diff options
Diffstat (limited to 'results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json')
| -rw-r--r-- | results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json | 2965 |
1 files changed, 2965 insertions, 0 deletions
diff --git a/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json b/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json new file mode 100644 index 0000000..1d1bf92 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json @@ -0,0 +1,2965 @@ +{ + "architecture": { + "base_width": 16, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 269722, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 200, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.05, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "method": "dualprop", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json", + "output_lr": 0.05, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.6441188268873426, + "train_seconds": 245.87645721435547, + "validation": { + "accuracy": 0.4738, + "examples": 5000, + "loss": 1.4967913818359375 + }, + "validation_seconds": 0.12893104553222656 + }, + { + "epoch": 2, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.1635693959765965, + "train_seconds": 245.52287817001343, + "validation": { + "accuracy": 0.6026, + "examples": 5000, + "loss": 1.153321435546875 + }, + "validation_seconds": 0.10486650466918945 + }, + { + "epoch": 3, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.9352410600874159, + "train_seconds": 245.50975513458252, + "validation": { + "accuracy": 0.6018, + "examples": 5000, + "loss": 1.329039013671875 + }, + "validation_seconds": 0.10483121871948242 + }, + { + "epoch": 4, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.7935926506890191, + "train_seconds": 245.52499318122864, + "validation": { + "accuracy": 0.7, + "examples": 5000, + "loss": 0.8777848876953125 + }, + "validation_seconds": 0.10486555099487305 + }, + { + "epoch": 5, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.7077589258935716, + "train_seconds": 245.52429103851318, + "validation": { + "accuracy": 0.7442, + "examples": 5000, + "loss": 0.7581129028320313 + }, + "validation_seconds": 0.1048576831817627 + }, + { + "epoch": 6, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6486876674969991, + "train_seconds": 245.51856327056885, + "validation": { + "accuracy": 0.7254, + "examples": 5000, + "loss": 0.8741846923828125 + }, + "validation_seconds": 0.10486769676208496 + }, + { + "epoch": 7, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6074715187284682, + "train_seconds": 245.71568083763123, + "validation": { + "accuracy": 0.7682, + "examples": 5000, + "loss": 0.6674996459960938 + }, + "validation_seconds": 0.10516047477722168 + }, + { + "epoch": 8, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5710463250690037, + "train_seconds": 245.7029790878296, + "validation": { + "accuracy": 0.7782, + "examples": 5000, + "loss": 0.6623569702148437 + }, + "validation_seconds": 0.10492897033691406 + }, + { + "epoch": 9, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5336540970166525, + "train_seconds": 245.72987985610962, + "validation": { + "accuracy": 0.786, + "examples": 5000, + "loss": 0.640069384765625 + }, + "validation_seconds": 0.10491442680358887 + }, + { + "epoch": 10, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5106024785730574, + "train_seconds": 245.71102261543274, + "validation": { + "accuracy": 0.7826, + "examples": 5000, + "loss": 0.6581781616210938 + }, + "validation_seconds": 0.104949951171875 + }, + { + "epoch": 11, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.488464806482527, + "train_seconds": 245.70360898971558, + "validation": { + "accuracy": 0.7776, + "examples": 5000, + "loss": 0.6763726928710937 + }, + "validation_seconds": 0.10500192642211914 + }, + { + "epoch": 12, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4659737181398604, + "train_seconds": 245.7096302509308, + "validation": { + "accuracy": 0.8018, + "examples": 5000, + "loss": 0.603849560546875 + }, + "validation_seconds": 0.10489654541015625 + }, + { + "epoch": 13, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4473519527753194, + "train_seconds": 245.70638012886047, + "validation": { + "accuracy": 0.7406, + "examples": 5000, + "loss": 0.8407753784179688 + }, + "validation_seconds": 0.10495471954345703 + }, + { + "epoch": 14, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4388091464042664, + "train_seconds": 245.70655918121338, + "validation": { + "accuracy": 0.8338, + "examples": 5000, + "loss": 0.48709592895507814 + }, + "validation_seconds": 0.10503792762756348 + }, + { + "epoch": 15, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.42087091799841986, + "train_seconds": 245.71817874908447, + "validation": { + "accuracy": 0.7902, + "examples": 5000, + "loss": 0.6748830810546875 + }, + "validation_seconds": 0.1049962043762207 + }, + { + "epoch": 16, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4085377493328518, + "train_seconds": 245.71288919448853, + "validation": { + "accuracy": 0.8178, + "examples": 5000, + "loss": 0.5664857482910156 + }, + "validation_seconds": 0.10489010810852051 + }, + { + "epoch": 17, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.39353587019178604, + "train_seconds": 245.71254181861877, + "validation": { + "accuracy": 0.8128, + "examples": 5000, + "loss": 0.5877752197265625 + }, + "validation_seconds": 0.1049199104309082 + }, + { + "epoch": 18, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.38593393679724797, + "train_seconds": 245.7317178249359, + "validation": { + "accuracy": 0.8052, + "examples": 5000, + "loss": 0.5994394897460937 + }, + "validation_seconds": 0.1049954891204834 + }, + { + "epoch": 19, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.37112634809547, + "train_seconds": 245.7781913280487, + "validation": { + "accuracy": 0.8426, + "examples": 5000, + "loss": 0.4838470458984375 + }, + "validation_seconds": 0.10508608818054199 + }, + { + "epoch": 20, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3595076397365994, + "train_seconds": 246.36518168449402, + "validation": { + "accuracy": 0.8094, + "examples": 5000, + "loss": 0.6470749267578125 + }, + "validation_seconds": 0.10510373115539551 + }, + { + "epoch": 21, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3523521614180671, + "train_seconds": 246.52931666374207, + "validation": { + "accuracy": 0.8242, + "examples": 5000, + "loss": 0.5845088745117187 + }, + "validation_seconds": 0.10520172119140625 + }, + { + "epoch": 22, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.34563857612080046, + "train_seconds": 246.01910829544067, + "validation": { + "accuracy": 0.8096, + "examples": 5000, + "loss": 0.62537509765625 + }, + "validation_seconds": 0.10520815849304199 + }, + { + "epoch": 23, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.33976192728678384, + "train_seconds": 245.8921320438385, + "validation": { + "accuracy": 0.8444, + "examples": 5000, + "loss": 0.484679150390625 + }, + "validation_seconds": 0.1049337387084961 + }, + { + "epoch": 24, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3225088294665019, + "train_seconds": 245.83658075332642, + "validation": { + "accuracy": 0.8396, + "examples": 5000, + "loss": 0.50716240234375 + }, + "validation_seconds": 0.1049196720123291 + }, + { + "epoch": 25, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.32271037464141844, + "train_seconds": 245.84278297424316, + "validation": { + "accuracy": 0.8406, + "examples": 5000, + "loss": 0.493395458984375 + }, + "validation_seconds": 0.10499238967895508 + }, + { + "epoch": 26, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3157399925496843, + "train_seconds": 245.84543251991272, + "validation": { + "accuracy": 0.858, + "examples": 5000, + "loss": 0.44974503173828123 + }, + "validation_seconds": 0.10491776466369629 + }, + { + "epoch": 27, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.30543111317952476, + "train_seconds": 245.86448335647583, + "validation": { + "accuracy": 0.8416, + "examples": 5000, + "loss": 0.5209959777832032 + }, + "validation_seconds": 0.10499191284179688 + }, + { + "epoch": 28, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.301433762105306, + "train_seconds": 245.840149641037, + "validation": { + "accuracy": 0.8244, + "examples": 5000, + "loss": 0.5745289489746094 + }, + "validation_seconds": 0.10509037971496582 + }, + { + "epoch": 29, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.29656535551018187, + "train_seconds": 245.8558783531189, + "validation": { + "accuracy": 0.8144, + "examples": 5000, + "loss": 0.6412824096679688 + }, + "validation_seconds": 0.10495448112487793 + }, + { + "epoch": 30, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.29441430922614203, + "train_seconds": 245.80975580215454, + "validation": { + "accuracy": 0.8356, + "examples": 5000, + "loss": 0.5149374877929688 + }, + "validation_seconds": 0.10511445999145508 + }, + { + "epoch": 31, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.28566161390410527, + "train_seconds": 245.73937225341797, + "validation": { + "accuracy": 0.842, + "examples": 5000, + "loss": 0.5199394226074219 + }, + "validation_seconds": 0.10500478744506836 + }, + { + "epoch": 32, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2805537829717, + "train_seconds": 245.7104251384735, + "validation": { + "accuracy": 0.8688, + "examples": 5000, + "loss": 0.4206529296875 + }, + "validation_seconds": 0.10506677627563477 + }, + { + "epoch": 33, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2817669762743844, + "train_seconds": 245.7130422592163, + "validation": { + "accuracy": 0.8566, + "examples": 5000, + "loss": 0.43376856689453125 + }, + "validation_seconds": 0.1049799919128418 + }, + { + "epoch": 34, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.27807320489618514, + "train_seconds": 245.71902012825012, + "validation": { + "accuracy": 0.8558, + "examples": 5000, + "loss": 0.46831655883789064 + }, + "validation_seconds": 0.10493636131286621 + }, + { + "epoch": 35, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.263952144712872, + "train_seconds": 245.93091106414795, + "validation": { + "accuracy": 0.8554, + "examples": 5000, + "loss": 0.4892894958496094 + }, + "validation_seconds": 0.10500526428222656 + }, + { + "epoch": 36, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2642927113586002, + "train_seconds": 246.30246591567993, + "validation": { + "accuracy": 0.8482, + "examples": 5000, + "loss": 0.47789179077148436 + }, + "validation_seconds": 0.10546326637268066 + }, + { + "epoch": 37, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2638369460807906, + "train_seconds": 246.64353728294373, + "validation": { + "accuracy": 0.8326, + "examples": 5000, + "loss": 0.5493978393554687 + }, + "validation_seconds": 0.10544991493225098 + }, + { + "epoch": 38, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.25538891750971476, + "train_seconds": 246.44886422157288, + "validation": { + "accuracy": 0.8574, + "examples": 5000, + "loss": 0.451583837890625 + }, + "validation_seconds": 0.10479283332824707 + }, + { + "epoch": 39, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.25059934278064305, + "train_seconds": 246.8432900905609, + "validation": { + "accuracy": 0.8636, + "examples": 5000, + "loss": 0.44829965209960937 + }, + "validation_seconds": 0.10497689247131348 + }, + { + "epoch": 40, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.24968606785933176, + "train_seconds": 246.86512732505798, + "validation": { + "accuracy": 0.857, + "examples": 5000, + "loss": 0.47868128662109377 + }, + "validation_seconds": 0.10499143600463867 + }, + { + "epoch": 41, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.24340666038990022, + "train_seconds": 246.9741039276123, + "validation": { + "accuracy": 0.863, + "examples": 5000, + "loss": 0.43970326538085935 + }, + "validation_seconds": 0.10489535331726074 + }, + { + "epoch": 42, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2426151712603039, + "train_seconds": 246.62085914611816, + "validation": { + "accuracy": 0.8678, + "examples": 5000, + "loss": 0.43893058471679686 + }, + "validation_seconds": 0.10530686378479004 + }, + { + "epoch": 43, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.23919208281305102, + "train_seconds": 246.47220969200134, + "validation": { + "accuracy": 0.8564, + "examples": 5000, + "loss": 0.4502086669921875 + }, + "validation_seconds": 0.1048734188079834 + }, + { + "epoch": 44, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2358168052620358, + "train_seconds": 246.44223856925964, + "validation": { + "accuracy": 0.8584, + "examples": 5000, + "loss": 0.4598034118652344 + }, + "validation_seconds": 0.10493111610412598 + }, + { + "epoch": 45, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.23206949197981092, + "train_seconds": 245.91084361076355, + "validation": { + "accuracy": 0.8434, + "examples": 5000, + "loss": 0.5504330444335938 + }, + "validation_seconds": 0.10497212409973145 + }, + { + "epoch": 46, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.23059860911899144, + "train_seconds": 246.39553570747375, + "validation": { + "accuracy": 0.8514, + "examples": 5000, + "loss": 0.4961731567382813 + }, + "validation_seconds": 0.10496854782104492 + }, + { + "epoch": 47, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2259720978895823, + "train_seconds": 246.28038716316223, + "validation": { + "accuracy": 0.8612, + "examples": 5000, + "loss": 0.4870090087890625 + }, + "validation_seconds": 0.10482335090637207 + }, + { + "epoch": 48, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2301974679734972, + "train_seconds": 245.90750241279602, + "validation": { + "accuracy": 0.8584, + "examples": 5000, + "loss": 0.4882284912109375 + }, + "validation_seconds": 0.10474634170532227 + }, + { + "epoch": 49, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.22358488881323071, + "train_seconds": 245.81437158584595, + "validation": { + "accuracy": 0.8672, + "examples": 5000, + "loss": 0.46109835205078126 + }, + "validation_seconds": 0.10482645034790039 + }, + { + "epoch": 50, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2217408281246821, + "train_seconds": 245.80533456802368, + "validation": { + "accuracy": 0.8702, + "examples": 5000, + "loss": 0.42463759765625 + }, + "validation_seconds": 0.10474348068237305 + }, + { + "epoch": 51, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.21557051228152382, + "train_seconds": 245.94883275032043, + "validation": { + "accuracy": 0.8518, + "examples": 5000, + "loss": 0.48058946533203123 + }, + "validation_seconds": 0.10477089881896973 + }, + { + "epoch": 52, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2120681472301483, + "train_seconds": 245.94964838027954, + "validation": { + "accuracy": 0.865, + "examples": 5000, + "loss": 0.44984417114257813 + }, + "validation_seconds": 0.10477733612060547 + }, + { + "epoch": 53, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.20977391286161212, + "train_seconds": 245.92657470703125, + "validation": { + "accuracy": 0.8518, + "examples": 5000, + "loss": 0.53985625 + }, + "validation_seconds": 0.10475325584411621 + }, + { + "epoch": 54, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2089330201546351, + "train_seconds": 245.93967461585999, + "validation": { + "accuracy": 0.856, + "examples": 5000, + "loss": 0.4909478576660156 + }, + "validation_seconds": 0.10521721839904785 + }, + { + "epoch": 55, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.21272749128341675, + "train_seconds": 246.59948468208313, + "validation": { + "accuracy": 0.8546, + "examples": 5000, + "loss": 0.4906638977050781 + }, + "validation_seconds": 0.10508394241333008 + }, + { + "epoch": 56, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.20717158926592932, + "train_seconds": 246.48116064071655, + "validation": { + "accuracy": 0.8504, + "examples": 5000, + "loss": 0.5430794372558594 + }, + "validation_seconds": 0.10493278503417969 + }, + { + "epoch": 57, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.20313986417717403, + "train_seconds": 246.57427525520325, + "validation": { + "accuracy": 0.842, + "examples": 5000, + "loss": 0.5557522338867188 + }, + "validation_seconds": 0.1049644947052002 + }, + { + "epoch": 58, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2027928885936737, + "train_seconds": 246.51094436645508, + "validation": { + "accuracy": 0.8666, + "examples": 5000, + "loss": 0.4270877563476563 + }, + "validation_seconds": 0.1048426628112793 + }, + { + "epoch": 59, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.20086540544033052, + "train_seconds": 245.954576253891, + "validation": { + "accuracy": 0.8526, + "examples": 5000, + "loss": 0.4917937927246094 + }, + "validation_seconds": 0.10494685173034668 + }, + { + "epoch": 60, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.19744458295504252, + "train_seconds": 246.04650807380676, + "validation": { + "accuracy": 0.8634, + "examples": 5000, + "loss": 0.4751939880371094 + }, + "validation_seconds": 0.10476160049438477 + }, + { + "epoch": 61, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.19625209110842812, + "train_seconds": 246.03907656669617, + "validation": { + "accuracy": 0.855, + "examples": 5000, + "loss": 0.5089188537597656 + }, + "validation_seconds": 0.10500311851501465 + }, + { + "epoch": 62, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.19375401494238112, + "train_seconds": 245.91973066329956, + "validation": { + "accuracy": 0.8516, + "examples": 5000, + "loss": 0.5393552185058593 + }, + "validation_seconds": 0.10500383377075195 + }, + { + "epoch": 63, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1935203471104304, + "train_seconds": 245.89782786369324, + "validation": { + "accuracy": 0.861, + "examples": 5000, + "loss": 0.4727430725097656 + }, + "validation_seconds": 0.1048274040222168 + }, + { + "epoch": 64, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18534126149018607, + "train_seconds": 245.9205825328827, + "validation": { + "accuracy": 0.8472, + "examples": 5000, + "loss": 0.5769324584960938 + }, + "validation_seconds": 0.10484862327575684 + }, + { + "epoch": 65, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.19316486716535355, + "train_seconds": 245.91525053977966, + "validation": { + "accuracy": 0.8758, + "examples": 5000, + "loss": 0.4404202941894531 + }, + "validation_seconds": 0.10477232933044434 + }, + { + "epoch": 66, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18819561854203543, + "train_seconds": 245.94995212554932, + "validation": { + "accuracy": 0.8632, + "examples": 5000, + "loss": 0.4956598876953125 + }, + "validation_seconds": 0.10492324829101562 + }, + { + "epoch": 67, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1861683264070087, + "train_seconds": 245.92055249214172, + "validation": { + "accuracy": 0.8396, + "examples": 5000, + "loss": 0.564496875 + }, + "validation_seconds": 0.10488319396972656 + }, + { + "epoch": 68, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18245379467010497, + "train_seconds": 245.90807962417603, + "validation": { + "accuracy": 0.855, + "examples": 5000, + "loss": 0.516592431640625 + }, + "validation_seconds": 0.10478520393371582 + }, + { + "epoch": 69, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18739715305169424, + "train_seconds": 245.9303948879242, + "validation": { + "accuracy": 0.8826, + "examples": 5000, + "loss": 0.4084698120117187 + }, + "validation_seconds": 0.10473227500915527 + }, + { + "epoch": 70, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18380963235431247, + "train_seconds": 245.92044949531555, + "validation": { + "accuracy": 0.8758, + "examples": 5000, + "loss": 0.43710973510742185 + }, + "validation_seconds": 0.1049046516418457 + }, + { + "epoch": 71, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18311760312716166, + "train_seconds": 245.96293902397156, + "validation": { + "accuracy": 0.8716, + "examples": 5000, + "loss": 0.4604240112304688 + }, + "validation_seconds": 0.1050562858581543 + }, + { + "epoch": 72, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17762095175319248, + "train_seconds": 245.93334341049194, + "validation": { + "accuracy": 0.8606, + "examples": 5000, + "loss": 0.5003765075683594 + }, + "validation_seconds": 0.10496139526367188 + }, + { + "epoch": 73, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1763817933294508, + "train_seconds": 245.9250099658966, + "validation": { + "accuracy": 0.882, + "examples": 5000, + "loss": 0.41763900756835937 + }, + "validation_seconds": 0.10477662086486816 + }, + { + "epoch": 74, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17314121738274893, + "train_seconds": 245.8988173007965, + "validation": { + "accuracy": 0.8886, + "examples": 5000, + "loss": 0.37928946533203123 + }, + "validation_seconds": 0.10474419593811035 + }, + { + "epoch": 75, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17797526009877523, + "train_seconds": 246.13520216941833, + "validation": { + "accuracy": 0.8752, + "examples": 5000, + "loss": 0.4501698974609375 + }, + "validation_seconds": 0.10515260696411133 + }, + { + "epoch": 76, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17506222959889306, + "train_seconds": 246.67396068572998, + "validation": { + "accuracy": 0.8754, + "examples": 5000, + "loss": 0.4668177490234375 + }, + "validation_seconds": 0.10496044158935547 + }, + { + "epoch": 77, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1717323899322086, + "train_seconds": 246.7168323993683, + "validation": { + "accuracy": 0.8626, + "examples": 5000, + "loss": 0.5042861206054687 + }, + "validation_seconds": 0.10485363006591797 + }, + { + "epoch": 78, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17638354170189963, + "train_seconds": 246.71074175834656, + "validation": { + "accuracy": 0.8708, + "examples": 5000, + "loss": 0.478639453125 + }, + "validation_seconds": 0.10538220405578613 + }, + { + "epoch": 79, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1677387358956867, + "train_seconds": 246.7130355834961, + "validation": { + "accuracy": 0.8472, + "examples": 5000, + "loss": 0.5276833251953125 + }, + "validation_seconds": 0.10472655296325684 + }, + { + "epoch": 80, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16351781239774493, + "train_seconds": 246.68999099731445, + "validation": { + "accuracy": 0.8682, + "examples": 5000, + "loss": 0.48907069091796873 + }, + "validation_seconds": 0.10485410690307617 + }, + { + "epoch": 81, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17571784110599095, + "train_seconds": 246.70585751533508, + "validation": { + "accuracy": 0.868, + "examples": 5000, + "loss": 0.4847575805664063 + }, + "validation_seconds": 0.10532069206237793 + }, + { + "epoch": 82, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1694504910853174, + "train_seconds": 246.70015454292297, + "validation": { + "accuracy": 0.8556, + "examples": 5000, + "loss": 0.488367724609375 + }, + "validation_seconds": 0.10495138168334961 + }, + { + "epoch": 83, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16532643840048047, + "train_seconds": 246.68881845474243, + "validation": { + "accuracy": 0.8416, + "examples": 5000, + "loss": 0.5911368530273438 + }, + "validation_seconds": 0.10479950904846191 + }, + { + "epoch": 84, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1684850957579083, + "train_seconds": 246.71207571029663, + "validation": { + "accuracy": 0.8732, + "examples": 5000, + "loss": 0.46909482421875 + }, + "validation_seconds": 0.1052088737487793 + }, + { + "epoch": 85, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1677253027809991, + "train_seconds": 246.69193530082703, + "validation": { + "accuracy": 0.8696, + "examples": 5000, + "loss": 0.4961885070800781 + }, + "validation_seconds": 0.10480213165283203 + }, + { + "epoch": 86, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16023339446915522, + "train_seconds": 246.63249588012695, + "validation": { + "accuracy": 0.8818, + "examples": 5000, + "loss": 0.4365656616210937 + }, + "validation_seconds": 0.1050879955291748 + }, + { + "epoch": 87, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1696844631936815, + "train_seconds": 246.65419721603394, + "validation": { + "accuracy": 0.8864, + "examples": 5000, + "loss": 0.394526025390625 + }, + "validation_seconds": 0.10512447357177734 + }, + { + "epoch": 88, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16108827577696908, + "train_seconds": 246.70431089401245, + "validation": { + "accuracy": 0.8798, + "examples": 5000, + "loss": 0.45328688354492186 + }, + "validation_seconds": 0.10512113571166992 + }, + { + "epoch": 89, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16032158904737898, + "train_seconds": 246.6812767982483, + "validation": { + "accuracy": 0.8816, + "examples": 5000, + "loss": 0.44019458618164065 + }, + "validation_seconds": 0.10508894920349121 + }, + { + "epoch": 90, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16225236432817247, + "train_seconds": 246.68167066574097, + "validation": { + "accuracy": 0.8624, + "examples": 5000, + "loss": 0.4962203369140625 + }, + "validation_seconds": 0.10485458374023438 + }, + { + "epoch": 91, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16403424759838317, + "train_seconds": 246.61265230178833, + "validation": { + "accuracy": 0.8774, + "examples": 5000, + "loss": 0.43822531127929687 + }, + "validation_seconds": 0.10491776466369629 + }, + { + "epoch": 92, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1596482249458631, + "train_seconds": 246.5586667060852, + "validation": { + "accuracy": 0.8592, + "examples": 5000, + "loss": 0.5626332885742188 + }, + "validation_seconds": 0.10481786727905273 + }, + { + "epoch": 93, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.15875176428423987, + "train_seconds": 246.6671712398529, + "validation": { + "accuracy": 0.8762, + "examples": 5000, + "loss": 0.45965128173828124 + }, + "validation_seconds": 0.10487985610961914 + }, + { + "epoch": 94, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16234073853757647, + "train_seconds": 246.63065123558044, + "validation": { + "accuracy": 0.868, + "examples": 5000, + "loss": 0.46680614624023437 + }, + "validation_seconds": 0.10476207733154297 + }, + { + "epoch": 95, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1534736484448115, + "train_seconds": 246.65962553024292, + "validation": { + "accuracy": 0.8672, + "examples": 5000, + "loss": 0.46624967041015625 + }, + "validation_seconds": 0.10492587089538574 + }, + { + "epoch": 96, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.15440344122515784, + "train_seconds": 246.66373991966248, + "validation": { + "accuracy": 0.8628, + "examples": 5000, + "loss": 0.4870942626953125 + }, + "validation_seconds": 0.10490155220031738 + }, + { + "epoch": 97, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.15565202848911286, + "train_seconds": 246.66854810714722, + "validation": { + "accuracy": 0.8712, + "examples": 5000, + "loss": 0.4484557861328125 + }, + "validation_seconds": 0.10535168647766113 + }, + { + "epoch": 98, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.15759743853674996, + "train_seconds": 246.67762279510498, + "validation": { + "accuracy": 0.8688, + "examples": 5000, + "loss": 0.49867626953125 + }, + "validation_seconds": 0.10507059097290039 + }, + { + "epoch": 99, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1574016169892417, + "train_seconds": 246.6524622440338, + "validation": { + "accuracy": 0.8764, + "examples": 5000, + "loss": 0.4572004699707031 + }, + "validation_seconds": 0.1049041748046875 + }, + { + "epoch": 100, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.14737063789367677, + "train_seconds": 246.63342666625977, + "validation": { + "accuracy": 0.8754, + "examples": 5000, + "loss": 0.4618093322753906 + }, + "validation_seconds": 0.10494613647460938 + }, + { + "epoch": 101, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.09404547926055061, + "train_seconds": 246.65588760375977, + "validation": { + "accuracy": 0.906, + "examples": 5000, + "loss": 0.3335665466308594 + }, + "validation_seconds": 0.10500240325927734 + }, + { + "epoch": 102, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07055688915782504, + "train_seconds": 246.65637946128845, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.3303562255859375 + }, + "validation_seconds": 0.10512924194335938 + }, + { + "epoch": 103, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.06393063375221358, + "train_seconds": 246.56659412384033, + "validation": { + "accuracy": 0.9106, + "examples": 5000, + "loss": 0.33267850341796873 + }, + "validation_seconds": 0.10535550117492676 + }, + { + "epoch": 104, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.056251141385237374, + "train_seconds": 246.58122849464417, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.34011267700195313 + }, + "validation_seconds": 0.10490632057189941 + }, + { + "epoch": 105, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.056818765364752874, + "train_seconds": 246.65290474891663, + "validation": { + "accuracy": 0.906, + "examples": 5000, + "loss": 0.3483814758300781 + }, + "validation_seconds": 0.10506415367126465 + }, + { + "epoch": 106, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.052452130284574294, + "train_seconds": 246.61904573440552, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.345201025390625 + }, + "validation_seconds": 0.10527157783508301 + }, + { + "epoch": 107, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.05003148424790965, + "train_seconds": 246.64277410507202, + "validation": { + "accuracy": 0.9102, + "examples": 5000, + "loss": 0.3456508056640625 + }, + "validation_seconds": 0.10487556457519531 + }, + { + "epoch": 108, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04837106784184774, + "train_seconds": 246.62830758094788, + "validation": { + "accuracy": 0.9082, + "examples": 5000, + "loss": 0.3555091491699219 + }, + "validation_seconds": 0.10479950904846191 + }, + { + "epoch": 109, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.046110330545239976, + "train_seconds": 246.64310789108276, + "validation": { + "accuracy": 0.9074, + "examples": 5000, + "loss": 0.36177589721679687 + }, + "validation_seconds": 0.10510897636413574 + }, + { + "epoch": 110, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04422715940343009, + "train_seconds": 246.65255451202393, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.35412333984375 + }, + "validation_seconds": 0.10495305061340332 + }, + { + "epoch": 111, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04302270753913456, + "train_seconds": 246.68583822250366, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.36001808471679686 + }, + "validation_seconds": 0.10499882698059082 + }, + { + "epoch": 112, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04221789358986749, + "train_seconds": 246.63890480995178, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.3641726318359375 + }, + "validation_seconds": 0.10499691963195801 + }, + { + "epoch": 113, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04019610329998864, + "train_seconds": 246.6600148677826, + "validation": { + "accuracy": 0.9112, + "examples": 5000, + "loss": 0.3682267639160156 + }, + "validation_seconds": 0.10518097877502441 + }, + { + "epoch": 114, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.038186527136299345, + "train_seconds": 246.6488721370697, + "validation": { + "accuracy": 0.907, + "examples": 5000, + "loss": 0.37181755981445314 + }, + "validation_seconds": 0.10516762733459473 + }, + { + "epoch": 115, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03772044310834673, + "train_seconds": 246.58393955230713, + "validation": { + "accuracy": 0.908, + "examples": 5000, + "loss": 0.37084036865234377 + }, + "validation_seconds": 0.10496854782104492 + }, + { + "epoch": 116, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03771770534714063, + "train_seconds": 246.46069836616516, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.3673696716308594 + }, + "validation_seconds": 0.10475349426269531 + }, + { + "epoch": 117, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03610480902194977, + "train_seconds": 246.45283031463623, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.3759578430175781 + }, + "validation_seconds": 0.10521340370178223 + }, + { + "epoch": 118, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03562721369995011, + "train_seconds": 246.5085368156433, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.377801953125 + }, + "validation_seconds": 0.10504674911499023 + }, + { + "epoch": 119, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0351684531306227, + "train_seconds": 246.4897859096527, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.3793407165527344 + }, + "validation_seconds": 0.1051180362701416 + }, + { + "epoch": 120, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03166843252413803, + "train_seconds": 246.504944562912, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.3807004577636719 + }, + "validation_seconds": 0.10520434379577637 + }, + { + "epoch": 121, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.032848191968599956, + "train_seconds": 246.5223376750946, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.3911223388671875 + }, + "validation_seconds": 0.10503935813903809 + }, + { + "epoch": 122, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.032460003905495007, + "train_seconds": 246.50015473365784, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.3859025024414062 + }, + "validation_seconds": 0.10502910614013672 + }, + { + "epoch": 123, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03331895860698488, + "train_seconds": 246.51648592948914, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.3818461975097656 + }, + "validation_seconds": 0.10492110252380371 + }, + { + "epoch": 124, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.030804785005582702, + "train_seconds": 246.51632118225098, + "validation": { + "accuracy": 0.9082, + "examples": 5000, + "loss": 0.384043212890625 + }, + "validation_seconds": 0.10508346557617188 + }, + { + "epoch": 125, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03100608646472295, + "train_seconds": 246.48151516914368, + "validation": { + "accuracy": 0.9068, + "examples": 5000, + "loss": 0.3950748840332031 + }, + "validation_seconds": 0.10526251792907715 + }, + { + "epoch": 126, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.030607278982798258, + "train_seconds": 246.49509143829346, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.38821942749023436 + }, + "validation_seconds": 0.1049349308013916 + }, + { + "epoch": 127, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02984628584285577, + "train_seconds": 246.48997020721436, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.39522205810546873 + }, + "validation_seconds": 0.1047677993774414 + }, + { + "epoch": 128, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.028140256262818974, + "train_seconds": 246.5008955001831, + "validation": { + "accuracy": 0.9058, + "examples": 5000, + "loss": 0.3940588623046875 + }, + "validation_seconds": 0.10520315170288086 + }, + { + "epoch": 129, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.029524326505263647, + "train_seconds": 246.49133825302124, + "validation": { + "accuracy": 0.906, + "examples": 5000, + "loss": 0.40357665405273435 + }, + "validation_seconds": 0.10487580299377441 + }, + { + "epoch": 130, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.027381085895167457, + "train_seconds": 246.53742814064026, + "validation": { + "accuracy": 0.9058, + "examples": 5000, + "loss": 0.40291564331054686 + }, + "validation_seconds": 0.10510921478271484 + }, + { + "epoch": 131, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02712565425733725, + "train_seconds": 246.5171241760254, + "validation": { + "accuracy": 0.9066, + "examples": 5000, + "loss": 0.39782152099609375 + }, + "validation_seconds": 0.10495710372924805 + }, + { + "epoch": 132, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.027452782110704316, + "train_seconds": 246.53909969329834, + "validation": { + "accuracy": 0.9072, + "examples": 5000, + "loss": 0.40547529907226565 + }, + "validation_seconds": 0.1050260066986084 + }, + { + "epoch": 133, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02682856381237507, + "train_seconds": 246.52162146568298, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.400748779296875 + }, + "validation_seconds": 0.10499382019042969 + }, + { + "epoch": 134, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02624693241649204, + "train_seconds": 246.5282382965088, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4016249328613281 + }, + "validation_seconds": 0.10514974594116211 + }, + { + "epoch": 135, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02609049816512399, + "train_seconds": 246.50827550888062, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.40999682006835936 + }, + "validation_seconds": 0.10509872436523438 + }, + { + "epoch": 136, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02532903560532464, + "train_seconds": 246.4754238128662, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.40476075439453124 + }, + "validation_seconds": 0.10525250434875488 + }, + { + "epoch": 137, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.025431149927112793, + "train_seconds": 246.53441071510315, + "validation": { + "accuracy": 0.9102, + "examples": 5000, + "loss": 0.40455454711914063 + }, + "validation_seconds": 0.10527610778808594 + }, + { + "epoch": 138, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02547713861929046, + "train_seconds": 246.49636459350586, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.4073498596191406 + }, + "validation_seconds": 0.10490679740905762 + }, + { + "epoch": 139, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02374524391161071, + "train_seconds": 246.48351001739502, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4119003784179687 + }, + "validation_seconds": 0.1051790714263916 + }, + { + "epoch": 140, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.023575902736849254, + "train_seconds": 246.4875295162201, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.4121861877441406 + }, + "validation_seconds": 0.10484886169433594 + }, + { + "epoch": 141, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.025533000514242385, + "train_seconds": 246.53514981269836, + "validation": { + "accuracy": 0.911, + "examples": 5000, + "loss": 0.40740313720703125 + }, + "validation_seconds": 0.1048581600189209 + }, + { + "epoch": 142, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.023307849234673713, + "train_seconds": 246.49939465522766, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4205747985839844 + }, + "validation_seconds": 0.10505390167236328 + }, + { + "epoch": 143, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.024081955377260843, + "train_seconds": 246.50040888786316, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.42347489013671874 + }, + "validation_seconds": 0.10478520393371582 + }, + { + "epoch": 144, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02217308875951502, + "train_seconds": 246.50287771224976, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.4230736328125 + }, + "validation_seconds": 0.1047658920288086 + }, + { + "epoch": 145, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.021583292372690306, + "train_seconds": 246.49319005012512, + "validation": { + "accuracy": 0.9068, + "examples": 5000, + "loss": 0.4246301025390625 + }, + "validation_seconds": 0.10511493682861328 + }, + { + "epoch": 146, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.024299849601421092, + "train_seconds": 246.4877734184265, + "validation": { + "accuracy": 0.9064, + "examples": 5000, + "loss": 0.4284990051269531 + }, + "validation_seconds": 0.10503721237182617 + }, + { + "epoch": 147, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02203905286722713, + "train_seconds": 246.4707317352295, + "validation": { + "accuracy": 0.9056, + "examples": 5000, + "loss": 0.43494686279296874 + }, + "validation_seconds": 0.10479950904846191 + }, + { + "epoch": 148, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0230322131584088, + "train_seconds": 246.5043556690216, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.4222664306640625 + }, + "validation_seconds": 0.1047523021697998 + }, + { + "epoch": 149, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.021366101835833654, + "train_seconds": 246.5034623146057, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.4292074523925781 + }, + "validation_seconds": 0.10499930381774902 + }, + { + "epoch": 150, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.021379361391067504, + "train_seconds": 246.49155592918396, + "validation": { + "accuracy": 0.9062, + "examples": 5000, + "loss": 0.4323758117675781 + }, + "validation_seconds": 0.10504913330078125 + }, + { + "epoch": 151, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.019433662343025206, + "train_seconds": 246.48384475708008, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.4278642761230469 + }, + "validation_seconds": 0.10531210899353027 + }, + { + "epoch": 152, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.019419856379098363, + "train_seconds": 246.4657862186432, + "validation": { + "accuracy": 0.908, + "examples": 5000, + "loss": 0.4250206359863281 + }, + "validation_seconds": 0.10491490364074707 + }, + { + "epoch": 153, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01779731034901407, + "train_seconds": 246.51931881904602, + "validation": { + "accuracy": 0.907, + "examples": 5000, + "loss": 0.4278398681640625 + }, + "validation_seconds": 0.10487937927246094 + }, + { + "epoch": 154, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01879941195083989, + "train_seconds": 246.67745804786682, + "validation": { + "accuracy": 0.9066, + "examples": 5000, + "loss": 0.42561909790039065 + }, + "validation_seconds": 0.10490798950195312 + }, + { + "epoch": 155, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017670769177211654, + "train_seconds": 246.60387182235718, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.42608068237304686 + }, + "validation_seconds": 0.10499191284179688 + }, + { + "epoch": 156, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01898041370842192, + "train_seconds": 246.49284768104553, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.42472913818359376 + }, + "validation_seconds": 0.10478997230529785 + }, + { + "epoch": 157, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01866564043296708, + "train_seconds": 246.5001609325409, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.42693660888671875 + }, + "validation_seconds": 0.10493803024291992 + }, + { + "epoch": 158, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.018088148384003175, + "train_seconds": 246.44772601127625, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.424379541015625 + }, + "validation_seconds": 0.1050882339477539 + }, + { + "epoch": 159, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01760909207902021, + "train_seconds": 246.48644709587097, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.4213317810058594 + }, + "validation_seconds": 0.10483264923095703 + }, + { + "epoch": 160, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01838235599050919, + "train_seconds": 246.49702405929565, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.4235311767578125 + }, + "validation_seconds": 0.10515522956848145 + }, + { + "epoch": 161, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017624795015901328, + "train_seconds": 246.47225213050842, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.42268819580078126 + }, + "validation_seconds": 0.10508155822753906 + }, + { + "epoch": 162, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016919317315518857, + "train_seconds": 246.48047471046448, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.42747427368164065 + }, + "validation_seconds": 0.10546684265136719 + }, + { + "epoch": 163, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.018289856963025198, + "train_seconds": 246.46548342704773, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.424314111328125 + }, + "validation_seconds": 0.10488176345825195 + }, + { + "epoch": 164, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01622216137184037, + "train_seconds": 246.47679567337036, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.42815198974609375 + }, + "validation_seconds": 0.10511016845703125 + }, + { + "epoch": 165, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01683462748808993, + "train_seconds": 246.48844981193542, + "validation": { + "accuracy": 0.9082, + "examples": 5000, + "loss": 0.42897827758789064 + }, + "validation_seconds": 0.10507750511169434 + }, + { + "epoch": 166, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01676006626519892, + "train_seconds": 246.48650288581848, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.4275100830078125 + }, + "validation_seconds": 0.1049661636352539 + }, + { + "epoch": 167, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016608415773842068, + "train_seconds": 246.47876834869385, + "validation": { + "accuracy": 0.908, + "examples": 5000, + "loss": 0.4286995666503906 + }, + "validation_seconds": 0.10474586486816406 + }, + { + "epoch": 168, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017247727344764603, + "train_seconds": 246.4671652317047, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.4272683166503906 + }, + "validation_seconds": 0.10498309135437012 + }, + { + "epoch": 169, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016528946716917885, + "train_seconds": 246.58322262763977, + "validation": { + "accuracy": 0.9106, + "examples": 5000, + "loss": 0.4269004150390625 + }, + "validation_seconds": 0.10540938377380371 + }, + { + "epoch": 170, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016474853175216252, + "train_seconds": 246.69525337219238, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.42872039794921873 + }, + "validation_seconds": 0.10507774353027344 + }, + { + "epoch": 171, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016024321196476617, + "train_seconds": 246.60345554351807, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4304836059570312 + }, + "validation_seconds": 0.10527491569519043 + }, + { + "epoch": 172, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017347405864298344, + "train_seconds": 246.6121952533722, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.42888103637695313 + }, + "validation_seconds": 0.10518264770507812 + }, + { + "epoch": 173, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016893391016125678, + "train_seconds": 246.6187880039215, + "validation": { + "accuracy": 0.9098, + "examples": 5000, + "loss": 0.4287686279296875 + }, + "validation_seconds": 0.1053009033203125 + }, + { + "epoch": 174, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016590854315045808, + "train_seconds": 246.58989000320435, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.42554988403320315 + }, + "validation_seconds": 0.10519289970397949 + }, + { + "epoch": 175, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01584643976688385, + "train_seconds": 246.5360791683197, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.425059716796875 + }, + "validation_seconds": 0.10481643676757812 + }, + { + "epoch": 176, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016978252725137605, + "train_seconds": 246.5840287208557, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.4239813842773438 + }, + "validation_seconds": 0.1047203540802002 + }, + { + "epoch": 177, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015669997653530703, + "train_seconds": 246.60086369514465, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.4284854431152344 + }, + "validation_seconds": 0.10500645637512207 + }, + { + "epoch": 178, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017344997499841783, + "train_seconds": 246.62754106521606, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4284616516113281 + }, + "validation_seconds": 0.10516667366027832 + }, + { + "epoch": 179, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016015476820866268, + "train_seconds": 246.58459210395813, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.4248768615722656 + }, + "validation_seconds": 0.1052248477935791 + }, + { + "epoch": 180, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016324492257336776, + "train_seconds": 246.54819989204407, + "validation": { + "accuracy": 0.9108, + "examples": 5000, + "loss": 0.42777464599609377 + }, + "validation_seconds": 0.10559630393981934 + }, + { + "epoch": 181, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01666753611365954, + "train_seconds": 246.57083463668823, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.42751864013671875 + }, + "validation_seconds": 0.10503649711608887 + }, + { + "epoch": 182, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016868565215335952, + "train_seconds": 246.51821660995483, + "validation": { + "accuracy": 0.91, + "examples": 5000, + "loss": 0.43049243774414064 + }, + "validation_seconds": 0.1048879623413086 + }, + { + "epoch": 183, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017401854633788266, + "train_seconds": 246.54656767845154, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.43035255126953126 + }, + "validation_seconds": 0.10518288612365723 + }, + { + "epoch": 184, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01609395571152369, + "train_seconds": 246.47771430015564, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.43076982421875 + }, + "validation_seconds": 0.10487246513366699 + }, + { + "epoch": 185, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015626564153035483, + "train_seconds": 246.0274088382721, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.43607783813476564 + }, + "validation_seconds": 0.10477232933044434 + }, + { + "epoch": 186, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016511785845789644, + "train_seconds": 246.0552146434784, + "validation": { + "accuracy": 0.9102, + "examples": 5000, + "loss": 0.43122318115234376 + }, + "validation_seconds": 0.10496950149536133 + }, + { + "epoch": 187, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016176714078916445, + "train_seconds": 246.01906418800354, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.43309248046875 + }, + "validation_seconds": 0.10488486289978027 + }, + { + "epoch": 188, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017060648574762874, + "train_seconds": 246.00013732910156, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.43197459106445313 + }, + "validation_seconds": 0.1047358512878418 + }, + { + "epoch": 189, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015920067586998144, + "train_seconds": 246.03866243362427, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.42890040283203124 + }, + "validation_seconds": 0.10486721992492676 + }, + { + "epoch": 190, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015953422194719315, + "train_seconds": 246.05727362632751, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.42850841064453127 + }, + "validation_seconds": 0.1049807071685791 + }, + { + "epoch": 191, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015881843958629502, + "train_seconds": 246.0005452632904, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4348256774902344 + }, + "validation_seconds": 0.10516357421875 + }, + { + "epoch": 192, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.014872584097584088, + "train_seconds": 246.02225160598755, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.43110437622070313 + }, + "validation_seconds": 0.10500121116638184 + }, + { + "epoch": 193, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016930411442783143, + "train_seconds": 246.41416931152344, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.4323291320800781 + }, + "validation_seconds": 0.1049497127532959 + }, + { + "epoch": 194, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015967371432648765, + "train_seconds": 246.44947600364685, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.4302525146484375 + }, + "validation_seconds": 0.10496187210083008 + }, + { + "epoch": 195, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015971045525537596, + "train_seconds": 246.44577145576477, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.42978770141601563 + }, + "validation_seconds": 0.10512423515319824 + }, + { + "epoch": 196, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015177000864677959, + "train_seconds": 246.432710647583, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.432722900390625 + }, + "validation_seconds": 0.10512948036193848 + }, + { + "epoch": 197, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017064794731885196, + "train_seconds": 246.43957209587097, + "validation": { + "accuracy": 0.9098, + "examples": 5000, + "loss": 0.43294736938476563 + }, + "validation_seconds": 0.10485196113586426 + }, + { + "epoch": 198, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015426826933191883, + "train_seconds": 246.46257519721985, + "validation": { + "accuracy": 0.9102, + "examples": 5000, + "loss": 0.4334692504882813 + }, + "validation_seconds": 0.10505104064941406 + }, + { + "epoch": 199, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016626012229753864, + "train_seconds": 246.4922273159027, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.430858544921875 + }, + "validation_seconds": 0.10512781143188477 + }, + { + "epoch": 200, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015557368935479057, + "train_seconds": 246.5367796421051, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.43468799438476563 + }, + "validation_seconds": 0.10501646995544434 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.9104, + "evaluation_seconds": 0.10538721084594727, + "examples": 5000, + "finite": true, + "loss": 0.43468799438476563 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2008608768, + "peak_memory_reserved_bytes": 2808086528, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 49286.44217157364, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 200, + "feedforward_example_passes": 10005000, + "forward_macs_per_example": 40551040, + "forward_parameter_count": 269722, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 144000000, + "logical_task_loss_queries": 0, + "num_trainable_layers": 20, + "ordinary_training_examples": 9000000, + "ordinary_validation_examples": 1005000, + "relaxation_example_passes": 144000000, + "training_example_presentations": 9000000 + } +} |
