_training - Step 47512: {'lr': 3.3072586070584664e-06, 'samples': 24326656, 'steps': 47512, 'loss/train': 2.5362508296966553}
+02/26/2022 15:07:00 - INFO - codeparrot_training - Step 47513: {'lr': 3.304606444707453e-06, 'samples': 24327168, 'steps': 47513, 'loss/train': 1.4630937576293945}
+02/26/2022 15:07:04 - INFO - codeparrot_training - Step 47514: {'lr': 3.3019553391211153e-06, 'samples': 24327680, 'steps': 47514, 'loss/train': 0.23679926991462708}
+02/26/2022 15:07:09 - INFO - codeparrot_training - Step 47515: {'lr': 3.299305290310889e-06, 'samples': 24328192, 'steps': 47515, 'loss/train': 1.9297516345977783}
+02/26/2022 15:07:12 - INFO - codeparrot_training - Step 47516: {'lr': 3.2966562982880977e-06, 'samples': 24328704, 'steps': 47516, 'loss/train': 1.1394567489624023}
+02/26/2022 15:07:18 - INFO - codeparrot_training - Step 47517: {'lr': 3.294008363064094e-06, 'samples': 24329216, 'steps': 47517, 'loss/train': 2.362560987472534}
+02/26/2022 15:07:22 - INFO - codeparrot_training - Step 47518: {'lr': 3.2913614846502015e-06, 'samples': 24329728, 'steps': 47518, 'loss/train': 1.7055608034133911}
+02/26/2022 15:07:27 - INFO - codeparrot_training - Step 47519: {'lr': 3.288715663057801e-06, 'samples': 24330240, 'steps': 47519, 'loss/train': 0.8176570534706116}
+02/26/2022 15:07:31 - INFO - codeparrot_training - Step 47520: {'lr': 3.286070898298188e-06, 'samples': 24330752, 'steps': 47520, 'loss/train': 1.3797684907913208}
+02/26/2022 15:07:36 - INFO - codeparrot_training - Step 47521: {'lr': 3.2834271903826873e-06, 'samples': 24331264, 'steps': 47521, 'loss/train': 1.334601640701294}
+02/26/2022 15:07:42 - INFO - codeparrot_training - Step 47522: {'lr': 3.280784539322679e-06, 'samples': 24331776, 'steps': 47522, 'loss/train': 2.0882368087768555}
+02/26/2022 15:07:46 - INFO - codeparrot_training - Step 47523: {'lr': 3.2781429451294586e-06, 'samples': 24332288, 'steps': 47523, 'loss/train': 3.377593517303467}
+02/26/2022 15:07:49 - INFO - codeparrot_training - Step 47524: {'lr': 3.2755024078142957e-06, 'samples': 24332800, 'steps': 47524, 'loss/train': 1.9391717910766602}
+02/26/2022 15:07:55 - INFO - codeparrot_training - Step 47525: {'lr': 3.2728629273885424e-06, 'samples': 24333312, 'steps': 47525, 'loss/train': 2.326525926589966}
+02/26/2022 15:07:58 - INFO - codeparrot_training - Step 47526: {'lr': 3.2702245038635227e-06, 'samples': 24333824, 'steps': 47526, 'loss/train': 1.3061593770980835}
+02/26/2022 15:08:04 - INFO - codeparrot_training - Step 47527: {'lr': 3.267587137250505e-06, 'samples': 24334336, 'steps': 47527, 'loss/train': 1.1045658588409424}
+02/26/2022 15:08:09 - INFO - codeparrot_training - Step 47528: {'lr': 3.2649508275607863e-06, 'samples': 24334848, 'steps': 47528, 'loss/train': 1.3623377084732056}
+02/26/2022 15:08:13 - INFO - codeparrot_training - Step 47529: {'lr': 3.262315574805663e-06, 'samples': 24335360, 'steps': 47529, 'loss/train': 1.696001648902893}
+02/26/2022 15:08:18 - INFO - codeparrot_training - Step 47530: {'lr': 3.2596813789964596e-06, 'samples': 24335872, 'steps': 47530, 'loss/train': 1.374510407447815}
+02/26/2022 15:08:22 - INFO - codeparrot_training - Step 47531: {'lr': 3.257048240144417e-06, 'samples': 24336384, 'steps': 47531, 'loss/train': 0.8178138732910156}
+02/26/2022 15:08:28 - INFO - codeparrot_training - Step 47532: {'lr': 3.254416158260831e-06, 'samples': 24336896, 'steps': 47532, 'loss/train': 1.0946320295333862}
+02/26/2022 15:08:32 - INFO - codeparrot_training - Step 47533: {'lr': 3.2517851333569716e-06, 'samples': 24337408, 'steps': 47533, 'loss/train': 0.9300341606140137}
+02/26/2022 15:08:37 - INFO - codeparrot_training - Step 47534: {'lr': 3.249155165444134e-06, 'samples': 24337920, 'steps': 47534, 'loss/train': 1.6627955436706543}
+02/26/2022 15:08:41 - INFO - codeparrot_training - Step 47535: {'lr': 3.246526254533533e-06, 'samples': 24338432, 'steps': 47535, 'loss/train': 1.8643912076950073}
+02/26/2022 15:08:46 - INFO - codeparrot_training - Step 47536: {'lr': 3.243898400636491e-06, 'samples': 24338944, 'steps': 47536, 'loss/train': 1.4736212491989136}
+02/26/2022 15:08:50 - INFO - codeparrot_training - Step 47537: {'lr': 3.2412716037641955e-06, 'samples': 24339456, 'steps': 47537, 'loss/train': 2.1177971363067627}
+02/26/2022 15:08:55 - INFO - codeparrot_training - Step 47538: {'lr': 3.2386458639279693e-06, 'samples': 24339968, 'steps': 47538, 'loss/train': 1.907954216003418}
+02/26/2022 15:08:59 - INFO - codeparrot_training - Step 47539: {'lr': 3.236021181139026e-06, 'samples': 24340480, 'steps': 47539, 'loss/train': 0.431205689907074}
+02/26/2022 15:09:02 - INFO - codeparrot_training - Step 47540: {'lr': 3.233397555408607e-06, 'samples': 24340992, 'steps': 47540, 'loss/train': 7.405642509460449}
+02/26/2022 15:09:08 - INFO - codeparrot_training - Step 47541: {'lr': 3.230774986747953e-06, 'samples': 24341504, 'steps': 47541, 'loss/train': 1.1553869247436523}
+02/26/2022 15:09:12 - INFO - codeparrot_training - Step 47542: {'lr': 3.228153475168305e-06, 'samples': 24342016, 'steps': 47542, 'loss/train': 2.4501562118530273}
+02/26/2022 15:09:17 - INFO - codeparrot_training - Step 47543: {'lr': 3.225533020680904e-06, 'samples': 24342528, 'steps': 47543, 'loss/train': 2.49342679977417}
+02/26/2022 15:09:21 - INFO - codeparrot_training - Step 47544: {'lr': 3.222913623296936e-06, 'samples': 24343040, 'steps': 47544, 'loss/train': 2.1924617290496826}
+02/26/2022 15:09:27 - INFO - codeparrot_training - Step 47545: {'lr': 3.220295283027641e-06, 'samples': 24343552, 'steps': 47545, 'loss/train': 1.7320542335510254}
+02/26/2022 15:09:30 - INFO - codeparrot_training - Step 47546: {'lr': 3.217677999884261e-06, 'samples': 24344064, 'steps': 47546, 'loss/train': 2.0045504570007324}
+02/26/2022 15:09:36 - INFO - codeparrot_training - Step 47547: {'lr': 3.215061773877953e-06, 'samples': 24344576, 'steps': 47547, 'loss/train': 1.5592222213745117}
+02/26/2022 15:09:40 - INFO - codeparrot_training - Step 47548: {'lr': 3.212446605020014e-06, 'samples': 24345088, 'steps': 47548, 'loss/train': 1.1184674501419067}
+02/26/2022 15:09:43 - INFO - codeparrot_training - Step 47549: {'lr': 3.2098324933215182e-06, 'samples': 24345600, 'steps': 47549, 'loss/train': 1.0899333953857422}
+02/26/2022 15:09:49 - INFO - codeparrot_training - Step 47550: {'lr': 3.20721943879379e-06, 'samples': 24346112, 'steps': 47550, 'loss/train': 1.733041524887085}
+02/26/2022 15:09:52 - INFO - codeparrot_training - Step 47551: {'lr': 3.204607441447932e-06, 'samples': 24346624, 'steps': 47551, 'loss/train': 1.2579232454299927}
+02/26/2022 15:09:58 - INFO - codeparrot_training - Step 47552: {'lr': 3.2019965012952125e-06, 'samples': 24347136, 'steps': 47552, 'loss/train': 0.6781299710273743}
+02/26/2022 15:10:01 - INFO - codeparrot_training - Step 47553: {'lr': 3.1993866183467335e-06, 'samples': 24347648, 'steps': 47553, 'loss/train': 0.7907813191413879}
+02/26/2022 15:10:07 - INFO - codeparrot_training - Step 47554: {'lr': 3.1967777926137363e-06, 'samples': 24348160, 'steps': 47554, 'loss/train': 0.8809179067611694}
+02/26/2022 15:10:10 - INFO - codeparrot_training - Step 47555: {'lr': 3.194170024107351e-06, 'samples': 24348672, 'steps': 47555, 'loss/train': 1.2193326950073242}
+02/26/2022 15:10:16 - INFO - codeparrot_training - Step 47556: {'lr': 3.191563312838819e-06, 'samples': 24349184, 'steps': 47556, 'loss/train': 1.982816457748413}
+02/26/2022 15:10:19 - INFO - codeparrot_training - Step 47557: {'lr': 3.1889576588192136e-06, 'samples': 24349696, 'steps': 47557, 'loss/train': 0.85979163646698}
+02/26/2022 15:10:25 - INFO - codeparrot_training - Step 47558: {'lr': 3.186353062059749e-06, 'samples': 24350208, 'steps': 47558, 'loss/train': 1.4743715524673462}
+02/26/2022 15:10:28 - INFO - codeparrot_training - Step 47559: {'lr': 3.183749522571583e-06, 'samples': 24350720, 'steps': 47559, 'loss/train': 1.5385243892669678}
+02/26/2022 15:10:34 - INFO - codeparrot_training - Step 47560: {'lr': 3.181147040365873e-06, 'samples': 24351232, 'steps': 47560, 'loss/train': 1.7087470293045044}
+02/26/2022 15:10:38 - INFO - codeparrot_training - Step 47561: {'lr': 3.1785456154537485e-06, 'samples': 24351744, 'steps': 47561, 'loss/train': 1.294171929359436}
+02/26/2022 15:10:43 - INFO - codeparrot_training - Step 47562: {'lr': 3.1759452478463404e-06, 'samples': 24352256, 'steps': 47562, 'loss/train': 1.5617821216583252}
+02/26/2022 15:10:47 - INFO - codeparrot_training - Step 47563: {'lr': 3.1733459375548335e-06, 'samples': 24352768, 'steps': 47563, 'loss/train': 2.0581600666046143}
+02/26/2022 15:10:52 - INFO - codeparrot_training - Step 47564: {'lr': 3.1707476845903025e-06, 'samples': 24353280, 'steps': 47564, 'loss/train': 1.7121607065200806}
+02/26/2022 15:11:00 - INFO - codeparrot_training - Step 47565: {'lr': 3.168150488963961e-06, 'samples': 24353792, 'steps': 47565, 'loss/train': 8.073162078857422}
+02/26/2022 15:11:03 - INFO - codeparrot_training - Step 47566: {'lr': 3.1655543506868557e-06, 'samples': 24354304, 'steps': 47566, 'loss/train': 1.8005565404891968}
+02/26/2022 15:11:09 - INFO - codeparrot_training - Step 47567: {'lr': 3.1629592697701435e-06, 'samples': 24354816, 'steps': 47567, 'loss/train': 0.9688205122947693}
+02/26/2022 15:11:12 - INFO - codeparrot_training - Step 47568: {'lr': 3.1603652462249e-06, 'samples': 24355328, 'steps': 47568, 'loss/train': 1.6241395473480225}
+02/26/2022 15:11:18 - INFO - codeparrot_training - Step 47569: {'lr': 3.1577722800623098e-06, 'samples': 24355840, 'steps': 47569, 'loss/train': 2.0020203590393066}
+02/26/2022 15:11:21 - INFO - codeparrot_training - Step 47570: {'lr': 3.1551803712934477e-06, 'samples': 24356352, 'steps': 47570, 'loss/train': 1.3297297954559326}
+02/26/2022 15:11:27 - INFO - codeparrot_training - Step 47571: {'lr': 3.1525895199293886e-06, 'samples': 24356864, 'steps': 47571, 'loss/train': 1.98710298538208}
+02/26/2022 15:11:30 - INFO - codeparrot_training - Step 47572: {'lr': 3.149999725981262e-06, 'samples': 24357376, 'steps': 47572, 'loss/train': 1.347033977508545}
+02/26/2022 15:11:36 - INFO - codeparrot_training - Step 47573: {'lr': 3.1474109894601423e-06, 'samples': 24357888, 'steps': 47573, 'loss/train': 1.4820574522018433}
+02/26/2022 15:11:39 - INFO - codeparrot_training - Step 47574: {'lr': 3.1448233103771596e-06, 'samples': 24358400, 'steps': 47574, 'loss/train': 1.6216740608215332}
+02/26/2022 15:11:47 - INFO - codeparrot_training - Step 47575: {'lr': 3.142236688743333e-06, 'samples': 24358912, 'steps': 47575, 'loss/train': 2.0336694717407227}
+02/26/2022 15:11:50 - INFO - codeparrot_training - Step 47576: {'lr': 3.1396511245697922e-06, 'samples': 24359424, 'steps': 47576, 'loss/train': 1.9226903915405273}
+02/26/2022 15:11:55 - INFO - codeparrot_training - Step 47577: {'lr': 3.137066617867612e-06, 'samples': 24359936, 'steps': 47577, 'loss/train': 0.7686557173728943}
+02/26/2022 15:11:59 - INFO - codeparrot_training - Step 47578: {'lr': 3.1344831686478393e-06, 'samples': 24360448, 'steps': 47578, 'loss/train': 1.770987629890442}
+02/26/2022 15:12:04 - INFO - codeparrot_training - Step 47579: {'lr': 3.131900776921548e-06, 'samples': 24360960, 'steps': 47579, 'loss/train': 1.230187177658081}
+02/26/2022 15:12:08 - INFO - codeparrot_training - Step 47580: {'lr': 3.129319442699785e-06, 'samples': 24361472, 'steps': 47580, 'loss/train': 1.4721728563308716}
+02/26/2022 15:12:13 - INFO - codeparrot_training - Step 47581: {'lr': 3.126739165993653e-06, 'samples': 24361984, 'steps': 47581, 'loss/train': 1.4379863739013672}
+02/26/2022 15:12:17 - INFO - codeparrot_training - Step 47582: {'lr': 3.12415994681417e-06, 'samples': 24362496, 'steps': 47582, 'loss/train': 3.3473212718963623}
+02/26/2022 15:12:22 - INFO - codeparrot_training - Step 47583: {'lr': 3.121581785172439e-06, 'samples': 24363008, 'steps': 47583, 'loss/train': 1.0834342241287231}
+02/26/2022 15:12:26 - INFO - codeparrot_training - Step 47584: {'lr': 3.119004681079396e-06, 'samples': 24363520, 'steps': 47584, 'loss/train': 1.609164834022522}
+02/26/2022 15:12:33 - INFO - codeparrot_training - Step 47585: {'lr': 3.11642863454617e-06, 'samples': 24364032, 'steps': 47585, 'loss/train': 1.1637941598892212}
+02/26/2022 15:12:36 - INFO - codeparrot_training - Step 47586: {'lr': 3.113853645583753e-06, 'samples': 24364544, 'steps': 47586, 'loss/train': 0.15011459589004517}
+02/26/2022 15:12:42 - INFO - codeparrot_training - Step 47587: {'lr': 3.111279714203219e-06, 'samples': 24365056, 'steps': 47587, 'loss/train': 1.823533058166504}
+02/26/2022 15:12:45 - INFO - codeparrot_training - Step 47588: {'lr': 3.1087068404155593e-06, 'samples': 24365568, 'steps': 47588, 'loss/train': 0.8395389318466187}
+02/26/2022 15:12:51 - INFO - codeparrot_training - Step 47589: {'lr': 3.1061350242317933e-06, 'samples': 24366080, 'steps': 47589, 'loss/train': 2.761305093765259}
+02/26/2022 15:12:54 - INFO - codeparrot_training - Step 47590: {'lr': 3.1035642656629393e-06, 'samples': 24366592, 'steps': 47590, 'loss/train': 2.197582721710205}
+02/26/2022 15:13:00 - INFO - codeparrot_training - Step 47591: {'lr': 3.1009945647200443e-06, 'samples': 24367104, 'steps': 47591, 'loss/train': 1.3107774257659912}
+02/26/2022 15:13:03 - INFO - codeparrot_training - Step 47592: {'lr': 3.098425921414072e-06, 'samples': 24367616, 'steps': 47592, 'loss/train': 2.3657925128936768}
+02/26/2022 15:13:09 - INFO - codeparrot_training - Step 47593: {'lr': 3.095858335756041e-06, 'samples': 24368128, 'steps': 47593, 'loss/train': 1.4001954793930054}
+02/26/2022 15:13:12 - INFO - codeparrot_training - Step 47594: {'lr': 3.0932918077569427e-06, 'samples': 24368640, 'steps': 47594, 'loss/train': 1.7272108793258667}
+02/26/2022 15:13:18 - INFO - codeparrot_training - Step 47595: {'lr': 3.090726337427796e-06, 'samples': 24369152, 'steps': 47595, 'loss/train': 2.380660057067871}
+02/26/2022 15:13:21 - INFO - codeparrot_training - Step 47596: {'lr': 3.0881619247795645e-06, 'samples': 24369664, 'steps': 47596, 'loss/train': 0.865058183670044}
+02/26/2022 15:13:29 - INFO - codeparrot_training - Step 47597: {'lr': 3.085598569823267e-06, 'samples': 24370176, 'steps': 47597, 'loss/train': 1.7573013305664062}
+02/26/2022 15:13:32 - INFO - codeparrot_training - Step 47598: {'lr': 3.083036272569839e-06, 'samples': 24370688, 'steps': 47598, 'loss/train': 1.7063276767730713}
+02/26/2022 15:13:37 - INFO - codeparrot_training - Step 47599: {'lr': 3.080475033030272e-06, 'samples': 24371200, 'steps': 47599, 'loss/train': 1.7071788311004639}
+02/26/2022 15:13:41 - INFO - codeparrot_training - Step 47600: {'lr': 3.077914851215585e-06, 'samples': 24371712, 'steps': 47600, 'loss/train': 1.8461300134658813}
+02/26/2022 15:13:46 - INFO - codeparrot_training - Step 47601: {'lr': 3.0753557271366584e-06, 'samples': 24372224, 'steps': 47601, 'loss/train': 1.123298168182373}
+02/26/2022 15:13:50 - INFO - codeparrot_training - Step 47602: {'lr': 3.0727976608045106e-06, 'samples': 24372736, 'steps': 47602, 'loss/train': 1.5736615657806396}
+02/26/2022 15:13:55 - INFO - codeparrot_training - Step 47603: {'lr': 3.0702406522301053e-06, 'samples': 24373248, 'steps': 47603, 'loss/train': 2.1834020614624023}
+02/26/2022 15:13:59 - INFO - codeparrot_training - Step 47604: {'lr': 3.067684701424378e-06, 'samples': 24373760, 'steps': 47604, 'loss/train': 1.6734988689422607}
+02/26/2022 15:14:04 - INFO - codeparrot_training - Step 47605: {'lr': 3.0651298083982925e-06, 'samples': 24374272, 'steps': 47605, 'loss/train': 2.1561081409454346}
+02/26/2022 15:14:08 - INFO - codeparrot_training - Step 47606: {'lr': 3.062575973162729e-06, 'samples': 24374784, 'steps': 47606, 'loss/train': 1.7029731273651123}
+02/26/2022 15:14:13 - INFO - codeparrot_training - Step 47607: {'lr': 3.060023195728734e-06, 'samples': 24375296, 'steps': 47607, 'loss/train': 2.3150179386138916}
+02/26/2022 15:14:17 - INFO - codeparrot_training - Step 47608: {'lr': 3.0574714761071596e-06, 'samples': 24375808, 'steps': 47608, 'loss/train': 1.840514898300171}
+02/26/2022 15:14:22 - INFO - codeparrot_training - Step 47609: {'lr': 3.05492081430897e-06, 'samples': 24376320, 'steps': 47609, 'loss/train': 1.8019163608551025}
+02/26/2022 15:14:26 - INFO - codeparrot_training - Step 47610: {'lr': 3.052371210345073e-06, 'samples': 24376832, 'steps': 47610, 'loss/train': 0.302370548248291}
+02/26/2022 15:14:33 - INFO - codeparrot_training - Step 47611: {'lr': 3.0498226642264314e-06, 'samples': 24377344, 'steps': 47611, 'loss/train': 1.8133140802383423}
+02/26/2022 15:14:36 - INFO - codeparrot_training - Step 47612: {'lr': 3.0472751759639263e-06, 'samples': 24377856, 'steps': 47612, 'loss/train': 0.9807099103927612}
+02/26/2022 15:14:42 - INFO - codeparrot_training - Step 47613: {'lr': 3.044728745568465e-06, 'samples': 24378368, 'steps': 47613, 'loss/train': 1.9523282051086426}
+02/26/2022 15:14:48 - INFO - codeparrot_training - Step 47614: {'lr': 3.042183373050983e-06, 'samples': 24378880, 'steps': 47614, 'loss/train': 2.3784940242767334}
+02/26/2022 15:14:51 - INFO - codeparrot_training - Step 47615: {'lr': 3.0396390584223333e-06, 'samples': 24379392, 'steps': 47615, 'loss/train': 0.07598035782575607}
+02/26/2022 15:14:57 - INFO - codeparrot_training - Step 47616: {'lr': 3.037095801693479e-06, 'samples': 24379904, 'steps': 47616, 'loss/train': 1.2915031909942627}
+02/26/2022 15:15:00 - INFO - codeparrot_training - Step 47617: {'lr': 3.0345536028752728e-06, 'samples': 24380416, 'steps': 47617, 'loss/train': 1.849656105041504}
+02/26/2022 15:15:05 - INFO - codeparrot_training - Step 47618: {'lr': 3.0320124619786506e-06, 'samples': 24380928, 'steps': 47618, 'loss/train': 2.1262218952178955}
+02/26/2022 15:15:09 - INFO - codeparrot_training - Step 47619: {'lr': 3.0294723790144362e-06, 'samples': 24381440, 'steps': 47619, 'loss/train': 1.899937391281128}
+02/26/2022 15:15:16 - INFO - codeparrot_training - Step 47620: {'lr': 3.026933353993566e-06, 'samples': 24381952, 'steps': 47620, 'loss/train': 1.5570577383041382}
+02/26/2022 15:15:20 - INFO - codeparrot_training - Step 47621: {'lr': 3.0243953869268647e-06, 'samples': 24382464, 'steps': 47621, 'loss/train': 1.6037689447402954}
+02/26/2022 15:15:25 - INFO - codeparrot_training - Step 47622: {'lr': 3.02185847782524e-06, 'samples': 24382976, 'steps': 47622, 'loss/train': 1.9914475679397583}
+02/26/2022 15:15:29 - INFO - codeparrot_training - Step 47623: {'lr': 3.0193226266995444e-06, 'samples': 24383488, 'steps': 47623, 'loss/train': 2.3011107444763184}
+02/26/2022 15:15:34 - INFO - codeparrot_training - Step 47624: {'lr': 3.016787833560658e-06, 'samples': 24384000, 'steps': 47624, 'loss/train': 0.1855044662952423}
+02/26/2022 15:15:37 - INFO - codeparrot_training - Step 47625: {'lr': 3.0142540984194056e-06, 'samples': 24384512, 'steps': 47625, 'loss/train': 1.5040817260742188}
+02/26/2022 15:15:43 - INFO - codeparrot_training - Step 47626: {'lr': 3.011721421286695e-06, 'samples': 24385024, 'steps': 47626, 'loss/train': 1.5098698139190674}
+02/26/2022 15:15:46 - INFO - codeparrot_training - Step 47627: {'lr': 3.0091898021733233e-06, 'samples': 24385536, 'steps': 47627, 'loss/train': 1.8019421100616455}
+02/26/2022 15:15:52 - INFO - codeparrot_training - Step 47628: {'lr': 3.0066592410901427e-06, 'samples': 24386048, 'steps': 47628, 'loss/train': 3.0033986568450928}
+02/26/2022 15:15:55 - INFO - codeparrot_training - Step 47629: {'lr': 3.004129738048006e-06, 'samples': 24386560, 'steps': 47629, 'loss/train': 1.6713452339172363}
+02/26/2022 15:16:01 - INFO - codeparrot_training - Step 47630: {'lr': 3.0016012930577652e-06, 'samples': 24387072, 'steps': 47630, 'loss/train': 1.8882120847702026}
+02/26/2022 15:16:05 - INFO - codeparrot_training - Step 47631: {'lr': 2.9990739061302176e-06, 'samples': 24387584, 'steps': 47631, 'loss/train': 1.9239529371261597}
+02/26/2022 15:16:12 - INFO - codeparrot_training - Step 47632: {'lr': 2.996547577276215e-06, 'samples': 24388096, 'steps': 47632, 'loss/train': 1.236438274383545}
+02/26/2022 15:16:15 - INFO - codeparrot_training - Step 47633: {'lr': 2.994022306506555e-06, 'samples': 24388608, 'steps': 47633, 'loss/train': 1.224778175354004}
+02/26/2022 15:16:21 - INFO - codeparrot_training - Step 47634: {'lr': 2.9914980938320903e-06, 'samples': 24389120, 'steps': 47634, 'loss/train': 2.2543954849243164}
+02/26/2022 15:16:25 - INFO - codeparrot_training - Step 47635: {'lr': 2.988974939263617e-06, 'samples': 24389632, 'steps': 47635, 'loss/train': 2.205904006958008}
+02/26/2022 15:16:30 - INFO - codeparrot_training - Step 47636: {'lr': 2.986452842811932e-06, 'samples': 24390144, 'steps': 47636, 'loss/train': 0.9505888223648071}
+02/26/2022 15:16:34 - INFO - codeparrot_training - Step 47637: {'lr': 2.9839318044878326e-06, 'samples': 24390656, 'steps': 47637, 'loss/train': 1.8812248706817627}
+02/26/2022 15:16:37 - INFO - codeparrot_training - Step 47638: {'lr': 2.981411824302144e-06, 'samples': 24391168, 'steps': 47638, 'loss/train': 2.115784168243408}
+02/26/2022 15:16:43 - INFO - codeparrot_training - Step 47639: {'lr': 2.978892902265662e-06, 'samples': 24391680, 'steps': 47639, 'loss/train': 1.4282594919204712}
+02/26/2022 15:16:46 - INFO - codeparrot_training - Step 47640: {'lr': 2.9763750383891565e-06, 'samples': 24392192, 'steps': 47640, 'loss/train': 0.3510097861289978}
+02/26/2022 15:16:52 - INFO - codeparrot_training - Step 47641: {'lr': 2.973858232683424e-06, 'samples': 24392704, 'steps': 47641, 'loss/train': 2.1685428619384766}
+02/26/2022 15:16:59 - INFO - codeparrot_training - Step 47642: {'lr': 2.9713424851592342e-06, 'samples': 24393216, 'steps': 47642, 'loss/train': 1.6736844778060913}
+02/26/2022 15:17:02 - INFO - codeparrot_training - Step 47643: {'lr': 2.968827795827356e-06, 'samples': 24393728, 'steps': 47643, 'loss/train': 1.7606958150863647}
+02/26/2022 15:17:08 - INFO - codeparrot_training - Step 47644: {'lr': 2.9663141646986137e-06, 'samples': 24394240, 'steps': 47644, 'loss/train': 0.2881993353366852}
+02/26/2022 15:17:11 - INFO - codeparrot_training - Step 47645: {'lr': 2.9638015917837213e-06, 'samples': 24394752, 'steps': 47645, 'loss/train': 1.7967191934585571}
+02/26/2022 15:17:17 - INFO - codeparrot_training - Step 47646: {'lr': 2.961290077093476e-06, 'samples': 24395264, 'steps': 47646, 'loss/train': 1.323774814605713}
+02/26/2022 15:17:20 - INFO - codeparrot_training - Step 47647: {'lr': 2.9587796206386185e-06, 'samples': 24395776, 'steps': 47647, 'loss/train': 1.4970837831497192}
+02/26/2022 15:17:26 - INFO - codeparrot_training - Step 47648: {'lr': 2.956270222429891e-06, 'samples': 24396288, 'steps': 47648, 'loss/train': 0.6243499517440796}
+02/26/2022 15:17:29 - INFO - codeparrot_training - Step 47649: {'lr': 2.9537618824780345e-06, 'samples': 24396800, 'steps': 47649, 'loss/train': 1.6987028121948242}
+02/26/2022 15:17:35 - INFO - codeparrot_training - Step 47650: {'lr': 2.9512546007938457e-06, 'samples': 24397312, 'steps': 47650, 'loss/train': 1.4116984605789185}
+02/26/2022 15:17:38 - INFO - codeparrot_training - Step 47651: {'lr': 2.948748377388039e-06, 'samples': 24397824, 'steps': 47651, 'loss/train': 0.46655547618865967}
+02/26/2022 15:17:46 - INFO - codeparrot_training - Step 47652: {'lr': 2.9462432122713545e-06, 'samples': 24398336, 'steps': 47652, 'loss/train': 1.8256189823150635}
+02/26/2022 15:17:49 - INFO - codeparrot_training - Step 47653: {'lr': 2.9437391054544795e-06, 'samples': 24398848, 'steps': 47653, 'loss/train': 2.047633647918701}
+02/26/2022 15:17:55 - INFO - codeparrot_training - Step 47654: {'lr': 2.94123605694821e-06, 'samples': 24399360, 'steps': 47654, 'loss/train': 0.8964255452156067}
+02/26/2022 15:17:58 - INFO - codeparrot_training - Step 47655: {'lr': 2.938734066763232e-06, 'samples': 24399872, 'steps': 47655, 'loss/train': 2.083402395248413}
+02/26/2022 15:18:04 - INFO - codeparrot_training - Step 47656: {'lr': 2.9362331349102323e-06, 'samples': 24400384, 'steps': 47656, 'loss/train': 1.3712934255599976}
+02/26/2022 15:18:07 - INFO - codeparrot_training - Step 47657: {'lr': 2.933733261399979e-06, 'samples': 24400896, 'steps': 47657, 'loss/train': 2.032747507095337}
+02/26/2022 15:18:13 - INFO - codeparrot_training - Step 47658: {'lr': 2.9312344462431863e-06, 'samples': 24401408, 'steps': 47658, 'loss/train': 0.589367151260376}
+02/26/2022 15:18:16 - INFO - codeparrot_training - Step 47659: {'lr': 2.928736689450512e-06, 'samples': 24401920, 'steps': 47659, 'loss/train': 2.638061761856079}
+02/26/2022 15:18:22 - INFO - codeparrot_training - Step 47660: {'lr': 2.92623999103267e-06, 'samples': 24402432, 'steps': 47660, 'loss/train': 1.7041376829147339}
+02/26/2022 15:18:25 - INFO - codeparrot_training - Step 47661: {'lr': 2.9237443510003737e-06, 'samples': 24402944, 'steps': 47661, 'loss/train': 1.6518422365188599}
+02/26/2022 15:18:31 - INFO - codeparrot_training - Step 47662: {'lr': 2.9212497693642814e-06, 'samples': 24403456, 'steps': 47662, 'loss/train': 0.4850243628025055}
+02/26/2022 15:18:34 - INFO - codeparrot_training - Step 47663: {'lr': 2.918756246135107e-06, 'samples': 24403968, 'steps': 47663, 'loss/train': 1.7888721227645874}
+02/26/2022 15:18:40 - INFO - codeparrot_training - Step 47664: {'lr': 2.9162637813235083e-06, 'samples': 24404480, 'steps': 47664, 'loss/train': 0.19029074907302856}
+02/26/2022 15:18:43 - INFO - codeparrot_training - Step 47665: {'lr': 2.913772374940227e-06, 'samples': 24404992, 'steps': 47665, 'loss/train': 2.127103328704834}
+02/26/2022 15:18:50 - INFO - codeparrot_training - Step 47666: {'lr': 2.9112820269958374e-06, 'samples': 24405504, 'steps': 47666, 'loss/train': 1.6891711950302124}
+02/26/2022 15:18:54 - INFO - codeparrot_training - Step 47667: {'lr': 2.9087927375010816e-06, 'samples': 24406016, 'steps': 47667, 'loss/train': 2.296438217163086}
+02/26/2022 15:19:00 - INFO - codeparrot_training - Step 47668: {'lr': 2.9063045064665894e-06, 'samples': 24406528, 'steps': 47668, 'loss/train': 1.5600039958953857}
+02/26/2022 15:19:03 - INFO - codeparrot_training - Step 47669: {'lr': 2.9038173339030194e-06, 'samples': 24407040, 'steps': 47669, 'loss/train': 1.7696064710617065}
+02/26/2022 15:19:07 - INFO - codeparrot_training - Step 47670: {'lr': 2.9013312198210573e-06, 'samples': 24407552, 'steps': 47670, 'loss/train': 2.2763867378234863}
+02/26/2022 15:19:12 - INFO - codeparrot_training - Step 47671: {'lr': 2.898846164231306e-06, 'samples': 24408064, 'steps': 47671, 'loss/train': 1.164310336112976}
+02/26/2022 15:19:16 - INFO - codeparrot_training - Step 47672: {'lr': 2.896362167144423e-06, 'samples': 24408576, 'steps': 47672, 'loss/train': 1.4535681009292603}
+02/26/2022 15:19:21 - INFO - codeparrot_training - Step 47673: {'lr': 2.8938792285710946e-06, 'samples': 24409088, 'steps': 47673, 'loss/train': 1.094604253768921}
+02/26/2022 15:19:25 - INFO - codeparrot_training - Step 47674: {'lr': 2.8913973485219236e-06, 'samples': 24409600, 'steps': 47674, 'loss/train': 0.3681040406227112}
+02/26/2022 15:19:30 - INFO - codeparrot_training - Step 47675: {'lr': 2.8889165270075122e-06, 'samples': 24410112, 'steps': 47675, 'loss/train': 2.456069231033325}
+02/26/2022 15:19:34 - INFO - codeparrot_training - Step 47676: {'lr': 2.8864367640385192e-06, 'samples': 24410624, 'steps': 47676, 'loss/train': 2.094545364379883}
+02/26/2022 15:19:39 - INFO - codeparrot_training - Step 47677: {'lr': 2.8839580596255743e-06, 'samples': 24411136, 'steps': 47677, 'loss/train': 0.7565767765045166}
+02/26/2022 15:19:43 - INFO - codeparrot_training - Step 47678: {'lr': 2.8814804137792803e-06, 'samples': 24411648, 'steps': 47678, 'loss/train': 1.3240689039230347}
+02/26/2022 15:19:50 - INFO - codeparrot_training - Step 47679: {'lr': 2.879003826510268e-06, 'samples': 24412160, 'steps': 47679, 'loss/train': 1.258461356163025}
+02/26/2022 15:19:54 - INFO - codeparrot_training - Step 47680: {'lr': 2.876528297829112e-06, 'samples': 24412672, 'steps': 47680, 'loss/train': 1.2884104251861572}
+02/26/2022 15:19:59 - INFO - codeparrot_training - Step 47681: {'lr': 2.87405382774647e-06, 'samples': 24413184, 'steps': 47681, 'loss/train': 2.0182693004608154}
+02/26/2022 15:20:03 - INFO - codeparrot_training - Step 47682: {'lr': 2.871580416272862e-06, 'samples': 24413696, 'steps': 47682, 'loss/train': 1.2920496463775635}
+02/26/2022 15:20:08 - INFO - codeparrot_training - Step 47683: {'lr': 2.8691080634190013e-06, 'samples': 24414208, 'steps': 47683, 'loss/train': 1.6101053953170776}
+02/26/2022 15:20:12 - INFO - codeparrot_training - Step 47684: {'lr': 2.866636769195352e-06, 'samples': 24414720, 'steps': 47684, 'loss/train': 0.8663586974143982}
+02/26/2022 15:20:17 - INFO - codeparrot_training - Step 47685: {'lr': 2.864166533612572e-06, 'samples': 24415232, 'steps': 47685, 'loss/train': 2.719475269317627}
+02/26/2022 15:20:21 - INFO - codeparrot_training - Step 47686: {'lr': 2.8616973566812087e-06, 'samples': 24415744, 'steps': 47686, 'loss/train': 2.119499921798706}
+02/26/2022 15:20:26 - INFO - codeparrot_training - Step 47687: {'lr': 2.8592292384118923e-06, 'samples': 24416256, 'steps': 47687, 'loss/train': 1.5559251308441162}
+02/26/2022 15:20:30 - INFO - codeparrot_training - Step 47688: {'lr': 2.8567621788151423e-06, 'samples': 24416768, 'steps': 47688, 'loss/train': 1.6287193298339844}
+02/26/2022 15:20:37 - INFO - codeparrot_training - Step 47689: {'lr': 2.8542961779015606e-06, 'samples': 24417280, 'steps': 47689, 'loss/train': 1.0163534879684448}
+02/26/2022 15:20:40 - INFO - codeparrot_training - Step 47690: {'lr': 2.851831235681668e-06, 'samples': 24417792, 'steps': 47690, 'loss/train': 1.458247423171997}
+02/26/2022 15:20:46 - INFO - codeparrot_training - Step 47691: {'lr': 2.849367352166066e-06, 'samples': 24418304, 'steps': 47691, 'loss/train': 2.1798388957977295}
+02/26/2022 15:20:49 - INFO - codeparrot_training - Step 47692: {'lr': 2.8469045273653294e-06, 'samples': 24418816, 'steps': 47692, 'loss/train': 1.5786243677139282}
+02/26/2022 15:20:55 - INFO - codeparrot_training - Step 47693: {'lr': 2.8444427612899502e-06, 'samples': 24419328, 'steps': 47693, 'loss/train': 1.2595247030258179}
+02/26/2022 15:20:58 - INFO - codeparrot_training - Step 47694: {'lr': 2.841982053950476e-06, 'samples': 24419840, 'steps': 47694, 'loss/train': 1.4161241054534912}
+02/26/2022 15:21:04 - INFO - codeparrot_training - Step 47695: {'lr': 2.839522405357481e-06, 'samples': 24420352, 'steps': 47695, 'loss/train': 1.810314416885376}
+02/26/2022 15:21:07 - INFO - codeparrot_training - Step 47696: {'lr': 2.8370638155215123e-06, 'samples': 24420864, 'steps': 47696, 'loss/train': 1.4527337551116943}
+02/26/2022 15:21:13 - INFO - codeparrot_training - Step 47697: {'lr': 2.8346062844530617e-06, 'samples': 24421376, 'steps': 47697, 'loss/train': 1.8947196006774902}
+02/26/2022 15:21:16 - INFO - codeparrot_training - Step 47698: {'lr': 2.8321498121626767e-06, 'samples': 24421888, 'steps': 47698, 'loss/train': 1.9455469846725464}
+02/26/2022 15:21:24 - INFO - codeparrot_training - Step 47699: {'lr': 2.829694398660848e-06, 'samples': 24422400, 'steps': 47699, 'loss/train': 0.8469429016113281}
+02/26/2022 15:21:27 - INFO - codeparrot_training - Step 47700: {'lr': 2.8272400439581513e-06, 'samples': 24422912, 'steps': 47700, 'loss/train': 0.3897291421890259}
+02/26/2022 15:21:32 - INFO - codeparrot_training - Step 47701: {'lr': 2.824786748065078e-06, 'samples': 24423424, 'steps': 47701, 'loss/train': 1.4133137464523315}
+02/26/2022 15:21:36 - INFO - codeparrot_training - Step 47702: {'lr': 2.8223345109921196e-06, 'samples': 24423936, 'steps': 47702, 'loss/train': 2.38584566116333}
+02/26/2022 15:21:41 - INFO - codeparrot_training - Step 47703: {'lr': 2.8198833327497952e-06, 'samples': 24424448, 'steps': 47703, 'loss/train': 0.43464207649230957}
+02/26/2022 15:21:45 - INFO - codeparrot_training - Step 47704: {'lr': 2.817433213348597e-06, 'samples': 24424960, 'steps': 47704, 'loss/train': 0.126841738820076}
+02/26/2022 15:21:51 - INFO - codeparrot_training - Step 47705: {'lr': 2.8149841527990437e-06, 'samples': 24425472, 'steps': 47705, 'loss/train': 0.4895975887775421}
+02/26/2022 15:21:54 - INFO - codeparrot_training - Step 47706: {'lr': 2.8125361511115997e-06, 'samples': 24425984, 'steps': 47706, 'loss/train': 2.1269664764404297}
+02/26/2022 15:21:59 - INFO - codeparrot_training - Step 47707: {'lr': 2.8100892082967565e-06, 'samples': 24426496, 'steps': 47707, 'loss/train': 1.6531054973602295}
+02/26/2022 15:22:03 - INFO - codeparrot_training - Step 47708: {'lr': 2.8076433243650056e-06, 'samples': 24427008, 'steps': 47708, 'loss/train': 1.432506799697876}
+02/26/2022 15:22:08 - INFO - codeparrot_training - Step 47709: {'lr': 2.8051984993268107e-06, 'samples': 24427520, 'steps': 47709, 'loss/train': 0.6510240435600281}
+02/26/2022 15:22:12 - INFO - codeparrot_training - Step 47710: {'lr': 2.802754733192664e-06, 'samples': 24428032, 'steps': 47710, 'loss/train': 1.3855977058410645}
+02/26/2022 15:22:17 - INFO - codeparrot_training - Step 47711: {'lr': 2.8003120259730007e-06, 'samples': 24428544, 'steps': 47711, 'loss/train': 2.165266752243042}
+02/26/2022 15:22:21 - INFO - codeparrot_training - Step 47712: {'lr': 2.7978703776783132e-06, 'samples': 24429056, 'steps': 47712, 'loss/train': 0.7595996856689453}
+02/26/2022 15:22:26 - INFO - codeparrot_training - Step 47713: {'lr': 2.795429788319065e-06, 'samples': 24429568, 'steps': 47713, 'loss/train': 1.2099123001098633}
+02/26/2022 15:22:30 - INFO - codeparrot_training - Step 47714: {'lr': 2.792990257905692e-06, 'samples': 24430080, 'steps': 47714, 'loss/train': 8.540417671203613}
+02/26/2022 15:22:37 - INFO - codeparrot_training - Step 47715: {'lr': 2.790551786448631e-06, 'samples': 24430592, 'steps': 47715, 'loss/train': 0.41386866569519043}
+02/26/2022 15:22:41 - INFO - codeparrot_training - Step 47716: {'lr': 2.7881143739583724e-06, 'samples': 24431104, 'steps': 47716, 'loss/train': 2.180248975753784}
+02/26/2022 15:22:46 - INFO - codeparrot_training - Step 47717: {'lr': 2.7856780204453256e-06, 'samples': 24431616, 'steps': 47717, 'loss/train': 1.8266843557357788}
+02/26/2022 15:22:50 - INFO - codeparrot_training - Step 47718: {'lr': 2.7832427259199546e-06, 'samples': 24432128, 'steps': 47718, 'loss/train': 0.36090385913848877}
+02/26/2022 15:22:56 - INFO - codeparrot_training - Step 47719: {'lr': 2.780808490392639e-06, 'samples': 24432640, 'steps': 47719, 'loss/train': 1.6662006378173828}
+02/26/2022 15:22:59 - INFO - codeparrot_training - Step 47720: {'lr': 2.778375313873871e-06, 'samples': 24433152, 'steps': 47720, 'loss/train': 1.184025764465332}
+02/26/2022 15:23:05 - INFO - codeparrot_training - Step 47721: {'lr': 2.7759431963740035e-06, 'samples': 24433664, 'steps': 47721, 'loss/train': 2.0826425552368164}
+02/26/2022 15:23:08 - INFO - codeparrot_training - Step 47722: {'lr': 2.7735121379035276e-06, 'samples': 24434176, 'steps': 47722, 'loss/train': 2.0559120178222656}
+02/26/2022 15:23:14 - INFO - codeparrot_training - Step 47723: {'lr': 2.771082138472797e-06, 'samples': 24434688, 'steps': 47723, 'loss/train': 1.30818510055542}
+02/26/2022 15:23:17 - INFO - codeparrot_training - Step 47724: {'lr': 2.7686531980922746e-06, 'samples': 24435200, 'steps': 47724, 'loss/train': 1.9541465044021606}
+02/26/2022 15:23:25 - INFO - codeparrot_training - Step 47725: {'lr': 2.766225316772314e-06, 'samples': 24435712, 'steps': 47725, 'loss/train': 1.82325279712677}
+02/26/2022 15:23:28 - INFO - codeparrot_training - Step 47726: {'lr': 2.7637984945233507e-06, 'samples': 24436224, 'steps': 47726, 'loss/train': 1.9219880104064941}
+02/26/2022 15:23:34 - INFO - codeparrot_training - Step 47727: {'lr': 2.761372731355766e-06, 'samples': 24436736, 'steps': 47727, 'loss/train': 1.6483488082885742}
+02/26/2022 15:23:37 - INFO - codeparrot_training - Step 47728: {'lr': 2.7589480272799395e-06, 'samples': 24437248, 'steps': 47728, 'loss/train': 0.16203279793262482}
+02/26/2022 15:23:43 - INFO - codeparrot_training - Step 47729: {'lr': 2.7565243823062803e-06, 'samples': 24437760, 'steps': 47729, 'loss/train': 0.5907553434371948}
+02/26/2022 15:23:47 - INFO - codeparrot_training - Step 47730: {'lr': 2.7541017964451685e-06, 'samples': 24438272, 'steps': 47730, 'loss/train': 1.9620444774627686}
+02/26/2022 15:23:52 - INFO - codeparrot_training - Step 47731: {'lr': 2.7516802697069855e-06, 'samples': 24438784, 'steps': 47731, 'loss/train': 1.0339369773864746}
+02/26/2022 15:23:56 - INFO - codeparrot_training - Step 47732: {'lr': 2.7492598021020833e-06, 'samples': 24439296, 'steps': 47732, 'loss/train': 1.3877111673355103}
+02/26/2022 15:24:01 - INFO - codeparrot_training - Step 47733: {'lr': 2.7468403936408147e-06, 'samples': 24439808, 'steps': 47733, 'loss/train': 1.1138633489608765}
+02/26/2022 15:24:05 - INFO - codeparrot_training - Step 47734: {'lr': 2.7444220443336165e-06, 'samples': 24440320, 'steps': 47734, 'loss/train': 1.8677597045898438}
+02/26/2022 15:24:12 - INFO - codeparrot_training - Step 47735: {'lr': 2.7420047541907576e-06, 'samples': 24440832, 'steps': 47735, 'loss/train': 1.9140660762786865}
+02/26/2022 15:24:16 - INFO - codeparrot_training - Step 47736: {'lr': 2.7395885232226748e-06, 'samples': 24441344, 'steps': 47736, 'loss/train': 1.9463571310043335}
+02/26/2022 15:24:21 - INFO - codeparrot_training - Step 47737: {'lr': 2.7371733514396645e-06, 'samples': 24441856, 'steps': 47737, 'loss/train': 1.8360120058059692}
+02/26/2022 15:24:24 - INFO - codeparrot_training - Step 47738: {'lr': 2.73475923885208e-06, 'samples': 24442368, 'steps': 47738, 'loss/train': 2.069361686706543}
+02/26/2022 15:24:30 - INFO - codeparrot_training - Step 47739: {'lr': 2.7323461854702746e-06, 'samples': 24442880, 'steps': 47739, 'loss/train': 1.7338221073150635}
+02/26/2022 15:24:34 - INFO - codeparrot_training - Step 47740: {'lr': 2.7299341913046006e-06, 'samples': 24443392, 'steps': 47740, 'loss/train': 2.8587684631347656}
+02/26/2022 15:24:39 - INFO - codeparrot_training - Step 47741: {'lr': 2.727523256365355e-06, 'samples': 24443904, 'steps': 47741, 'loss/train': 2.726759433746338}
+02/26/2022 15:24:43 - INFO - codeparrot_training - Step 47742: {'lr': 2.7251133806628916e-06, 'samples': 24444416, 'steps': 47742, 'loss/train': 2.069056510925293}
+02/26/2022 15:24:48 - INFO - codeparrot_training - Step 47743: {'lr': 2.7227045642075064e-06, 'samples': 24444928, 'steps': 47743, 'loss/train': 0.9621003866195679}
+02/26/2022 15:24:52 - INFO - codeparrot_training - Step 47744: {'lr': 2.7202968070095537e-06, 'samples': 24445440, 'steps': 47744, 'loss/train': 2.2205052375793457}
+02/26/2022 15:24:59 - INFO - codeparrot_training - Step 47745: {'lr': 2.7178901090793296e-06, 'samples': 24445952, 'steps': 47745, 'loss/train': 1.2422583103179932}
+02/26/2022 15:25:03 - INFO - codeparrot_training - Step 47746: {'lr': 2.715484470427132e-06, 'samples': 24446464, 'steps': 47746, 'loss/train': 2.436001777648926}
+02/26/2022 15:25:08 - INFO - codeparrot_training - Step 47747: {'lr': 2.713079891063286e-06, 'samples': 24446976, 'steps': 47747, 'loss/train': 1.2731224298477173}
+02/26/2022 15:25:12 - INFO - codeparrot_training - Step 47748: {'lr': 2.710676370998061e-06, 'samples': 24447488, 'steps': 47748, 'loss/train': 1.7909307479858398}
+02/26/2022 15:25:17 - INFO - codeparrot_training - Step 47749: {'lr': 2.7082739102418097e-06, 'samples': 24448000, 'steps': 47749, 'loss/train': 3.53122878074646}
+02/26/2022 15:25:21 - INFO - codeparrot_training - Step 47750: {'lr': 2.7058725088047465e-06, 'samples': 24448512, 'steps': 47750, 'loss/train': 0.3870745003223419}
+02/26/2022 15:25:26 - INFO - codeparrot_training - Step 47751: {'lr': 2.7034721666972518e-06, 'samples': 24449024, 'steps': 47751, 'loss/train': 2.2625205516815186}
+02/26/2022 15:25:30 - INFO - codeparrot_training - Step 47752: {'lr': 2.7010728839295118e-06, 'samples': 24449536, 'steps': 47752, 'loss/train': 1.1067169904708862}
+02/26/2022 15:25:35 - INFO - codeparrot_training - Step 47753: {'lr': 2.6986746605118795e-06, 'samples': 24450048, 'steps': 47753, 'loss/train': 1.8368946313858032}
+02/26/2022 15:25:39 - INFO - codeparrot_training - Step 47754: {'lr': 2.6962774964545687e-06, 'samples': 24450560, 'steps': 47754, 'loss/train': 1.2937439680099487}
+02/26/2022 15:25:44 - INFO - codeparrot_training - Step 47755: {'lr': 2.693881391767905e-06, 'samples': 24451072, 'steps': 47755, 'loss/train': 1.3347302675247192}
+02/26/2022 15:25:48 - INFO - codeparrot_training - Step 47756: {'lr': 2.691486346462102e-06, 'samples': 24451584, 'steps': 47756, 'loss/train': 1.0759286880493164}
+02/26/2022 15:25:53 - INFO - codeparrot_training - Step 47757: {'lr': 2.689092360547485e-06, 'samples': 24452096, 'steps': 47757, 'loss/train': 0.9846163392066956}
+02/26/2022 15:25:57 - INFO - codeparrot_training - Step 47758: {'lr': 2.68669943403424e-06, 'samples': 24452608, 'steps': 47758, 'loss/train': 0.17225030064582825}
+02/26/2022 15:26:02 - INFO - codeparrot_training - Step 47759: {'lr': 2.6843075669326366e-06, 'samples': 24453120, 'steps': 47759, 'loss/train': 1.6041593551635742}
+02/26/2022 15:26:06 - INFO - codeparrot_training - Step 47760: {'lr': 2.681916759252917e-06, 'samples': 24453632, 'steps': 47760, 'loss/train': 1.8171963691711426}
+02/26/2022 15:26:13 - INFO - codeparrot_training - Step 47761: {'lr': 2.6795270110053784e-06, 'samples': 24454144, 'steps': 47761, 'loss/train': 1.549590826034546}
+02/26/2022 15:26:16 - INFO - codeparrot_training - Step 47762: {'lr': 2.677138322200151e-06, 'samples': 24454656, 'steps': 47762, 'loss/train': 1.2060056924819946}
+02/26/2022 15:26:22 - INFO - codeparrot_training - Step 47763: {'lr': 2.6747506928475882e-06, 'samples': 24455168, 'steps': 47763, 'loss/train': 2.1794633865356445}
+02/26/2022 15:26:25 - INFO - codeparrot_training - Step 47764: {'lr': 2.672364122957793e-06, 'samples': 24455680, 'steps': 47764, 'loss/train': 1.4822766780853271}
+02/26/2022 15:26:31 - INFO - codeparrot_training - Step 47765: {'lr': 2.6699786125411187e-06, 'samples': 24456192, 'steps': 47765, 'loss/train': 1.182654857635498}
+02/26/2022 15:26:34 - INFO - codeparrot_training - Step 47766: {'lr': 2.667594161607695e-06, 'samples': 24456704, 'steps': 47766, 'loss/train': 1.8918431997299194}
+02/26/2022 15:26:40 - INFO - codeparrot_training - Step 47767: {'lr': 2.665210770167764e-06, 'samples': 24457216, 'steps': 47767, 'loss/train': 2.267146587371826}
+02/26/2022 15:26:43 - INFO - codeparrot_training - Step 47768: {'lr': 2.6628284382315125e-06, 'samples': 24457728, 'steps': 47768, 'loss/train': 0.549653172492981}
+02/26/2022 15:26:49 - INFO - codeparrot_training - Step 47769: {'lr': 2.6604471658091823e-06, 'samples': 24458240, 'steps': 47769, 'loss/train': 2.4738872051239014}
+02/26/2022 15:26:52 - INFO - codeparrot_training - Step 47770: {'lr': 2.658066952910959e-06, 'samples': 24458752, 'steps': 47770, 'loss/train': 1.3863539695739746}
+02/26/2022 15:27:00 - INFO - codeparrot_training - Step 47771: {'lr': 2.6556877995470572e-06, 'samples': 24459264, 'steps': 47771, 'loss/train': 1.9759031534194946}
+02/26/2022 15:27:03 - INFO - codeparrot_training - Step 47772: {'lr': 2.6533097057276077e-06, 'samples': 24459776, 'steps': 47772, 'loss/train': 1.876699686050415}
+02/26/2022 15:27:09 - INFO - codeparrot_training - Step 47773: {'lr': 2.650932671462852e-06, 'samples': 24460288, 'steps': 47773, 'loss/train': 1.3217414617538452}
+02/26/2022 15:27:12 - INFO - codeparrot_training - Step 47774: {'lr': 2.648556696762977e-06, 'samples': 24460800, 'steps': 47774, 'loss/train': 1.2455205917358398}
+02/26/2022 15:27:18 - INFO - codeparrot_training - Step 47775: {'lr': 2.6461817816381407e-06, 'samples': 24461312, 'steps': 47775, 'loss/train': 1.4605668783187866}
+02/26/2022 15:27:21 - INFO - codeparrot_training - Step 47776: {'lr': 2.6438079260984737e-06, 'samples': 24461824, 'steps': 47776, 'loss/train': 0.34825846552848816}
+02/26/2022 15:27:27 - INFO - codeparrot_training - Step 47777: {'lr': 2.641435130154246e-06, 'samples': 24462336, 'steps': 47777, 'loss/train': 2.0794758796691895}
+02/26/2022 15:27:30 - INFO - codeparrot_training - Step 47778: {'lr': 2.639063393815533e-06, 'samples': 24462848, 'steps': 47778, 'loss/train': 1.456945776939392}
+02/26/2022 15:27:36 - INFO - codeparrot_training - Step 47779: {'lr': 2.636692717092548e-06, 'samples': 24463360, 'steps': 47779, 'loss/train': 2.4288454055786133}
+02/26/2022 15:27:39 - INFO - codeparrot_training - Step 47780: {'lr': 2.634323099995395e-06, 'samples': 24463872, 'steps': 47780, 'loss/train': 2.041590452194214}
+02/26/2022 15:27:46 - INFO - codeparrot_training - Step 47781: {'lr': 2.6319545425342593e-06, 'samples': 24464384, 'steps': 47781, 'loss/train': 2.0931224822998047}
+02/26/2022 15:27:52 - INFO - codeparrot_training - Step 47782: {'lr': 2.6295870447193006e-06, 'samples': 24464896, 'steps': 47782, 'loss/train': 0.8226175308227539}
+02/26/2022 15:27:55 - INFO - codeparrot_training - Step 47783: {'lr': 2.6272206065606486e-06, 'samples': 24465408, 'steps': 47783, 'loss/train': 1.8073890209197998}
+02/26/2022 15:28:01 - INFO - codeparrot_training - Step 47784: {'lr': 2.6248552280684067e-06, 'samples': 24465920, 'steps': 47784, 'loss/train': 2.063500165939331}
+02/26/2022 15:28:04 - INFO - codeparrot_training - Step 47785: {'lr': 2.6224909092527337e-06, 'samples': 24466432, 'steps': 47785, 'loss/train': 1.3700064420700073}
+02/26/2022 15:28:10 - INFO - codeparrot_training - Step 47786: {'lr': 2.620127650123788e-06, 'samples': 24466944, 'steps': 47786, 'loss/train': 1.7232751846313477}
+02/26/2022 15:28:13 - INFO - codeparrot_training - Step 47787: {'lr': 2.6177654506916447e-06, 'samples': 24467456, 'steps': 47787, 'loss/train': 2.575131416320801}
+02/26/2022 15:28:19 - INFO - codeparrot_training - Step 47788: {'lr': 2.6154043109664626e-06, 'samples': 24467968, 'steps': 47788, 'loss/train': 3.261383056640625}
+02/26/2022 15:28:22 - INFO - codeparrot_training - Step 47789: {'lr': 2.613044230958289e-06, 'samples': 24468480, 'steps': 47789, 'loss/train': 1.7220776081085205}
+02/26/2022 15:28:28 - INFO - codeparrot_training - Step 47790: {'lr': 2.6106852106773106e-06, 'samples': 24468992, 'steps': 47790, 'loss/train': 1.222286343574524}
+02/26/2022 15:28:31 - INFO - codeparrot_training - Step 47791: {'lr': 2.608327250133602e-06, 'samples': 24469504, 'steps': 47791, 'loss/train': 0.5067892074584961}
+02/26/2022 15:28:38 - INFO - codeparrot_training - Step 47792: {'lr': 2.6059703493372665e-06, 'samples': 24470016, 'steps': 47792, 'loss/train': 1.6780601739883423}
+02/26/2022 15:28:42 - INFO - codeparrot_training - Step 47793: {'lr': 2.6036145082983797e-06, 'samples': 24470528, 'steps': 47793, 'loss/train': 0.21104831993579865}
+02/26/2022 15:28:47 - INFO - codeparrot_training - Step 47794: {'lr': 2.601259727027072e-06, 'samples': 24471040, 'steps': 47794, 'loss/train': 2.0346317291259766}
+02/26/2022 15:28:51 - INFO - codeparrot_training - Step 47795: {'lr': 2.5989060055333913e-06, 'samples': 24471552, 'steps': 47795, 'loss/train': 0.6854222416877747}
+02/26/2022 15:28:56 - INFO - codeparrot_training - Step 47796: {'lr': 2.596553343827468e-06, 'samples': 24472064, 'steps': 47796, 'loss/train': 1.9851348400115967}
+02/26/2022 15:29:00 - INFO - codeparrot_training - Step 47797: {'lr': 2.594201741919322e-06, 'samples': 24472576, 'steps': 47797, 'loss/train': 1.6300476789474487}
+02/26/2022 15:29:05 - INFO - codeparrot_training - Step 47798: {'lr': 2.5918511998190565e-06, 'samples': 24473088, 'steps': 47798, 'loss/train': 2.656505584716797}
+02/26/2022 15:29:09 - INFO - codeparrot_training - Step 47799: {'lr': 2.5895017175367463e-06, 'samples': 24473600, 'steps': 47799, 'loss/train': 1.9577513933181763}
+02/26/2022 15:29:12 - INFO - codeparrot_training - Step 47800: {'lr': 2.5871532950824397e-06, 'samples': 24474112, 'steps': 47800, 'loss/train': 1.0012179613113403}
+02/26/2022 15:29:18 - INFO - codeparrot_training - Step 47801: {'lr': 2.584805932466239e-06, 'samples': 24474624, 'steps': 47801, 'loss/train': 1.2555547952651978}
+02/26/2022 15:29:21 - INFO - codeparrot_training - Step 47802: {'lr': 2.5824596296981362e-06, 'samples': 24475136, 'steps': 47802, 'loss/train': 1.833644151687622}
+02/26/2022 15:29:27 - INFO - codeparrot_training - Step 47803: {'lr': 2.580114386788207e-06, 'samples': 24475648, 'steps': 47803, 'loss/train': 0.8827839493751526}
+02/26/2022 15:29:30 - INFO - codeparrot_training - Step 47804: {'lr': 2.5777702037465267e-06, 'samples': 24476160, 'steps': 47804, 'loss/train': 2.377434015274048}
+02/26/2022 15:29:36 - INFO - codeparrot_training - Step 47805: {'lr': 2.5754270805830872e-06, 'samples': 24476672, 'steps': 47805, 'loss/train': 1.6889660358428955}
+02/26/2022 15:29:43 - INFO - codeparrot_training - Step 47806: {'lr': 2.573085017307991e-06, 'samples': 24477184, 'steps': 47806, 'loss/train': 1.3994858264923096}
+02/26/2022 15:29:47 - INFO - codeparrot_training - Step 47807: {'lr': 2.570744013931203e-06, 'samples': 24477696, 'steps': 47807, 'loss/train': 1.5294290781021118}
+02/26/2022 15:29:52 - INFO - codeparrot_training - Step 47808: {'lr': 2.568404070462771e-06, 'samples': 24478208, 'steps': 47808, 'loss/train': 2.0072031021118164}
+02/26/2022 15:29:56 - INFO - codeparrot_training - Step 47809: {'lr': 2.5660651869127415e-06, 'samples': 24478720, 'steps': 47809, 'loss/train': 2.090956926345825}
+02/26/2022 15:30:01 - INFO - codeparrot_training - Step 47810: {'lr': 2.563727363291135e-06, 'samples': 24479232, 'steps': 47810, 'loss/train': 2.6381170749664307}
+02/26/2022 15:30:04 - INFO - codeparrot_training - Step 47811: {'lr': 2.5613905996079156e-06, 'samples': 24479744, 'steps': 47811, 'loss/train': 0.9159024357795715}
+02/26/2022 15:30:10 - INFO - codeparrot_training - Step 47812: {'lr': 2.5590548958731586e-06, 'samples': 24480256, 'steps': 47812, 'loss/train': 2.4449822902679443}
+02/26/2022 15:30:14 - INFO - codeparrot_training - Step 47813: {'lr': 2.556720252096828e-06, 'samples': 24480768, 'steps': 47813, 'loss/train': 1.6399636268615723}
+02/26/2022 15:30:19 - INFO - codeparrot_training - Step 47814: {'lr': 2.5543866682889437e-06, 'samples': 24481280, 'steps': 47814, 'loss/train': 1.7094807624816895}
+02/26/2022 15:30:23 - INFO - codeparrot_training - Step 47815: {'lr': 2.5520541444594703e-06, 'samples': 24481792, 'steps': 47815, 'loss/train': 0.7401649951934814}
+02/26/2022 15:30:30 - INFO - codeparrot_training - Step 47816: {'lr': 2.5497226806184548e-06, 'samples': 24482304, 'steps': 47816, 'loss/train': 0.8024858236312866}
+02/26/2022 15:30:33 - INFO - codeparrot_training - Step 47817: {'lr': 2.5473922767758063e-06, 'samples': 24482816, 'steps': 47817, 'loss/train': 2.027305841445923}
+02/26/2022 15:30:39 - INFO - codeparrot_training - Step 47818: {'lr': 2.5450629329416277e-06, 'samples': 24483328, 'steps': 47818, 'loss/train': 1.7449239492416382}
+02/26/2022 15:30:42 - INFO - codeparrot_training - Step 47819: {'lr': 2.5427346491257717e-06, 'samples': 24483840, 'steps': 47819, 'loss/train': 1.380434274673462}
+02/26/2022 15:30:48 - INFO - codeparrot_training - Step 47820: {'lr': 2.5404074253382592e-06, 'samples': 24484352, 'steps': 47820, 'loss/train': 0.8455139994621277}
+02/26/2022 15:30:52 - INFO - codeparrot_training - Step 47821: {'lr': 2.5380812615890813e-06, 'samples': 24484864, 'steps': 47821, 'loss/train': 2.0565526485443115}
+02/26/2022 15:30:57 - INFO - codeparrot_training - Step 47822: {'lr': 2.5357561578882026e-06, 'samples': 24485376, 'steps': 47822, 'loss/train': 2.365363597869873}
+02/26/2022 15:31:01 - INFO - codeparrot_training - Step 47823: {'lr': 2.533432114245532e-06, 'samples': 24485888, 'steps': 47823, 'loss/train': 0.7811762690544128}
+02/26/2022 15:31:06 - INFO - codeparrot_training - Step 47824: {'lr': 2.531109130671061e-06, 'samples': 24486400, 'steps': 47824, 'loss/train': 1.7682344913482666}
+02/26/2022 15:31:10 - INFO - codeparrot_training - Step 47825: {'lr': 2.528787207174754e-06, 'samples': 24486912, 'steps': 47825, 'loss/train': 0.8853068351745605}
+02/26/2022 15:31:15 - INFO - codeparrot_training - Step 47826: {'lr': 2.526466343766548e-06, 'samples': 24487424, 'steps': 47826, 'loss/train': 2.8831467628479004}
+02/26/2022 15:31:19 - INFO - codeparrot_training - Step 47827: {'lr': 2.5241465404563512e-06, 'samples': 24487936, 'steps': 47827, 'loss/train': 1.5609880685806274}
+02/26/2022 15:31:26 - INFO - codeparrot_training - Step 47828: {'lr': 2.5218277972541557e-06, 'samples': 24488448, 'steps': 47828, 'loss/train': 1.6217362880706787}
+02/26/2022 15:31:29 - INFO - codeparrot_training - Step 47829: {'lr': 2.5195101141698706e-06, 'samples': 24488960, 'steps': 47829, 'loss/train': 1.2341086864471436}
+02/26/2022 15:31:35 - INFO - codeparrot_training - Step 47830: {'lr': 2.5171934912133765e-06, 'samples': 24489472, 'steps': 47830, 'loss/train': 1.317106008529663}
+02/26/2022 15:31:38 - INFO - codeparrot_training - Step 47831: {'lr': 2.514877928394693e-06, 'samples': 24489984, 'steps': 47831, 'loss/train': 1.6443696022033691}
+02/26/2022 15:31:44 - INFO - codeparrot_training - Step 47832: {'lr': 2.512563425723646e-06, 'samples': 24490496, 'steps': 47832, 'loss/train': 1.7121108770370483}
+02/26/2022 15:31:47 - INFO - codeparrot_training - Step 47833: {'lr': 2.5102499832102275e-06, 'samples': 24491008, 'steps': 47833, 'loss/train': 2.020407199859619}
+02/26/2022 15:31:53 - INFO - codeparrot_training - Step 47834: {'lr': 2.5079376008642906e-06, 'samples': 24491520, 'steps': 47834, 'loss/train': 1.2551398277282715}
+02/26/2022 15:31:56 - INFO - codeparrot_training - Step 47835: {'lr': 2.505626278695744e-06, 'samples': 24492032, 'steps': 47835, 'loss/train': 2.0137217044830322}
+02/26/2022 15:32:02 - INFO - codeparrot_training - Step 47836: {'lr': 2.503316016714552e-06, 'samples': 24492544, 'steps': 47836, 'loss/train': 2.295029640197754}
+02/26/2022 15:32:05 - INFO - codeparrot_training - Step 47837: {'lr': 2.50100681493054e-06, 'samples': 24493056, 'steps': 47837, 'loss/train': 1.4651015996932983}
+02/26/2022 15:32:13 - INFO - codeparrot_training - Step 47838: {'lr': 2.4986986733536164e-06, 'samples': 24493568, 'steps': 47838, 'loss/train': 1.918484091758728}
+02/26/2022 15:32:16 - INFO - codeparrot_training - Step 47839: {'lr': 2.496391591993663e-06, 'samples': 24494080, 'steps': 47839, 'loss/train': 1.6724934577941895}
+02/26/2022 15:32:22 - INFO - codeparrot_training - Step 47840: {'lr': 2.494085570860616e-06, 'samples': 24494592, 'steps': 47840, 'loss/train': 1.6329292058944702}
+02/26/2022 15:32:25 - INFO - codeparrot_training - Step 47841: {'lr': 2.4917806099643003e-06, 'samples': 24495104, 'steps': 47841, 'loss/train': 2.21370005607605}
+02/26/2022 15:32:31 - INFO - codeparrot_training - Step 47842: {'lr': 2.4894767093145975e-06, 'samples': 24495616, 'steps': 47842, 'loss/train': 2.584339141845703}
+02/26/2022 15:32:34 - INFO - codeparrot_training - Step 47843: {'lr': 2.4871738689213886e-06, 'samples': 24496128, 'steps': 47843, 'loss/train': 1.3005484342575073}
+02/26/2022 15:32:39 - INFO - codeparrot_training - Step 47844: {'lr': 2.4848720887945263e-06, 'samples': 24496640, 'steps': 47844, 'loss/train': 1.1254960298538208}
+02/26/2022 15:32:43 - INFO - codeparrot_training - Step 47845: {'lr': 2.482571368943892e-06, 'samples': 24497152, 'steps': 47845, 'loss/train': 0.4927946925163269}
+02/26/2022 15:32:49 - INFO - codeparrot_training - Step 47846: {'lr': 2.4802717093792835e-06, 'samples': 24497664, 'steps': 47846, 'loss/train': 1.3502267599105835}
+02/26/2022 15:32:52 - INFO - codeparrot_training - Step 47847: {'lr': 2.4779731101106372e-06, 'samples': 24498176, 'steps': 47847, 'loss/train': 0.7948471307754517}
+02/26/2022 15:32:58 - INFO - codeparrot_training - Step 47848: {'lr': 2.4756755711477508e-06, 'samples': 24498688, 'steps': 47848, 'loss/train': 1.4448049068450928}
+02/26/2022 15:33:01 - INFO - codeparrot_training - Step 47849: {'lr': 2.4733790925004773e-06, 'samples': 24499200, 'steps': 47849, 'loss/train': 2.192993640899658}
+02/26/2022 15:33:07 - INFO - codeparrot_training - Step 47850: {'lr': 2.4710836741786145e-06, 'samples': 24499712, 'steps': 47850, 'loss/train': 2.073532819747925}
+02/26/2022 15:33:10 - INFO - codeparrot_training - Step 47851: {'lr': 2.4687893161920717e-06, 'samples': 24500224, 'steps': 47851, 'loss/train': 1.7523607015609741}
+02/26/2022 15:33:17 - INFO - codeparrot_training - Step 47852: {'lr': 2.466496018550618e-06, 'samples': 24500736, 'steps': 47852, 'loss/train': 2.029283285140991}
+02/26/2022 15:33:21 - INFO - codeparrot_training - Step 47853: {'lr': 2.464203781264107e-06, 'samples': 24501248, 'steps': 47853, 'loss/train': 2.250612735748291}
+02/26/2022 15:33:26 - INFO - codeparrot_training - Step 47854: {'lr': 2.461912604342309e-06, 'samples': 24501760, 'steps': 47854, 'loss/train': 1.9850369691848755}
+02/26/2022 15:33:30 - INFO - codeparrot_training - Step 47855: {'lr': 2.4596224877951044e-06, 'samples': 24502272, 'steps': 47855, 'loss/train': 2.4468674659729004}
+02/26/2022 15:33:35 - INFO - codeparrot_training - Step 47856: {'lr': 2.457333431632264e-06, 'samples': 24502784, 'steps': 47856, 'loss/train': 1.7805763483047485}
+02/26/2022 15:33:39 - INFO - codeparrot_training - Step 47857: {'lr': 2.45504543586364e-06, 'samples': 24503296, 'steps': 47857, 'loss/train': 0.8310919404029846}
+02/26/2022 15:33:45 - INFO - codeparrot_training - Step 47858: {'lr': 2.452758500498947e-06, 'samples': 24503808, 'steps': 47858, 'loss/train': 3.1714062690734863}
+02/26/2022 15:33:48 - INFO - codeparrot_training - Step 47859: {'lr': 2.450472625548067e-06, 'samples': 24504320, 'steps': 47859, 'loss/train': 2.0159599781036377}
+02/26/2022 15:33:54 - INFO - codeparrot_training - Step 47860: {'lr': 2.4481878110207134e-06, 'samples': 24504832, 'steps': 47860, 'loss/train': 2.3763718605041504}
+02/26/2022 15:33:57 - INFO - codeparrot_training - Step 47861: {'lr': 2.44590405692674e-06, 'samples': 24505344, 'steps': 47861, 'loss/train': 3.1528851985931396}
+02/26/2022 15:34:02 - INFO - codeparrot_training - Step 47862: {'lr': 2.443621363275916e-06, 'samples': 24505856, 'steps': 47862, 'loss/train': 1.9872863292694092}
+02/26/2022 15:34:06 - INFO - codeparrot_training - Step 47863: {'lr': 2.4413397300780127e-06, 'samples': 24506368, 'steps': 47863, 'loss/train': 1.5886955261230469}
+02/26/2022 15:34:13 - INFO - codeparrot_training - Step 47864: {'lr': 2.439059157342799e-06, 'samples': 24506880, 'steps': 47864, 'loss/train': 2.1267330646514893}
+02/26/2022 15:34:17 - INFO - codeparrot_training - Step 47865: {'lr': 2.4367796450800174e-06, 'samples': 24507392, 'steps': 47865, 'loss/train': 2.1475813388824463}
+02/26/2022 15:34:22 - INFO - codeparrot_training - Step 47866: {'lr': 2.434501193299493e-06, 'samples': 24507904, 'steps': 47866, 'loss/train': 1.3327138423919678}
+02/26/2022 15:34:25 - INFO - codeparrot_training - Step 47867: {'lr': 2.432223802010969e-06, 'samples': 24508416, 'steps': 47867, 'loss/train': 1.3871217966079712}
+02/26/2022 15:34:31 - INFO - codeparrot_training - Step 47868: {'lr': 2.429947471224131e-06, 'samples': 24508928, 'steps': 47868, 'loss/train': 2.093010902404785}
+02/26/2022 15:34:34 - INFO - codeparrot_training - Step 47869: {'lr': 2.4276722009488328e-06, 'samples': 24509440, 'steps': 47869, 'loss/train': 0.7602604627609253}
+02/26/2022 15:34:40 - INFO - codeparrot_training - Step 47870: {'lr': 2.425397991194761e-06, 'samples': 24509952, 'steps': 47870, 'loss/train': 1.63302743434906}
+02/26/2022 15:34:43 - INFO - codeparrot_training - Step 47871: {'lr': 2.4231248419716856e-06, 'samples': 24510464, 'steps': 47871, 'loss/train': 2.0227210521698}
+02/26/2022 15:34:49 - INFO - codeparrot_training - Step 47872: {'lr': 2.420852753289293e-06, 'samples': 24510976, 'steps': 47872, 'loss/train': 2.16332745552063}
+02/26/2022 15:34:52 - INFO - codeparrot_training - Step 47873: {'lr': 2.418581725157382e-06, 'samples': 24511488, 'steps': 47873, 'loss/train': 0.573132336139679}
+02/26/2022 15:35:00 - INFO - codeparrot_training - Step 47874: {'lr': 2.4163117575856376e-06, 'samples': 24512000, 'steps': 47874, 'loss/train': 1.449105978012085}
+02/26/2022 15:35:03 - INFO - codeparrot_training - Step 47875: {'lr': 2.414042850583803e-06, 'samples': 24512512, 'steps': 47875, 'loss/train': 2.2010200023651123}
+02/26/2022 15:35:09 - INFO - codeparrot_training - Step 47876: {'lr': 2.4117750041615926e-06, 'samples': 24513024, 'steps': 47876, 'loss/train': 1.3662270307540894}
+02/26/2022 15:35:12 - INFO - codeparrot_training - Step 47877: {'lr': 2.409508218328693e-06, 'samples': 24513536, 'steps': 47877, 'loss/train': 0.9514808654785156}
+02/26/2022 15:35:18 - INFO - codeparrot_training - Step 47878: {'lr': 2.4072424930948744e-06, 'samples': 24514048, 'steps': 47878, 'loss/train': 2.7840561866760254}
+02/26/2022 15:35:21 - INFO - codeparrot_training - Step 47879: {'lr': 2.4049778284697953e-06, 'samples': 24514560, 'steps': 47879, 'loss/train': 1.0101137161254883}
+02/26/2022 15:35:27 - INFO - codeparrot_training - Step 47880: {'lr': 2.4027142244631706e-06, 'samples': 24515072, 'steps': 47880, 'loss/train': 1.3735039234161377}
+02/26/2022 15:35:30 - INFO - codeparrot_training - Step 47881: {'lr': 2.4004516810846864e-06, 'samples': 24515584, 'steps': 47881, 'loss/train': 1.14693284034729}
+02/26/2022 15:35:36 - INFO - codeparrot_training - Step 47882: {'lr': 2.3981901983440855e-06, 'samples': 24516096, 'steps': 47882, 'loss/train': 1.460396647453308}
+02/26/2022 15:35:39 - INFO - codeparrot_training - Step 47883: {'lr': 2.3959297762509714e-06, 'samples': 24516608, 'steps': 47883, 'loss/train': 1.0896514654159546}
+02/26/2022 15:35:46 - INFO - codeparrot_training - Step 47884: {'lr': 2.3936704148151134e-06, 'samples': 24517120, 'steps': 47884, 'loss/train': 2.7399637699127197}
+02/26/2022 15:35:50 - INFO - codeparrot_training - Step 47885: {'lr': 2.3914121140460877e-06, 'samples': 24517632, 'steps': 47885, 'loss/train': 2.3188326358795166}
+02/26/2022 15:35:55 - INFO - codeparrot_training - Step 47886: {'lr': 2.3891548739536918e-06, 'samples': 24518144, 'steps': 47886, 'loss/train': 1.9701364040374756}
+02/26/2022 15:35:59 - INFO - codeparrot_training - Step 47887: {'lr': 2.3868986945474735e-06, 'samples': 24518656, 'steps': 47887, 'loss/train': 0.6902536749839783}
+02/26/2022 15:36:04 - INFO - codeparrot_training - Step 47888: {'lr': 2.3846435758372033e-06, 'samples': 24519168, 'steps': 47888, 'loss/train': 1.5973650217056274}
+02/26/2022 15:36:08 - INFO - codeparrot_training - Step 47889: {'lr': 2.382389517832456e-06, 'samples': 24519680, 'steps': 47889, 'loss/train': 1.7861621379852295}
+02/26/2022 15:36:13 - INFO - codeparrot_training - Step 47890: {'lr': 2.380136520542947e-06, 'samples': 24520192, 'steps': 47890, 'loss/train': 1.3621666431427002}
+02/26/2022 15:36:17 - INFO - codeparrot_training - Step 47891: {'lr': 2.377884583978307e-06, 'samples': 24520704, 'steps': 47891, 'loss/train': 1.4045051336288452}
+02/26/2022 15:36:22 - INFO - codeparrot_training - Step 47892: {'lr': 2.375633708148195e-06, 'samples': 24521216, 'steps': 47892, 'loss/train': 1.2694860696792603}
+02/26/2022 15:36:26 - INFO - codeparrot_training - Step 47893: {'lr': 2.3733838930621865e-06, 'samples': 24521728, 'steps': 47893, 'loss/train': 2.413146495819092}
+02/26/2022 15:36:31 - INFO - codeparrot_training - Step 47894: {'lr': 2.371135138730024e-06, 'samples': 24522240, 'steps': 47894, 'loss/train': 1.6571993827819824}
+02/26/2022 15:36:37 - INFO - codeparrot_training - Step 47895: {'lr': 2.368887445161255e-06, 'samples': 24522752, 'steps': 47895, 'loss/train': 1.3963418006896973}
+02/26/2022 15:36:40 - INFO - codeparrot_training - Step 47896: {'lr': 2.3666408123655393e-06, 'samples': 24523264, 'steps': 47896, 'loss/train': 8.758283615112305}
+02/26/2022 15:36:44 - INFO - codeparrot_training - Step 47897: {'lr': 2.364395240352535e-06, 'samples': 24523776, 'steps': 47897, 'loss/train': 1.697182059288025}
+02/26/2022 15:36:49 - INFO - codeparrot_training - Step 47898: {'lr': 2.36215072913179e-06, 'samples': 24524288, 'steps': 47898, 'loss/train': 1.6604324579238892}
+02/26/2022 15:36:53 - INFO - codeparrot_training - Step 47899: {'lr': 2.3599072787129637e-06, 'samples': 24524800, 'steps': 47899, 'loss/train': 0.9347661137580872}
+02/26/2022 15:37:00 - INFO - codeparrot_training - Step 47900: {'lr': 2.357664889105687e-06, 'samples': 24525312, 'steps': 47900, 'loss/train': 1.2365621328353882}
+02/26/2022 15:37:04 - INFO - codeparrot_training - Step 47901: {'lr': 2.3554235603195362e-06, 'samples': 24525824, 'steps': 47901, 'loss/train': 0.9771762490272522}
+02/26/2022 15:37:09 - INFO - codeparrot_training - Step 47902: {'lr': 2.3531832923640862e-06, 'samples': 24526336, 'steps': 47902, 'loss/train': 0.9857110381126404}
+02/26/2022 15:37:13 - INFO - codeparrot_training - Step 47903: {'lr': 2.3509440852489683e-06, 'samples': 24526848, 'steps': 47903, 'loss/train': 1.839868426322937}
+02/26/2022 15:37:18 - INFO - codeparrot_training - Step 47904: {'lr': 2.348705938983786e-06, 'samples': 24527360, 'steps': 47904, 'loss/train': 2.1556990146636963}
+02/26/2022 15:37:24 - INFO - codeparrot_training - Step 47905: {'lr': 2.3464688535781155e-06, 'samples': 24527872, 'steps': 47905, 'loss/train': 1.6833938360214233}
+02/26/2022 15:37:27 - INFO - codeparrot_training - Step 47906: {'lr': 2.3442328290415037e-06, 'samples': 24528384, 'steps': 47906, 'loss/train': 1.6223562955856323}
+02/26/2022 15:37:33 - INFO - codeparrot_training - Step 47907: {'lr': 2.3419978653835826e-06, 'samples': 24528896, 'steps': 47907, 'loss/train': 1.238167643547058}
+02/26/2022 15:37:36 - INFO - codeparrot_training - Step 47908: {'lr': 2.3397639626139002e-06, 'samples': 24529408, 'steps': 47908, 'loss/train': 1.3795874118804932}
+02/26/2022 15:37:44 - INFO - codeparrot_training - Step 47909: {'lr': 2.3375311207420037e-06, 'samples': 24529920, 'steps': 47909, 'loss/train': 0.9087982773780823}
+02/26/2022 15:37:47 - INFO - codeparrot_training - Step 47910: {'lr': 2.3352993397775247e-06, 'samples': 24530432, 'steps': 47910, 'loss/train': 1.6121654510498047}
+02/26/2022 15:37:53 - INFO - codeparrot_training - Step 47911: {'lr': 2.3330686197299276e-06, 'samples': 24530944, 'steps': 47911, 'loss/train': 1.5290254354476929}
+02/26/2022 15:37:56 - INFO - codeparrot_training - Step 47912: {'lr': 2.330838960608872e-06, 'samples': 24531456, 'steps': 47912, 'loss/train': 1.2620149850845337}
+02/26/2022 15:38:02 - INFO - codeparrot_training - Step 47913: {'lr': 2.328610362423822e-06, 'samples': 24531968, 'steps': 47913, 'loss/train': 0.2835817039012909}
+02/26/2022 15:38:05 - INFO - codeparrot_training - Step 47914: {'lr': 2.326382825184381e-06, 'samples': 24532480, 'steps': 47914, 'loss/train': 2.129727602005005}
+02/26/2022 15:38:11 - INFO - codeparrot_training - Step 47915: {'lr': 2.3241563489000692e-06, 'samples': 24532992, 'steps': 47915, 'loss/train': 1.2792174816131592}
+02/26/2022 15:38:14 - INFO - codeparrot_training - Step 47916: {'lr': 2.3219309335804073e-06, 'samples': 24533504, 'steps': 47916, 'loss/train': 0.8699167370796204}
+02/26/2022 15:38:20 - INFO - codeparrot_training - Step 47917: {'lr': 2.31970657923497e-06, 'samples': 24534016, 'steps': 47917, 'loss/train': 2.230705976486206}
+02/26/2022 15:38:23 - INFO - codeparrot_training - Step 47918: {'lr': 2.317483285873251e-06, 'samples': 24534528, 'steps': 47918, 'loss/train': 0.9881263375282288}
+02/26/2022 15:38:30 - INFO - codeparrot_training - Step 47919: {'lr': 2.315261053504797e-06, 'samples': 24535040, 'steps': 47919, 'loss/train': 2.4287497997283936}
+02/26/2022 15:38:34 - INFO - codeparrot_training - Step 47920: {'lr': 2.3130398821391008e-06, 'samples': 24535552, 'steps': 47920, 'loss/train': 2.214642286300659}
+02/26/2022 15:38:39 - INFO - codeparrot_training - Step 47921: {'lr': 2.3108197717857105e-06, 'samples': 24536064, 'steps': 47921, 'loss/train': 1.6973623037338257}
+02/26/2022 15:38:43 - INFO - codeparrot_training - Step 47922: {'lr': 2.3086007224540906e-06, 'samples': 24536576, 'steps': 47922, 'loss/train': 2.3015148639678955}
+02/26/2022 15:38:48 - INFO - codeparrot_training - Step 47923: {'lr': 2.306382734153817e-06, 'samples': 24537088, 'steps': 47923, 'loss/train': 1.8384315967559814}
+02/26/2022 15:38:52 - INFO - codeparrot_training - Step 47924: {'lr': 2.3041658068942984e-06, 'samples': 24537600, 'steps': 47924, 'loss/train': 2.262763023376465}
+02/26/2022 15:38:57 - INFO - codeparrot_training - Step 47925: {'lr': 2.3019499406851106e-06, 'samples': 24538112, 'steps': 47925, 'loss/train': 1.6985430717468262}
+02/26/2022 15:39:01 - INFO - codeparrot_training - Step 47926: {'lr': 2.299735135535691e-06, 'samples': 24538624, 'steps': 47926, 'loss/train': 1.980005145072937}
+02/26/2022 15:39:06 - INFO - codeparrot_training - Step 47927: {'lr': 2.297521391455587e-06, 'samples': 24539136, 'steps': 47927, 'loss/train': 1.6654679775238037}
+02/26/2022 15:39:10 - INFO - codeparrot_training - Step 47928: {'lr': 2.2953087084542355e-06, 'samples': 24539648, 'steps': 47928, 'loss/train': 1.1862330436706543}
+02/26/2022 15:39:17 - INFO - codeparrot_training - Step 47929: {'lr': 2.2930970865411015e-06, 'samples': 24540160, 'steps': 47929, 'loss/train': 2.939847469329834}
+02/26/2022 15:39:20 - INFO - codeparrot_training - Step 47930: {'lr': 2.290886525725705e-06, 'samples': 24540672, 'steps': 47930, 'loss/train': 2.215400457382202}
+02/26/2022 15:39:26 - INFO - codeparrot_training - Step 47931: {'lr': 2.2886770260175104e-06, 'samples': 24541184, 'steps': 47931, 'loss/train': 1.4276045560836792}
+02/26/2022 15:39:29 - INFO - codeparrot_training - Step 47932: {'lr': 2.2864685874259273e-06, 'samples': 24541696, 'steps': 47932, 'loss/train': 1.2141623497009277}
+02/26/2022 15:39:35 - INFO - codeparrot_training - Step 47933: {'lr': 2.2842612099604753e-06, 'samples': 24542208, 'steps': 47933, 'loss/train': 1.406009316444397}
+02/26/2022 15:39:38 - INFO - codeparrot_training - Step 47934: {'lr': 2.2820548936305918e-06, 'samples': 24542720, 'steps': 47934, 'loss/train': 1.1579920053482056}
+02/26/2022 15:39:44 - INFO - codeparrot_training - Step 47935: {'lr': 2.2798496384457414e-06, 'samples': 24543232, 'steps': 47935, 'loss/train': 1.4007172584533691}
+02/26/2022 15:39:47 - INFO - codeparrot_training - Step 47936: {'lr': 2.2776454444153326e-06, 'samples': 24543744, 'steps': 47936, 'loss/train': 1.006408452987671}
+02/26/2022 15:39:53 - INFO - codeparrot_training - Step 47937: {'lr': 2.275442311548831e-06, 'samples': 24544256, 'steps': 47937, 'loss/train': 1.9216763973236084}
+02/26/2022 15:39:56 - INFO - codeparrot_training - Step 47938: {'lr': 2.2732402398556727e-06, 'samples': 24544768, 'steps': 47938, 'loss/train': 0.3714328706264496}
+02/26/2022 15:40:02 - INFO - codeparrot_training - Step 47939: {'lr': 2.2710392293452954e-06, 'samples': 24545280, 'steps': 47939, 'loss/train': 2.0875027179718018}
+02/26/2022 15:40:05 - INFO - codeparrot_training - Step 47940: {'lr': 2.268839280027135e-06, 'samples': 24545792, 'steps': 47940, 'loss/train': 1.7391934394836426}
+02/26/2022 15:40:11 - INFO - codeparrot_training - Step 47941: {'lr': 2.2666403919106016e-06, 'samples': 24546304, 'steps': 47941, 'loss/train': 2.399202585220337}
+02/26/2022 15:40:14 - INFO - codeparrot_training - Step 47942: {'lr': 2.264442565005076e-06, 'samples': 24546816, 'steps': 47942, 'loss/train': 1.592761516571045}
+02/26/2022 15:40:20 - INFO - codeparrot_training - Step 47943: {'lr': 2.262245799320051e-06, 'samples': 24547328, 'steps': 47943, 'loss/train': 0.9611049294471741}
+02/26/2022 15:40:23 - INFO - codeparrot_training - Step 47944: {'lr': 2.2600500948648793e-06, 'samples': 24547840, 'steps': 47944, 'loss/train': 1.998374342918396}
+02/26/2022 15:40:31 - INFO - codeparrot_training - Step 47945: {'lr': 2.257855451649027e-06, 'samples': 24548352, 'steps': 47945, 'loss/train': 1.0205129384994507}
+02/26/2022 15:40:34 - INFO - codeparrot_training - Step 47946: {'lr': 2.2556618696818187e-06, 'samples': 24548864, 'steps': 47946, 'loss/train': 1.6476207971572876}
+02/26/2022 15:40:40 - INFO - codeparrot_training - Step 47947: {'lr': 2.25346934897272e-06, 'samples': 24549376, 'steps': 47947, 'loss/train': 2.0553367137908936}
+02/26/2022 15:40:43 - INFO - codeparrot_training - Step 47948: {'lr': 2.251277889531056e-06, 'samples': 24549888, 'steps': 47948, 'loss/train': 1.4443731307983398}
+02/26/2022 15:40:49 - INFO - codeparrot_training - Step 47949: {'lr': 2.249087491366264e-06, 'samples': 24550400, 'steps': 47949, 'loss/train': 2.54288387298584}
+02/26/2022 15:40:52 - INFO - codeparrot_training - Step 47950: {'lr': 2.2468981544876978e-06, 'samples': 24550912, 'steps': 47950, 'loss/train': 1.1472249031066895}
+02/26/2022 15:40:58 - INFO - codeparrot_training - Step 47951: {'lr': 2.2447098789047658e-06, 'samples': 24551424, 'steps': 47951, 'loss/train': 1.5696851015090942}
+02/26/2022 15:41:01 - INFO - codeparrot_training - Step 47952: {'lr': 2.2425226646268228e-06, 'samples': 24551936, 'steps': 47952, 'loss/train': 0.9127157926559448}
+02/26/2022 15:41:07 - INFO - codeparrot_training - Step 47953: {'lr': 2.240336511663249e-06, 'samples': 24552448, 'steps': 47953, 'loss/train': 1.50718355178833}
+02/26/2022 15:41:10 - INFO - codeparrot_training - Step 47954: {'lr': 2.238151420023371e-06, 'samples': 24552960, 'steps': 47954, 'loss/train': 1.2241694927215576}
+02/26/2022 15:41:18 - INFO - codeparrot_training - Step 47955: {'lr': 2.235967389716598e-06, 'samples': 24553472, 'steps': 47955, 'loss/train': 1.1092032194137573}
+02/26/2022 15:41:21 - INFO - codeparrot_training - Step 47956: {'lr': 2.233784420752255e-06, 'samples': 24553984, 'steps': 47956, 'loss/train': 1.4935420751571655}
+02/26/2022 15:41:27 - INFO - codeparrot_training - Step 47957: {'lr': 2.2316025131397245e-06, 'samples': 24554496, 'steps': 47957, 'loss/train': 2.387575149536133}
+02/26/2022 15:41:30 - INFO - codeparrot_training - Step 47958: {'lr': 2.2294216668883315e-06, 'samples': 24555008, 'steps': 47958, 'loss/train': 2.4424500465393066}
+02/26/2022 15:41:36 - INFO - codeparrot_training - Step 47959: {'lr': 2.2272418820074024e-06, 'samples': 24555520, 'steps': 47959, 'loss/train': 1.5509511232376099}
+02/26/2022 15:41:39 - INFO - codeparrot_training - Step 47960: {'lr': 2.2250631585063187e-06, 'samples': 24556032, 'steps': 47960, 'loss/train': 0.11174685508012772}
+02/26/2022 15:41:45 - INFO - codeparrot_training - Step 47961: {'lr': 2.22288549639435e-06, 'samples': 24556544, 'steps': 47961, 'loss/train': 2.091057538986206}
+02/26/2022 15:41:48 - INFO - codeparrot_training - Step 47962: {'lr': 2.2207088956808785e-06, 'samples': 24557056, 'steps': 47962, 'loss/train': 1.0130252838134766}
+02/26/2022 15:41:54 - INFO - codeparrot_training - Step 47963: {'lr': 2.2185333563752296e-06, 'samples': 24557568, 'steps': 47963, 'loss/train': 3.0432074069976807}
+02/26/2022 15:41:57 - INFO - codeparrot_training - Step 47964: {'lr': 2.2163588784867015e-06, 'samples': 24558080, 'steps': 47964, 'loss/train': 1.931697130203247}
+02/26/2022 15:42:04 - INFO - codeparrot_training - Step 47965: {'lr': 2.2141854620245927e-06, 'samples': 24558592, 'steps': 47965, 'loss/train': 1.3123613595962524}
+02/26/2022 15:42:08 - INFO - codeparrot_training - Step 47966: {'lr': 2.2120131069982563e-06, 'samples': 24559104, 'steps': 47966, 'loss/train': 1.4895377159118652}
+02/26/2022 15:42:13 - INFO - codeparrot_training - Step 47967: {'lr': 2.209841813416935e-06, 'samples': 24559616, 'steps': 47967, 'loss/train': 0.5027288794517517}
+02/26/2022 15:42:17 - INFO - codeparrot_training - Step 47968: {'lr': 2.2076715812900107e-06, 'samples': 24560128, 'steps': 47968, 'loss/train': 1.7456347942352295}
+02/26/2022 15:42:22 - INFO - codeparrot_training - Step 47969: {'lr': 2.2055024106266976e-06, 'samples': 24560640, 'steps': 47969, 'loss/train': 2.1062068939208984}
+02/26/2022 15:42:25 - INFO - codeparrot_training - Step 47970: {'lr': 2.2033343014363494e-06, 'samples': 24561152, 'steps': 47970, 'loss/train': 1.4630951881408691}
+02/26/2022 15:42:31 - INFO - codeparrot_training - Step 47971: {'lr': 2.201167253728237e-06, 'samples': 24561664, 'steps': 47971, 'loss/train': 1.4736186265945435}
+02/26/2022 15:42:35 - INFO - codeparrot_training - Step 47972: {'lr': 2.199001267511658e-06, 'samples': 24562176, 'steps': 47972, 'loss/train': 1.333685278892517}
+02/26/2022 15:42:40 - INFO - codeparrot_training - Step 47973: {'lr': 2.196836342795827e-06, 'samples': 24562688, 'steps': 47973, 'loss/train': 0.21811817586421967}
+02/26/2022 15:42:44 - INFO - codeparrot_training - Step 47974: {'lr': 2.1946724795900984e-06, 'samples': 24563200, 'steps': 47974, 'loss/train': 2.17622709274292}
+02/26/2022 15:42:51 - INFO - codeparrot_training - Step 47975: {'lr': 2.1925096779036868e-06, 'samples': 24563712, 'steps': 47975, 'loss/train': 1.3610150814056396}
+02/26/2022 15:42:55 - INFO - codeparrot_training - Step 47976: {'lr': 2.19034793774589e-06, 'samples': 24564224, 'steps': 47976, 'loss/train': 1.4638632535934448}
+02/26/2022 15:43:00 - INFO - codeparrot_training - Step 47977: {'lr': 2.188187259125951e-06, 'samples': 24564736, 'steps': 47977, 'loss/train': 0.938526451587677}
+02/26/2022 15:43:04 - INFO - codeparrot_training - Step 47978: {'lr': 2.1860276420531124e-06, 'samples': 24565248, 'steps': 47978, 'loss/train': 0.755862832069397}
+02/26/2022 15:43:09 - INFO - codeparrot_training - Step 47979: {'lr': 2.183869086536672e-06, 'samples': 24565760, 'steps': 47979, 'loss/train': 2.3575170040130615}
+02/26/2022 15:43:13 - INFO - codeparrot_training - Step 47980: {'lr': 2.181711592585817e-06, 'samples': 24566272, 'steps': 47980, 'loss/train': 1.8960074186325073}
+02/26/2022 15:43:18 - INFO - codeparrot_training - Step 47981: {'lr': 2.1795551602098175e-06, 'samples': 24566784, 'steps': 47981, 'loss/train': 0.8626107573509216}
+02/26/2022 15:43:22 - INFO - codeparrot_training - Step 47982: {'lr': 2.1773997894179444e-06, 'samples': 24567296, 'steps': 47982, 'loss/train': 1.5351321697235107}
+02/26/2022 15:43:27 - INFO - codeparrot_training - Step 47983: {'lr': 2.1752454802193567e-06, 'samples': 24567808, 'steps': 47983, 'loss/train': 1.9386399984359741}
+02/26/2022 15:43:31 - INFO - codeparrot_training - Step 47984: {'lr': 2.1730922326233804e-06, 'samples': 24568320, 'steps': 47984, 'loss/train': 1.9180781841278076}
+02/26/2022 15:43:36 - INFO - codeparrot_training - Step 47985: {'lr': 2.1709400466391194e-06, 'samples': 24568832, 'steps': 47985, 'loss/train': 1.9001375436782837}
+02/26/2022 15:43:42 - INFO - codeparrot_training - Step 47986: {'lr': 2.1687889222758995e-06, 'samples': 24569344, 'steps': 47986, 'loss/train': 1.4409270286560059}
+02/26/2022 15:43:45 - INFO - codeparrot_training - Step 47987: {'lr': 2.166638859542852e-06, 'samples': 24569856, 'steps': 47987, 'loss/train': 0.44483739137649536}
+02/26/2022 15:43:51 - INFO - codeparrot_training - Step 47988: {'lr': 2.1644898584492756e-06, 'samples': 24570368, 'steps': 47988, 'loss/train': 1.263789176940918}
+02/26/2022 15:43:54 - INFO - codeparrot_training - Step 47989: {'lr': 2.1623419190043014e-06, 'samples': 24570880, 'steps': 47989, 'loss/train': 1.4638453722000122}
+02/26/2022 15:44:01 - INFO - codeparrot_training - Step 47990: {'lr': 2.1601950412171723e-06, 'samples': 24571392, 'steps': 47990, 'loss/train': 0.4411675035953522}
+02/26/2022 15:44:05 - INFO - codeparrot_training - Step 47991: {'lr': 2.158049225097047e-06, 'samples': 24571904, 'steps': 47991, 'loss/train': 2.0408377647399902}
+02/26/2022 15:44:11 - INFO - codeparrot_training - Step 47992: {'lr': 2.1559044706531415e-06, 'samples': 24572416, 'steps': 47992, 'loss/train': 0.5840863585472107}
+02/26/2022 15:44:14 - INFO - codeparrot_training - Step 47993: {'lr': 2.15376077789467e-06, 'samples': 24572928, 'steps': 47993, 'loss/train': 0.5253375172615051}
+02/26/2022 15:44:20 - INFO - codeparrot_training - Step 47994: {'lr': 2.151618146830764e-06, 'samples': 24573440, 'steps': 47994, 'loss/train': 2.518958330154419}
+02/26/2022 15:44:23 - INFO - codeparrot_training - Step 47995: {'lr': 2.149476577470638e-06, 'samples': 24573952, 'steps': 47995, 'loss/train': 1.8539888858795166}
+02/26/2022 15:44:29 - INFO - codeparrot_training - Step 47996: {'lr': 2.1473360698234245e-06, 'samples': 24574464, 'steps': 47996, 'loss/train': 1.164291262626648}
+02/26/2022 15:44:32 - INFO - codeparrot_training - Step 47997: {'lr': 2.1451966238983656e-06, 'samples': 24574976, 'steps': 47997, 'loss/train': 1.8377363681793213}
+02/26/2022 15:44:38 - INFO - codeparrot_training - Step 47998: {'lr': 2.143058239704565e-06, 'samples': 24575488, 'steps': 47998, 'loss/train': 2.02597975730896}
+02/26/2022 15:44:41 - INFO - codeparrot_training - Step 47999: {'lr': 2.140920917251182e-06, 'samples': 24576000, 'steps': 47999, 'loss/train': 1.8916027545928955}
+02/26/2022 15:44:41 - INFO - codeparrot_training - Evaluating and saving model checkpoint