dzungpham/graphcodebert-code-classification

1{2  "best_global_step": 1000,3  "best_metric": 0.7549859375827388,4  "best_model_checkpoint": "./output_checkpoints/graphcodebert-robust/checkpoint-1000",5  "epoch": 0.0768,6  "eval_steps": 1000,7  "global_step": 1200,8  "is_hyper_param_search": false,9  "is_local_process_zero": true,10  "is_world_process_zero": true,11  "log_history": [12    {13      "epoch": 0.00064,14      "grad_norm": 1.6144306659698486,15      "learning_rate": 1.1520000000000002e-08,16      "loss": 0.729,17      "step": 1018    },19    {20      "epoch": 0.00128,21      "grad_norm": 2.0952296257019043,22      "learning_rate": 2.4320000000000002e-08,23      "loss": 0.7295,24      "step": 2025    },26    {27      "epoch": 0.00192,28      "grad_norm": 1.3587689399719238,29      "learning_rate": 3.7120000000000004e-08,30      "loss": 0.73,31      "step": 3032    },33    {34      "epoch": 0.00256,35      "grad_norm": 1.2531732320785522,36      "learning_rate": 4.9920000000000006e-08,37      "loss": 0.7221,38      "step": 4039    },40    {41      "epoch": 0.0032,42      "grad_norm": 1.437932014465332,43      "learning_rate": 6.272000000000001e-08,44      "loss": 0.7209,45      "step": 5046    },47    {48      "epoch": 0.00384,49      "grad_norm": 1.418426752090454,50      "learning_rate": 7.552e-08,51      "loss": 0.729,52      "step": 6053    },54    {55      "epoch": 0.00448,56      "grad_norm": 1.9476298093795776,57      "learning_rate": 8.832e-08,58      "loss": 0.7242,59      "step": 7060    },61    {62      "epoch": 0.00512,63      "grad_norm": 1.7948051691055298,64      "learning_rate": 1.0112000000000001e-07,65      "loss": 0.7227,66      "step": 8067    },68    {69      "epoch": 0.00576,70      "grad_norm": 1.6534360647201538,71      "learning_rate": 1.1392e-07,72      "loss": 0.7234,73      "step": 9074    },75    {76      "epoch": 0.0064,77      "grad_norm": 1.0920158624649048,78      "learning_rate": 1.2672e-07,79      "loss": 0.7328,80      "step": 10081    },82    {83      "epoch": 0.00704,84      "grad_norm": 1.977837085723877,85      "learning_rate": 1.3952000000000002e-07,86      "loss": 0.7263,87      "step": 11088    },89    {90      "epoch": 0.00768,91      "grad_norm": 1.388983130455017,92      "learning_rate": 1.5232000000000003e-07,93      "loss": 0.7286,94      "step": 12095    },96    {97      "epoch": 0.00832,98      "grad_norm": 1.2956682443618774,99      "learning_rate": 1.6512e-07,100      "loss": 0.7251,101      "step": 130102    },103    {104      "epoch": 0.00896,105      "grad_norm": 1.8125052452087402,106      "learning_rate": 1.7792e-07,107      "loss": 0.7251,108      "step": 140109    },110    {111      "epoch": 0.0096,112      "grad_norm": 1.626846194267273,113      "learning_rate": 1.9072e-07,114      "loss": 0.727,115      "step": 150116    },117    {118      "epoch": 0.01024,119      "grad_norm": 2.3243086338043213,120      "learning_rate": 2.0352e-07,121      "loss": 0.726,122      "step": 160123    },124    {125      "epoch": 0.01088,126      "grad_norm": 1.4734737873077393,127      "learning_rate": 2.1632e-07,128      "loss": 0.7252,129      "step": 170130    },131    {132      "epoch": 0.01152,133      "grad_norm": 2.090498685836792,134      "learning_rate": 2.2912e-07,135      "loss": 0.7273,136      "step": 180137    },138    {139      "epoch": 0.01216,140      "grad_norm": 1.7563093900680542,141      "learning_rate": 2.4192000000000004e-07,142      "loss": 0.719,143      "step": 190144    },145    {146      "epoch": 0.0128,147      "grad_norm": 1.449843168258667,148      "learning_rate": 2.5472000000000005e-07,149      "loss": 0.7237,150      "step": 200151    },152    {153      "epoch": 0.01344,154      "grad_norm": 141396.296875,155      "learning_rate": 5.350742447516642e-07,156      "loss": 0.7217,157      "step": 210158    },159    {160      "epoch": 0.01408,161      "grad_norm": 102339.1640625,162      "learning_rate": 5.606758832565284e-07,163      "loss": 0.7215,164      "step": 220165    },166    {167      "epoch": 0.01472,168      "grad_norm": 134052.9375,169      "learning_rate": 5.862775217613928e-07,170      "loss": 0.7115,171      "step": 230172    },173    {174      "epoch": 0.01536,175      "grad_norm": 87181.984375,176      "learning_rate": 6.118791602662571e-07,177      "loss": 0.7241,178      "step": 240179    },180    {181      "epoch": 0.016,182      "grad_norm": 100231.328125,183      "learning_rate": 6.374807987711214e-07,184      "loss": 0.71,185      "step": 250186    },187    {188      "epoch": 0.01664,189      "grad_norm": 136721.484375,190      "learning_rate": 6.630824372759858e-07,191      "loss": 0.7188,192      "step": 260193    },194    {195      "epoch": 0.01728,196      "grad_norm": 115868.8125,197      "learning_rate": 6.8868407578085e-07,198      "loss": 0.7199,199      "step": 270200    },201    {202      "epoch": 0.01792,203      "grad_norm": 70205.1484375,204      "learning_rate": 7.142857142857143e-07,205      "loss": 0.7299,206      "step": 280207    },208    {209      "epoch": 0.01856,210      "grad_norm": 98926.4453125,211      "learning_rate": 7.398873527905787e-07,212      "loss": 0.7159,213      "step": 290214    },215    {216      "epoch": 0.0192,217      "grad_norm": 134108.140625,218      "learning_rate": 7.65488991295443e-07,219      "loss": 0.7122,220      "step": 300221    },222    {223      "epoch": 0.01984,224      "grad_norm": 103719.140625,225      "learning_rate": 7.910906298003073e-07,226      "loss": 0.7185,227      "step": 310228    },229    {230      "epoch": 0.02048,231      "grad_norm": 85624.953125,232      "learning_rate": 8.166922683051716e-07,233      "loss": 0.718,234      "step": 320235    },236    {237      "epoch": 0.02112,238      "grad_norm": 138824.15625,239      "learning_rate": 8.422939068100359e-07,240      "loss": 0.713,241      "step": 330242    },243    {244      "epoch": 0.02176,245      "grad_norm": 73629.0859375,246      "learning_rate": 8.678955453149002e-07,247      "loss": 0.7186,248      "step": 340249    },250    {251      "epoch": 0.0224,252      "grad_norm": 132493.0,253      "learning_rate": 8.934971838197646e-07,254      "loss": 0.7133,255      "step": 350256    },257    {258      "epoch": 0.02304,259      "grad_norm": 85223.625,260      "learning_rate": 9.190988223246289e-07,261      "loss": 0.7124,262      "step": 360263    },264    {265      "epoch": 0.02368,266      "grad_norm": 77868.78125,267      "learning_rate": 9.447004608294931e-07,268      "loss": 0.7058,269      "step": 370270    },271    {272      "epoch": 0.02432,273      "grad_norm": 75874.3046875,274      "learning_rate": 9.703020993343575e-07,275      "loss": 0.7139,276      "step": 380277    },278    {279      "epoch": 0.02496,280      "grad_norm": 151937.703125,281      "learning_rate": 9.959037378392218e-07,282      "loss": 0.713,283      "step": 390284    },285    {286      "epoch": 0.0256,287      "grad_norm": 161711.671875,288      "learning_rate": 1.021505376344086e-06,289      "loss": 0.7137,290      "step": 400291    },292    {293      "epoch": 0.02624,294      "grad_norm": 90800.234375,295      "learning_rate": 1.0471070148489503e-06,296      "loss": 0.7091,297      "step": 410298    },299    {300      "epoch": 0.02688,301      "grad_norm": 82131.34375,302      "learning_rate": 1.0727086533538148e-06,303      "loss": 0.7098,304      "step": 420305    },306    {307      "epoch": 0.02752,308      "grad_norm": 92818.9140625,309      "learning_rate": 1.0983102918586791e-06,310      "loss": 0.7099,311      "step": 430312    },313    {314      "epoch": 0.02816,315      "grad_norm": 88555.5078125,316      "learning_rate": 1.1239119303635434e-06,317      "loss": 0.7086,318      "step": 440319    },320    {321      "epoch": 0.0288,322      "grad_norm": 73428.6015625,323      "learning_rate": 1.1495135688684077e-06,324      "loss": 0.7117,325      "step": 450326    },327    {328      "epoch": 0.02944,329      "grad_norm": 128938.7421875,330      "learning_rate": 1.175115207373272e-06,331      "loss": 0.7182,332      "step": 460333    },334    {335      "epoch": 0.03008,336      "grad_norm": 102742.3359375,337      "learning_rate": 1.2007168458781362e-06,338      "loss": 0.7108,339      "step": 470340    },341    {342      "epoch": 0.03072,343      "grad_norm": 73825.8125,344      "learning_rate": 1.2263184843830007e-06,345      "loss": 0.7087,346      "step": 480347    },348    {349      "epoch": 0.03136,350      "grad_norm": 110930.75,351      "learning_rate": 1.251920122887865e-06,352      "loss": 0.7232,353      "step": 490354    },355    {356      "epoch": 0.032,357      "grad_norm": 95068.84375,358      "learning_rate": 1.2775217613927293e-06,359      "loss": 0.703,360      "step": 500361    },362    {363      "epoch": 0.03264,364      "grad_norm": 118731.9296875,365      "learning_rate": 1.3031233998975938e-06,366      "loss": 0.7063,367      "step": 510368    },369    {370      "epoch": 0.03328,371      "grad_norm": 80511.828125,372      "learning_rate": 1.3287250384024578e-06,373      "loss": 0.7143,374      "step": 520375    },376    {377      "epoch": 0.03392,378      "grad_norm": 84864.484375,379      "learning_rate": 1.354326676907322e-06,380      "loss": 0.7055,381      "step": 530382    },383    {384      "epoch": 0.03456,385      "grad_norm": 107800.109375,386      "learning_rate": 1.3799283154121864e-06,387      "loss": 0.7119,388      "step": 540389    },390    {391      "epoch": 0.0352,392      "grad_norm": 83667.671875,393      "learning_rate": 1.4055299539170509e-06,394      "loss": 0.7082,395      "step": 550396    },397    {398      "epoch": 0.03584,399      "grad_norm": 75656.4140625,400      "learning_rate": 1.4311315924219151e-06,401      "loss": 0.7062,402      "step": 560403    },404    {405      "epoch": 0.03648,406      "grad_norm": 79985.875,407      "learning_rate": 1.4567332309267796e-06,408      "loss": 0.7155,409      "step": 570410    },411    {412      "epoch": 0.03712,413      "grad_norm": 76334.078125,414      "learning_rate": 1.4823348694316437e-06,415      "loss": 0.7075,416      "step": 580417    },418    {419      "epoch": 0.03776,420      "grad_norm": 140764.03125,421      "learning_rate": 1.507936507936508e-06,422      "loss": 0.7065,423      "step": 590424    },425    {426      "epoch": 0.0384,427      "grad_norm": 100877.296875,428      "learning_rate": 1.5335381464413722e-06,429      "loss": 0.7096,430      "step": 600431    },432    {433      "epoch": 0.03904,434      "grad_norm": 104088.1171875,435      "learning_rate": 1.5591397849462367e-06,436      "loss": 0.6987,437      "step": 610438    },439    {440      "epoch": 0.03968,441      "grad_norm": 80806.2265625,442      "learning_rate": 1.584741423451101e-06,443      "loss": 0.707,444      "step": 620445    },446    {447      "epoch": 0.04032,448      "grad_norm": 109884.765625,449      "learning_rate": 1.6103430619559655e-06,450      "loss": 0.6991,451      "step": 630452    },453    {454      "epoch": 0.04096,455      "grad_norm": 79944.890625,456      "learning_rate": 1.6359447004608298e-06,457      "loss": 0.7047,458      "step": 640459    },460    {461      "epoch": 0.0416,462      "grad_norm": 93673.3828125,463      "learning_rate": 1.6615463389656938e-06,464      "loss": 0.6971,465      "step": 650466    },467    {468      "epoch": 0.04224,469      "grad_norm": 76641.265625,470      "learning_rate": 1.6871479774705581e-06,471      "loss": 0.6957,472      "step": 660473    },474    {475      "epoch": 0.04288,476      "grad_norm": 73583.5546875,477      "learning_rate": 1.7127496159754226e-06,478      "loss": 0.7028,479      "step": 670480    },481    {482      "epoch": 0.04352,483      "grad_norm": 75177.9609375,484      "learning_rate": 1.7383512544802869e-06,485      "loss": 0.7012,486      "step": 680487    },488    {489      "epoch": 0.04416,490      "grad_norm": 78340.8515625,491      "learning_rate": 1.7639528929851512e-06,492      "loss": 0.6987,493      "step": 690494    },495    {496      "epoch": 0.0448,497      "grad_norm": 86004.1171875,498      "learning_rate": 1.7895545314900157e-06,499      "loss": 0.7061,500      "step": 700501    },502    {503      "epoch": 0.04544,504      "grad_norm": 94212.0390625,505      "learning_rate": 1.8151561699948797e-06,506      "loss": 0.6993,507      "step": 710508    },509    {510      "epoch": 0.04608,511      "grad_norm": 83918.2421875,512      "learning_rate": 1.840757808499744e-06,513      "loss": 0.7009,514      "step": 720515    },516    {517      "epoch": 0.04672,518      "grad_norm": 68374.3125,519      "learning_rate": 1.8663594470046085e-06,520      "loss": 0.6964,521      "step": 730522    },523    {524      "epoch": 0.04736,525      "grad_norm": 90348.78125,526      "learning_rate": 1.8919610855094728e-06,527      "loss": 0.7011,528      "step": 740529    },530    {531      "epoch": 0.048,532      "grad_norm": 146658.0,533      "learning_rate": 1.9175627240143373e-06,534      "loss": 0.7003,535      "step": 750536    },537    {538      "epoch": 0.04864,539      "grad_norm": 112037.1640625,540      "learning_rate": 1.9431643625192015e-06,541      "loss": 0.7051,542      "step": 760543    },544    {545      "epoch": 0.04928,546      "grad_norm": 70628.625,547      "learning_rate": 1.9687660010240654e-06,548      "loss": 0.6923,549      "step": 770550    },551    {552      "epoch": 0.04992,553      "grad_norm": 109922.125,554      "learning_rate": 1.99436763952893e-06,555      "loss": 0.6893,556      "step": 780557    },558    {559      "epoch": 0.05056,560      "grad_norm": 135306.375,561      "learning_rate": 2.0199692780337944e-06,562      "loss": 0.7008,563      "step": 790564    },565    {566      "epoch": 0.0512,567      "grad_norm": 82354.8046875,568      "learning_rate": 2.0455709165386586e-06,569      "loss": 0.705,570      "step": 800571    },572    {573      "epoch": 0.05184,574      "grad_norm": 95951.671875,575      "learning_rate": 2.071172555043523e-06,576      "loss": 0.6912,577      "step": 810578    },579    {580      "epoch": 0.05248,581      "grad_norm": 96797.4609375,582      "learning_rate": 2.096774193548387e-06,583      "loss": 0.6922,584      "step": 820585    },586    {587      "epoch": 0.05312,588      "grad_norm": 87190.625,589      "learning_rate": 2.122375832053252e-06,590      "loss": 0.6946,591      "step": 830592    },593    {594      "epoch": 0.05376,595      "grad_norm": 87958.5625,596      "learning_rate": 2.1479774705581158e-06,597      "loss": 0.6949,598      "step": 840599    },600    {601      "epoch": 0.0544,602      "grad_norm": 77217.1796875,603      "learning_rate": 2.17357910906298e-06,604      "loss": 0.6928,605      "step": 850606    },607    {608      "epoch": 0.05504,609      "grad_norm": 117156.5546875,610      "learning_rate": 2.1991807475678443e-06,611      "loss": 0.692,612      "step": 860613    },614    {615      "epoch": 0.05568,616      "grad_norm": 94618.6875,617      "learning_rate": 2.224782386072709e-06,618      "loss": 0.6976,619      "step": 870620    },621    {622      "epoch": 0.05632,623      "grad_norm": 71444.6484375,624      "learning_rate": 2.2503840245775733e-06,625      "loss": 0.6989,626      "step": 880627    },628    {629      "epoch": 0.05696,630      "grad_norm": 159991.609375,631      "learning_rate": 2.2759856630824376e-06,632      "loss": 0.6928,633      "step": 890634    },635    {636      "epoch": 0.0576,637      "grad_norm": 81899.6875,638      "learning_rate": 2.301587301587302e-06,639      "loss": 0.691,640      "step": 900641    },642    {643      "epoch": 0.05824,644      "grad_norm": 110817.3671875,645      "learning_rate": 2.327188940092166e-06,646      "loss": 0.6858,647      "step": 910648    },649    {650      "epoch": 0.05888,651      "grad_norm": 105698.109375,652      "learning_rate": 2.3527905785970304e-06,653      "loss": 0.6965,654      "step": 920655    },656    {657      "epoch": 0.05952,658      "grad_norm": 76475.0,659      "learning_rate": 2.3783922171018947e-06,660      "loss": 0.6901,661      "step": 930662    },663    {664      "epoch": 0.06016,665      "grad_norm": 96672.6796875,666      "learning_rate": 2.403993855606759e-06,667      "loss": 0.6908,668      "step": 940669    },670    {671      "epoch": 0.0608,672      "grad_norm": 114510.8125,673      "learning_rate": 2.4295954941116232e-06,674      "loss": 0.6904,675      "step": 950676    },677    {678      "epoch": 0.06144,679      "grad_norm": 62412.4375,680      "learning_rate": 2.455197132616488e-06,681      "loss": 0.6855,682      "step": 960683    },684    {685      "epoch": 0.06208,686      "grad_norm": 92860.7109375,687      "learning_rate": 2.4807987711213518e-06,688      "loss": 0.6752,689      "step": 970690    },691    {692      "epoch": 0.06272,693      "grad_norm": 75184.359375,694      "learning_rate": 2.506400409626216e-06,695      "loss": 0.6868,696      "step": 980697    },698    {699      "epoch": 0.06336,700      "grad_norm": 77771.1640625,701      "learning_rate": 2.5320020481310808e-06,702      "loss": 0.6941,703      "step": 990704    },705    {706      "epoch": 0.064,707      "grad_norm": 65366.796875,708      "learning_rate": 2.557603686635945e-06,709      "loss": 0.6808,710      "step": 1000711    },712    {713      "epoch": 0.064,714      "eval_accuracy": 0.75744,715      "eval_loss": 0.6539617776870728,716      "eval_macro_f1": 0.7549859375827388,717      "eval_runtime": 1576.6702,718      "eval_samples_per_second": 63.425,719      "eval_steps_per_second": 0.496,720      "step": 1000721    },722    {723      "epoch": 0.06464,724      "grad_norm": 73310.6171875,725      "learning_rate": 2.583205325140809e-06,726      "loss": 0.6866,727      "step": 1010728    },729    {730      "epoch": 0.06528,731      "grad_norm": 80602.859375,732      "learning_rate": 2.6088069636456736e-06,733      "loss": 0.6873,734      "step": 1020735    },736    {737      "epoch": 0.06592,738      "grad_norm": 121537.0234375,739      "learning_rate": 2.634408602150538e-06,740      "loss": 0.6806,741      "step": 1030742    },743    {744      "epoch": 0.06656,745      "grad_norm": 105537.46875,746      "learning_rate": 2.6600102406554026e-06,747      "loss": 0.6835,748      "step": 1040749    },750    {751      "epoch": 0.0672,752      "grad_norm": 188847.71875,753      "learning_rate": 2.6856118791602664e-06,754      "loss": 0.687,755      "step": 1050756    },757    {758      "epoch": 0.06784,759      "grad_norm": 73677.8359375,760      "learning_rate": 2.7112135176651307e-06,761      "loss": 0.6848,762      "step": 1060763    },764    {765      "epoch": 0.06848,766      "grad_norm": 72158.984375,767      "learning_rate": 2.736815156169995e-06,768      "loss": 0.6833,769      "step": 1070770    },771    {772      "epoch": 0.06912,773      "grad_norm": 73585.1015625,774      "learning_rate": 2.7624167946748593e-06,775      "loss": 0.6766,776      "step": 1080777    },778    {779      "epoch": 0.06976,780      "grad_norm": 78721.0390625,781      "learning_rate": 2.788018433179724e-06,782      "loss": 0.6796,783      "step": 1090784    },785    {786      "epoch": 0.0704,787      "grad_norm": 72044.3515625,788      "learning_rate": 2.813620071684588e-06,789      "loss": 0.673,790      "step": 1100791    },792    {793      "epoch": 0.07104,794      "grad_norm": 100059.984375,795      "learning_rate": 2.8392217101894525e-06,796      "loss": 0.6783,797      "step": 1110798    },799    {800      "epoch": 0.07168,801      "grad_norm": 92175.921875,802      "learning_rate": 2.864823348694317e-06,803      "loss": 0.6901,804      "step": 1120805    },806    {807      "epoch": 0.07232,808      "grad_norm": 86143.453125,809      "learning_rate": 2.8904249871991806e-06,810      "loss": 0.6769,811      "step": 1130812    },813    {814      "epoch": 0.07296,815      "grad_norm": 101410.171875,816      "learning_rate": 2.9160266257040453e-06,817      "loss": 0.6781,818      "step": 1140819    },820    {821      "epoch": 0.0736,822      "grad_norm": 67173.296875,823      "learning_rate": 2.9416282642089096e-06,824      "loss": 0.6737,825      "step": 1150826    },827    {828      "epoch": 0.07424,829      "grad_norm": 100701.8203125,830      "learning_rate": 2.967229902713774e-06,831      "loss": 0.6741,832      "step": 1160833    },834    {835      "epoch": 0.07488,836      "grad_norm": 75457.328125,837      "learning_rate": 2.992831541218638e-06,838      "loss": 0.6742,839      "step": 1170840    },841    {842      "epoch": 0.07552,843      "grad_norm": 97755.9921875,844      "learning_rate": 3.018433179723503e-06,845      "loss": 0.6845,846      "step": 1180847    },848    {849      "epoch": 0.07616,850      "grad_norm": 53426.1171875,851      "learning_rate": 3.0440348182283667e-06,852      "loss": 0.6718,853      "step": 1190854    },855    {856      "epoch": 0.0768,857      "grad_norm": 71654.625,858      "learning_rate": 3.069636456733231e-06,859      "loss": 0.6798,860      "step": 1200861    }862  ],863  "logging_steps": 10,864  "max_steps": 78125,865  "num_input_tokens_seen": 0,866  "num_train_epochs": 5,867  "save_steps": 200,868  "stateful_callbacks": {869    "EarlyStoppingCallback": {870      "args": {871        "early_stopping_patience": 3,872        "early_stopping_threshold": 0.0873      },874      "attributes": {875        "early_stopping_patience_counter": 0876      }877    },878    "TrainerControl": {879      "args": {880        "should_epoch_stop": false,881        "should_evaluate": false,882        "should_log": false,883        "should_save": true,884        "should_training_stop": false885      },886      "attributes": {}887    }888  },889  "total_flos": 1.010052097089216e+16,890  "train_batch_size": 32,891  "trial_name": null,892  "trial_params": null893}894