|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 10.689778829184034, |
|
"eval_steps": 1000000, |
|
"global_step": 569359, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 5.000000000000001e-07, |
|
"loss": 10.329, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 1.0000000000000002e-06, |
|
"loss": 10.2616, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 1.5e-06, |
|
"loss": 10.2023, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 2.0000000000000003e-06, |
|
"loss": 10.1392, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 2.5e-06, |
|
"loss": 10.0621, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 3e-06, |
|
"loss": 9.9676, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 3.5e-06, |
|
"loss": 9.857, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 4.000000000000001e-06, |
|
"loss": 9.731, |
|
"step": 4000 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"learning_rate": 4.5e-06, |
|
"loss": 9.5912, |
|
"step": 4500 |
|
}, |
|
{ |
|
"epoch": 0.09, |
|
"learning_rate": 5e-06, |
|
"loss": 9.4379, |
|
"step": 5000 |
|
}, |
|
{ |
|
"epoch": 0.1, |
|
"learning_rate": 5.500000000000001e-06, |
|
"loss": 9.2753, |
|
"step": 5500 |
|
}, |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 6e-06, |
|
"loss": 9.1037, |
|
"step": 6000 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"learning_rate": 6.5000000000000004e-06, |
|
"loss": 8.9243, |
|
"step": 6500 |
|
}, |
|
{ |
|
"epoch": 0.13, |
|
"learning_rate": 7e-06, |
|
"loss": 8.7434, |
|
"step": 7000 |
|
}, |
|
{ |
|
"epoch": 0.14, |
|
"learning_rate": 7.500000000000001e-06, |
|
"loss": 8.5585, |
|
"step": 7500 |
|
}, |
|
{ |
|
"epoch": 0.15, |
|
"learning_rate": 8.000000000000001e-06, |
|
"loss": 8.3759, |
|
"step": 8000 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"learning_rate": 8.5e-06, |
|
"loss": 8.1965, |
|
"step": 8500 |
|
}, |
|
{ |
|
"epoch": 0.17, |
|
"learning_rate": 9e-06, |
|
"loss": 8.0233, |
|
"step": 9000 |
|
}, |
|
{ |
|
"epoch": 0.18, |
|
"learning_rate": 9.5e-06, |
|
"loss": 7.8604, |
|
"step": 9500 |
|
}, |
|
{ |
|
"epoch": 0.19, |
|
"learning_rate": 1e-05, |
|
"loss": 7.7141, |
|
"step": 10000 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"learning_rate": 9.999059478574923e-06, |
|
"loss": 7.5804, |
|
"step": 10500 |
|
}, |
|
{ |
|
"epoch": 0.21, |
|
"learning_rate": 9.998118957149845e-06, |
|
"loss": 7.4725, |
|
"step": 11000 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 9.997178435724768e-06, |
|
"loss": 7.3918, |
|
"step": 11500 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 9.99623791429969e-06, |
|
"loss": 7.3219, |
|
"step": 12000 |
|
}, |
|
{ |
|
"epoch": 0.23, |
|
"learning_rate": 9.99529739287461e-06, |
|
"loss": 7.2767, |
|
"step": 12500 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"learning_rate": 9.994356871449533e-06, |
|
"loss": 7.2424, |
|
"step": 13000 |
|
}, |
|
{ |
|
"epoch": 0.25, |
|
"learning_rate": 9.993416350024455e-06, |
|
"loss": 7.2083, |
|
"step": 13500 |
|
}, |
|
{ |
|
"epoch": 0.26, |
|
"learning_rate": 9.992475828599375e-06, |
|
"loss": 7.1793, |
|
"step": 14000 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"learning_rate": 9.991535307174298e-06, |
|
"loss": 7.1464, |
|
"step": 14500 |
|
}, |
|
{ |
|
"epoch": 0.28, |
|
"learning_rate": 9.99059478574922e-06, |
|
"loss": 7.1186, |
|
"step": 15000 |
|
}, |
|
{ |
|
"epoch": 0.29, |
|
"learning_rate": 9.989654264324142e-06, |
|
"loss": 7.099, |
|
"step": 15500 |
|
}, |
|
{ |
|
"epoch": 0.3, |
|
"learning_rate": 9.988713742899064e-06, |
|
"loss": 7.0719, |
|
"step": 16000 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"learning_rate": 9.987773221473987e-06, |
|
"loss": 7.0556, |
|
"step": 16500 |
|
}, |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 9.986832700048909e-06, |
|
"loss": 7.0397, |
|
"step": 17000 |
|
}, |
|
{ |
|
"epoch": 0.33, |
|
"learning_rate": 9.98589217862383e-06, |
|
"loss": 7.0193, |
|
"step": 17500 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 9.984951657198752e-06, |
|
"loss": 7.0082, |
|
"step": 18000 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"learning_rate": 9.984011135773674e-06, |
|
"loss": 6.9907, |
|
"step": 18500 |
|
}, |
|
{ |
|
"epoch": 0.36, |
|
"learning_rate": 9.983070614348596e-06, |
|
"loss": 6.9759, |
|
"step": 19000 |
|
}, |
|
{ |
|
"epoch": 0.37, |
|
"learning_rate": 9.982130092923518e-06, |
|
"loss": 6.9642, |
|
"step": 19500 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 9.981189571498439e-06, |
|
"loss": 6.9474, |
|
"step": 20000 |
|
}, |
|
{ |
|
"epoch": 0.38, |
|
"learning_rate": 9.980249050073361e-06, |
|
"loss": 6.932, |
|
"step": 20500 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"learning_rate": 9.979308528648283e-06, |
|
"loss": 6.9252, |
|
"step": 21000 |
|
}, |
|
{ |
|
"epoch": 0.4, |
|
"learning_rate": 9.978368007223206e-06, |
|
"loss": 6.9113, |
|
"step": 21500 |
|
}, |
|
{ |
|
"epoch": 0.41, |
|
"learning_rate": 9.977427485798126e-06, |
|
"loss": 6.9012, |
|
"step": 22000 |
|
}, |
|
{ |
|
"epoch": 0.42, |
|
"learning_rate": 9.976486964373048e-06, |
|
"loss": 6.8903, |
|
"step": 22500 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"learning_rate": 9.97554644294797e-06, |
|
"loss": 6.8752, |
|
"step": 23000 |
|
}, |
|
{ |
|
"epoch": 0.44, |
|
"learning_rate": 9.974605921522893e-06, |
|
"loss": 6.8672, |
|
"step": 23500 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 9.973665400097815e-06, |
|
"loss": 6.8566, |
|
"step": 24000 |
|
}, |
|
{ |
|
"epoch": 0.46, |
|
"learning_rate": 9.972724878672737e-06, |
|
"loss": 6.8466, |
|
"step": 24500 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"learning_rate": 9.97178435724766e-06, |
|
"loss": 6.8442, |
|
"step": 25000 |
|
}, |
|
{ |
|
"epoch": 0.48, |
|
"learning_rate": 9.970843835822582e-06, |
|
"loss": 6.8263, |
|
"step": 25500 |
|
}, |
|
{ |
|
"epoch": 0.49, |
|
"learning_rate": 9.969903314397504e-06, |
|
"loss": 6.8213, |
|
"step": 26000 |
|
}, |
|
{ |
|
"epoch": 0.5, |
|
"learning_rate": 9.968962792972425e-06, |
|
"loss": 6.8155, |
|
"step": 26500 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"learning_rate": 9.968022271547347e-06, |
|
"loss": 6.8035, |
|
"step": 27000 |
|
}, |
|
{ |
|
"epoch": 0.52, |
|
"learning_rate": 9.967081750122269e-06, |
|
"loss": 6.7945, |
|
"step": 27500 |
|
}, |
|
{ |
|
"epoch": 0.53, |
|
"learning_rate": 9.96614122869719e-06, |
|
"loss": 6.7894, |
|
"step": 28000 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 9.965200707272112e-06, |
|
"loss": 6.7814, |
|
"step": 28500 |
|
}, |
|
{ |
|
"epoch": 0.54, |
|
"learning_rate": 9.964260185847034e-06, |
|
"loss": 6.7751, |
|
"step": 29000 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"learning_rate": 9.963319664421956e-06, |
|
"loss": 6.7699, |
|
"step": 29500 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 9.962379142996879e-06, |
|
"loss": 6.756, |
|
"step": 30000 |
|
}, |
|
{ |
|
"epoch": 0.57, |
|
"learning_rate": 9.961438621571801e-06, |
|
"loss": 6.7525, |
|
"step": 30500 |
|
}, |
|
{ |
|
"epoch": 0.58, |
|
"learning_rate": 9.960498100146723e-06, |
|
"loss": 6.7494, |
|
"step": 31000 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"learning_rate": 9.959557578721644e-06, |
|
"loss": 6.7416, |
|
"step": 31500 |
|
}, |
|
{ |
|
"epoch": 0.6, |
|
"learning_rate": 9.958617057296566e-06, |
|
"loss": 6.735, |
|
"step": 32000 |
|
}, |
|
{ |
|
"epoch": 0.61, |
|
"learning_rate": 9.957676535871488e-06, |
|
"loss": 6.7314, |
|
"step": 32500 |
|
}, |
|
{ |
|
"epoch": 0.62, |
|
"learning_rate": 9.95673601444641e-06, |
|
"loss": 6.7211, |
|
"step": 33000 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"learning_rate": 9.955795493021333e-06, |
|
"loss": 6.7171, |
|
"step": 33500 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 9.954854971596253e-06, |
|
"loss": 6.7101, |
|
"step": 34000 |
|
}, |
|
{ |
|
"epoch": 0.65, |
|
"learning_rate": 9.953914450171175e-06, |
|
"loss": 6.707, |
|
"step": 34500 |
|
}, |
|
{ |
|
"epoch": 0.66, |
|
"learning_rate": 9.952973928746098e-06, |
|
"loss": 6.7042, |
|
"step": 35000 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 9.95203340732102e-06, |
|
"loss": 6.6962, |
|
"step": 35500 |
|
}, |
|
{ |
|
"epoch": 0.68, |
|
"learning_rate": 9.95109288589594e-06, |
|
"loss": 6.6903, |
|
"step": 36000 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 9.950152364470863e-06, |
|
"loss": 6.6825, |
|
"step": 36500 |
|
}, |
|
{ |
|
"epoch": 0.69, |
|
"learning_rate": 9.949211843045785e-06, |
|
"loss": 6.6813, |
|
"step": 37000 |
|
}, |
|
{ |
|
"epoch": 0.7, |
|
"learning_rate": 9.948271321620707e-06, |
|
"loss": 6.6745, |
|
"step": 37500 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"learning_rate": 9.94733080019563e-06, |
|
"loss": 6.6646, |
|
"step": 38000 |
|
}, |
|
{ |
|
"epoch": 0.72, |
|
"learning_rate": 9.946390278770552e-06, |
|
"loss": 6.6639, |
|
"step": 38500 |
|
}, |
|
{ |
|
"epoch": 0.73, |
|
"learning_rate": 9.945449757345474e-06, |
|
"loss": 6.6576, |
|
"step": 39000 |
|
}, |
|
{ |
|
"epoch": 0.74, |
|
"learning_rate": 9.944509235920396e-06, |
|
"loss": 6.6543, |
|
"step": 39500 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"learning_rate": 9.943568714495317e-06, |
|
"loss": 6.6494, |
|
"step": 40000 |
|
}, |
|
{ |
|
"epoch": 0.76, |
|
"learning_rate": 9.942628193070239e-06, |
|
"loss": 6.6475, |
|
"step": 40500 |
|
}, |
|
{ |
|
"epoch": 0.77, |
|
"learning_rate": 9.941687671645161e-06, |
|
"loss": 6.6446, |
|
"step": 41000 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 9.940747150220082e-06, |
|
"loss": 6.6374, |
|
"step": 41500 |
|
}, |
|
{ |
|
"epoch": 0.79, |
|
"learning_rate": 9.939806628795004e-06, |
|
"loss": 6.6312, |
|
"step": 42000 |
|
}, |
|
{ |
|
"epoch": 0.8, |
|
"learning_rate": 9.938866107369926e-06, |
|
"loss": 6.6318, |
|
"step": 42500 |
|
}, |
|
{ |
|
"epoch": 0.81, |
|
"learning_rate": 9.937925585944848e-06, |
|
"loss": 6.6253, |
|
"step": 43000 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"learning_rate": 9.93698506451977e-06, |
|
"loss": 6.6248, |
|
"step": 43500 |
|
}, |
|
{ |
|
"epoch": 0.83, |
|
"learning_rate": 9.936044543094693e-06, |
|
"loss": 6.616, |
|
"step": 44000 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 9.935104021669615e-06, |
|
"loss": 6.6109, |
|
"step": 44500 |
|
}, |
|
{ |
|
"epoch": 0.84, |
|
"learning_rate": 9.934163500244537e-06, |
|
"loss": 6.6124, |
|
"step": 45000 |
|
}, |
|
{ |
|
"epoch": 0.85, |
|
"learning_rate": 9.933222978819458e-06, |
|
"loss": 6.6054, |
|
"step": 45500 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"learning_rate": 9.93228245739438e-06, |
|
"loss": 6.5991, |
|
"step": 46000 |
|
}, |
|
{ |
|
"epoch": 0.87, |
|
"learning_rate": 9.931341935969302e-06, |
|
"loss": 6.6008, |
|
"step": 46500 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 9.930401414544225e-06, |
|
"loss": 6.5993, |
|
"step": 47000 |
|
}, |
|
{ |
|
"epoch": 0.89, |
|
"learning_rate": 9.929460893119145e-06, |
|
"loss": 6.5916, |
|
"step": 47500 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 9.928520371694067e-06, |
|
"loss": 6.5909, |
|
"step": 48000 |
|
}, |
|
{ |
|
"epoch": 0.91, |
|
"learning_rate": 9.92757985026899e-06, |
|
"loss": 6.586, |
|
"step": 48500 |
|
}, |
|
{ |
|
"epoch": 0.92, |
|
"learning_rate": 9.926639328843912e-06, |
|
"loss": 6.5913, |
|
"step": 49000 |
|
}, |
|
{ |
|
"epoch": 0.93, |
|
"learning_rate": 9.925698807418834e-06, |
|
"loss": 6.5776, |
|
"step": 49500 |
|
}, |
|
{ |
|
"epoch": 0.94, |
|
"learning_rate": 9.924758285993755e-06, |
|
"loss": 6.5808, |
|
"step": 50000 |
|
}, |
|
{ |
|
"epoch": 0.95, |
|
"learning_rate": 9.923817764568677e-06, |
|
"loss": 6.5807, |
|
"step": 50500 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 9.9228772431436e-06, |
|
"loss": 6.5744, |
|
"step": 51000 |
|
}, |
|
{ |
|
"epoch": 0.97, |
|
"learning_rate": 9.921936721718521e-06, |
|
"loss": 6.5638, |
|
"step": 51500 |
|
}, |
|
{ |
|
"epoch": 0.98, |
|
"learning_rate": 9.920996200293444e-06, |
|
"loss": 6.5719, |
|
"step": 52000 |
|
}, |
|
{ |
|
"epoch": 0.99, |
|
"learning_rate": 9.920055678868366e-06, |
|
"loss": 6.5617, |
|
"step": 52500 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 9.919115157443288e-06, |
|
"loss": 6.5638, |
|
"step": 53000 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 9.91817463601821e-06, |
|
"loss": 6.5574, |
|
"step": 53500 |
|
}, |
|
{ |
|
"epoch": 1.01, |
|
"learning_rate": 9.917234114593131e-06, |
|
"loss": 6.5556, |
|
"step": 54000 |
|
}, |
|
{ |
|
"epoch": 1.02, |
|
"learning_rate": 9.916293593168053e-06, |
|
"loss": 6.5515, |
|
"step": 54500 |
|
}, |
|
{ |
|
"epoch": 1.03, |
|
"learning_rate": 9.915353071742975e-06, |
|
"loss": 6.5479, |
|
"step": 55000 |
|
}, |
|
{ |
|
"epoch": 1.04, |
|
"learning_rate": 9.914412550317896e-06, |
|
"loss": 6.541, |
|
"step": 55500 |
|
}, |
|
{ |
|
"epoch": 1.05, |
|
"learning_rate": 9.913472028892818e-06, |
|
"loss": 6.5401, |
|
"step": 56000 |
|
}, |
|
{ |
|
"epoch": 1.06, |
|
"learning_rate": 9.91253150746774e-06, |
|
"loss": 6.5378, |
|
"step": 56500 |
|
}, |
|
{ |
|
"epoch": 1.07, |
|
"learning_rate": 9.911590986042663e-06, |
|
"loss": 6.5379, |
|
"step": 57000 |
|
}, |
|
{ |
|
"epoch": 1.08, |
|
"learning_rate": 9.910650464617585e-06, |
|
"loss": 6.5332, |
|
"step": 57500 |
|
}, |
|
{ |
|
"epoch": 1.09, |
|
"learning_rate": 9.909709943192507e-06, |
|
"loss": 6.5325, |
|
"step": 58000 |
|
}, |
|
{ |
|
"epoch": 1.1, |
|
"learning_rate": 9.90876942176743e-06, |
|
"loss": 6.5297, |
|
"step": 58500 |
|
}, |
|
{ |
|
"epoch": 1.11, |
|
"learning_rate": 9.907828900342352e-06, |
|
"loss": 6.523, |
|
"step": 59000 |
|
}, |
|
{ |
|
"epoch": 1.12, |
|
"learning_rate": 9.906888378917272e-06, |
|
"loss": 6.5198, |
|
"step": 59500 |
|
}, |
|
{ |
|
"epoch": 1.13, |
|
"learning_rate": 9.905947857492194e-06, |
|
"loss": 6.5197, |
|
"step": 60000 |
|
}, |
|
{ |
|
"epoch": 1.14, |
|
"learning_rate": 9.905007336067117e-06, |
|
"loss": 6.5193, |
|
"step": 60500 |
|
}, |
|
{ |
|
"epoch": 1.15, |
|
"learning_rate": 9.904066814642039e-06, |
|
"loss": 6.5161, |
|
"step": 61000 |
|
}, |
|
{ |
|
"epoch": 1.15, |
|
"learning_rate": 9.90312629321696e-06, |
|
"loss": 6.5136, |
|
"step": 61500 |
|
}, |
|
{ |
|
"epoch": 1.16, |
|
"learning_rate": 9.902185771791882e-06, |
|
"loss": 6.5058, |
|
"step": 62000 |
|
}, |
|
{ |
|
"epoch": 1.17, |
|
"learning_rate": 9.901245250366804e-06, |
|
"loss": 6.5103, |
|
"step": 62500 |
|
}, |
|
{ |
|
"epoch": 1.18, |
|
"learning_rate": 9.900304728941726e-06, |
|
"loss": 6.5016, |
|
"step": 63000 |
|
}, |
|
{ |
|
"epoch": 1.19, |
|
"learning_rate": 9.899364207516648e-06, |
|
"loss": 6.4994, |
|
"step": 63500 |
|
}, |
|
{ |
|
"epoch": 1.2, |
|
"learning_rate": 9.898423686091569e-06, |
|
"loss": 6.5003, |
|
"step": 64000 |
|
}, |
|
{ |
|
"epoch": 1.21, |
|
"learning_rate": 9.897483164666491e-06, |
|
"loss": 6.4922, |
|
"step": 64500 |
|
}, |
|
{ |
|
"epoch": 1.22, |
|
"learning_rate": 9.896542643241414e-06, |
|
"loss": 6.4953, |
|
"step": 65000 |
|
}, |
|
{ |
|
"epoch": 1.23, |
|
"learning_rate": 9.895602121816336e-06, |
|
"loss": 6.4871, |
|
"step": 65500 |
|
}, |
|
{ |
|
"epoch": 1.24, |
|
"learning_rate": 9.894661600391258e-06, |
|
"loss": 6.4882, |
|
"step": 66000 |
|
}, |
|
{ |
|
"epoch": 1.25, |
|
"learning_rate": 9.89372107896618e-06, |
|
"loss": 6.4836, |
|
"step": 66500 |
|
}, |
|
{ |
|
"epoch": 1.26, |
|
"learning_rate": 9.892780557541102e-06, |
|
"loss": 6.4858, |
|
"step": 67000 |
|
}, |
|
{ |
|
"epoch": 1.27, |
|
"learning_rate": 9.891840036116023e-06, |
|
"loss": 6.4862, |
|
"step": 67500 |
|
}, |
|
{ |
|
"epoch": 1.28, |
|
"learning_rate": 9.890899514690945e-06, |
|
"loss": 6.4821, |
|
"step": 68000 |
|
}, |
|
{ |
|
"epoch": 1.29, |
|
"learning_rate": 9.889958993265868e-06, |
|
"loss": 6.484, |
|
"step": 68500 |
|
}, |
|
{ |
|
"epoch": 1.3, |
|
"learning_rate": 9.889018471840788e-06, |
|
"loss": 6.4788, |
|
"step": 69000 |
|
}, |
|
{ |
|
"epoch": 1.3, |
|
"learning_rate": 9.88807795041571e-06, |
|
"loss": 6.4759, |
|
"step": 69500 |
|
}, |
|
{ |
|
"epoch": 1.31, |
|
"learning_rate": 9.887137428990633e-06, |
|
"loss": 6.4748, |
|
"step": 70000 |
|
}, |
|
{ |
|
"epoch": 1.32, |
|
"learning_rate": 9.886196907565555e-06, |
|
"loss": 6.4667, |
|
"step": 70500 |
|
}, |
|
{ |
|
"epoch": 1.33, |
|
"learning_rate": 9.885256386140477e-06, |
|
"loss": 6.4713, |
|
"step": 71000 |
|
}, |
|
{ |
|
"epoch": 1.34, |
|
"learning_rate": 9.8843158647154e-06, |
|
"loss": 6.4643, |
|
"step": 71500 |
|
}, |
|
{ |
|
"epoch": 1.35, |
|
"learning_rate": 9.883375343290322e-06, |
|
"loss": 6.4652, |
|
"step": 72000 |
|
}, |
|
{ |
|
"epoch": 1.36, |
|
"learning_rate": 9.882434821865244e-06, |
|
"loss": 6.4617, |
|
"step": 72500 |
|
}, |
|
{ |
|
"epoch": 1.37, |
|
"learning_rate": 9.881494300440166e-06, |
|
"loss": 6.4558, |
|
"step": 73000 |
|
}, |
|
{ |
|
"epoch": 1.38, |
|
"learning_rate": 9.880553779015087e-06, |
|
"loss": 6.4555, |
|
"step": 73500 |
|
}, |
|
{ |
|
"epoch": 1.39, |
|
"learning_rate": 9.879613257590009e-06, |
|
"loss": 6.4556, |
|
"step": 74000 |
|
}, |
|
{ |
|
"epoch": 1.4, |
|
"learning_rate": 9.878672736164931e-06, |
|
"loss": 6.4593, |
|
"step": 74500 |
|
}, |
|
{ |
|
"epoch": 1.41, |
|
"learning_rate": 9.877732214739853e-06, |
|
"loss": 6.4507, |
|
"step": 75000 |
|
}, |
|
{ |
|
"epoch": 1.42, |
|
"learning_rate": 9.876791693314774e-06, |
|
"loss": 6.4496, |
|
"step": 75500 |
|
}, |
|
{ |
|
"epoch": 1.43, |
|
"learning_rate": 9.875851171889696e-06, |
|
"loss": 6.4463, |
|
"step": 76000 |
|
}, |
|
{ |
|
"epoch": 1.44, |
|
"learning_rate": 9.874910650464618e-06, |
|
"loss": 6.4519, |
|
"step": 76500 |
|
}, |
|
{ |
|
"epoch": 1.45, |
|
"learning_rate": 9.87397012903954e-06, |
|
"loss": 6.4496, |
|
"step": 77000 |
|
}, |
|
{ |
|
"epoch": 1.46, |
|
"learning_rate": 9.873029607614463e-06, |
|
"loss": 6.4449, |
|
"step": 77500 |
|
}, |
|
{ |
|
"epoch": 1.46, |
|
"learning_rate": 9.872089086189383e-06, |
|
"loss": 6.44, |
|
"step": 78000 |
|
}, |
|
{ |
|
"epoch": 1.47, |
|
"learning_rate": 9.871148564764306e-06, |
|
"loss": 6.4384, |
|
"step": 78500 |
|
}, |
|
{ |
|
"epoch": 1.48, |
|
"learning_rate": 9.870208043339228e-06, |
|
"loss": 6.4376, |
|
"step": 79000 |
|
}, |
|
{ |
|
"epoch": 1.49, |
|
"learning_rate": 9.86926752191415e-06, |
|
"loss": 6.4316, |
|
"step": 79500 |
|
}, |
|
{ |
|
"epoch": 1.5, |
|
"learning_rate": 9.868327000489072e-06, |
|
"loss": 6.4302, |
|
"step": 80000 |
|
}, |
|
{ |
|
"epoch": 1.51, |
|
"learning_rate": 9.867386479063995e-06, |
|
"loss": 6.4286, |
|
"step": 80500 |
|
}, |
|
{ |
|
"epoch": 1.52, |
|
"learning_rate": 9.866445957638917e-06, |
|
"loss": 6.4263, |
|
"step": 81000 |
|
}, |
|
{ |
|
"epoch": 1.53, |
|
"learning_rate": 9.865505436213837e-06, |
|
"loss": 6.4286, |
|
"step": 81500 |
|
}, |
|
{ |
|
"epoch": 1.54, |
|
"learning_rate": 9.86456491478876e-06, |
|
"loss": 6.423, |
|
"step": 82000 |
|
}, |
|
{ |
|
"epoch": 1.55, |
|
"learning_rate": 9.863624393363682e-06, |
|
"loss": 6.428, |
|
"step": 82500 |
|
}, |
|
{ |
|
"epoch": 1.56, |
|
"learning_rate": 9.862683871938602e-06, |
|
"loss": 6.43, |
|
"step": 83000 |
|
}, |
|
{ |
|
"epoch": 1.57, |
|
"learning_rate": 9.861743350513525e-06, |
|
"loss": 6.4203, |
|
"step": 83500 |
|
}, |
|
{ |
|
"epoch": 1.58, |
|
"learning_rate": 9.860802829088447e-06, |
|
"loss": 6.4214, |
|
"step": 84000 |
|
}, |
|
{ |
|
"epoch": 1.59, |
|
"learning_rate": 9.859862307663369e-06, |
|
"loss": 6.417, |
|
"step": 84500 |
|
}, |
|
{ |
|
"epoch": 1.6, |
|
"learning_rate": 9.858921786238291e-06, |
|
"loss": 6.4206, |
|
"step": 85000 |
|
}, |
|
{ |
|
"epoch": 1.61, |
|
"learning_rate": 9.857981264813214e-06, |
|
"loss": 6.4175, |
|
"step": 85500 |
|
}, |
|
{ |
|
"epoch": 1.61, |
|
"learning_rate": 9.857040743388136e-06, |
|
"loss": 6.4181, |
|
"step": 86000 |
|
}, |
|
{ |
|
"epoch": 1.62, |
|
"learning_rate": 9.856100221963058e-06, |
|
"loss": 6.4088, |
|
"step": 86500 |
|
}, |
|
{ |
|
"epoch": 1.63, |
|
"learning_rate": 9.85515970053798e-06, |
|
"loss": 6.4108, |
|
"step": 87000 |
|
}, |
|
{ |
|
"epoch": 1.64, |
|
"learning_rate": 9.854219179112901e-06, |
|
"loss": 6.4097, |
|
"step": 87500 |
|
}, |
|
{ |
|
"epoch": 1.65, |
|
"learning_rate": 9.853278657687823e-06, |
|
"loss": 6.4071, |
|
"step": 88000 |
|
}, |
|
{ |
|
"epoch": 1.66, |
|
"learning_rate": 9.852338136262745e-06, |
|
"loss": 6.4051, |
|
"step": 88500 |
|
}, |
|
{ |
|
"epoch": 1.67, |
|
"learning_rate": 9.851397614837666e-06, |
|
"loss": 6.4045, |
|
"step": 89000 |
|
}, |
|
{ |
|
"epoch": 1.68, |
|
"learning_rate": 9.850457093412588e-06, |
|
"loss": 6.4081, |
|
"step": 89500 |
|
}, |
|
{ |
|
"epoch": 1.69, |
|
"learning_rate": 9.84951657198751e-06, |
|
"loss": 6.4021, |
|
"step": 90000 |
|
}, |
|
{ |
|
"epoch": 1.7, |
|
"learning_rate": 9.848576050562433e-06, |
|
"loss": 6.3969, |
|
"step": 90500 |
|
}, |
|
{ |
|
"epoch": 1.71, |
|
"learning_rate": 9.847635529137355e-06, |
|
"loss": 6.4018, |
|
"step": 91000 |
|
}, |
|
{ |
|
"epoch": 1.72, |
|
"learning_rate": 9.846695007712277e-06, |
|
"loss": 6.3984, |
|
"step": 91500 |
|
}, |
|
{ |
|
"epoch": 1.73, |
|
"learning_rate": 9.845754486287198e-06, |
|
"loss": 6.396, |
|
"step": 92000 |
|
}, |
|
{ |
|
"epoch": 1.74, |
|
"learning_rate": 9.84481396486212e-06, |
|
"loss": 6.398, |
|
"step": 92500 |
|
}, |
|
{ |
|
"epoch": 1.75, |
|
"learning_rate": 9.843873443437042e-06, |
|
"loss": 6.394, |
|
"step": 93000 |
|
}, |
|
{ |
|
"epoch": 1.76, |
|
"learning_rate": 9.842932922011964e-06, |
|
"loss": 6.3921, |
|
"step": 93500 |
|
}, |
|
{ |
|
"epoch": 1.76, |
|
"learning_rate": 9.841992400586887e-06, |
|
"loss": 6.3893, |
|
"step": 94000 |
|
}, |
|
{ |
|
"epoch": 1.77, |
|
"learning_rate": 9.841051879161809e-06, |
|
"loss": 6.389, |
|
"step": 94500 |
|
}, |
|
{ |
|
"epoch": 1.78, |
|
"learning_rate": 9.84011135773673e-06, |
|
"loss": 6.3841, |
|
"step": 95000 |
|
}, |
|
{ |
|
"epoch": 1.79, |
|
"learning_rate": 9.839170836311652e-06, |
|
"loss": 6.3856, |
|
"step": 95500 |
|
}, |
|
{ |
|
"epoch": 1.8, |
|
"learning_rate": 9.838230314886574e-06, |
|
"loss": 6.3836, |
|
"step": 96000 |
|
}, |
|
{ |
|
"epoch": 1.81, |
|
"learning_rate": 9.837289793461496e-06, |
|
"loss": 6.3868, |
|
"step": 96500 |
|
}, |
|
{ |
|
"epoch": 1.82, |
|
"learning_rate": 9.836349272036417e-06, |
|
"loss": 6.3852, |
|
"step": 97000 |
|
}, |
|
{ |
|
"epoch": 1.83, |
|
"learning_rate": 9.835408750611339e-06, |
|
"loss": 6.3818, |
|
"step": 97500 |
|
}, |
|
{ |
|
"epoch": 1.84, |
|
"learning_rate": 9.834468229186261e-06, |
|
"loss": 6.3824, |
|
"step": 98000 |
|
}, |
|
{ |
|
"epoch": 1.85, |
|
"learning_rate": 9.833527707761183e-06, |
|
"loss": 6.3787, |
|
"step": 98500 |
|
}, |
|
{ |
|
"epoch": 1.86, |
|
"learning_rate": 9.832587186336106e-06, |
|
"loss": 6.3771, |
|
"step": 99000 |
|
}, |
|
{ |
|
"epoch": 1.87, |
|
"learning_rate": 9.831646664911028e-06, |
|
"loss": 6.3777, |
|
"step": 99500 |
|
}, |
|
{ |
|
"epoch": 1.88, |
|
"learning_rate": 9.83070614348595e-06, |
|
"loss": 6.3782, |
|
"step": 100000 |
|
}, |
|
{ |
|
"epoch": 1.89, |
|
"learning_rate": 9.829765622060872e-06, |
|
"loss": 6.3745, |
|
"step": 100500 |
|
}, |
|
{ |
|
"epoch": 1.9, |
|
"learning_rate": 9.828825100635795e-06, |
|
"loss": 6.3721, |
|
"step": 101000 |
|
}, |
|
{ |
|
"epoch": 1.91, |
|
"learning_rate": 9.827884579210715e-06, |
|
"loss": 6.372, |
|
"step": 101500 |
|
}, |
|
{ |
|
"epoch": 1.92, |
|
"learning_rate": 9.826944057785637e-06, |
|
"loss": 6.3696, |
|
"step": 102000 |
|
}, |
|
{ |
|
"epoch": 1.92, |
|
"learning_rate": 9.82600353636056e-06, |
|
"loss": 6.3688, |
|
"step": 102500 |
|
}, |
|
{ |
|
"epoch": 1.93, |
|
"learning_rate": 9.82506301493548e-06, |
|
"loss": 6.3719, |
|
"step": 103000 |
|
}, |
|
{ |
|
"epoch": 1.94, |
|
"learning_rate": 9.824122493510402e-06, |
|
"loss": 6.3643, |
|
"step": 103500 |
|
}, |
|
{ |
|
"epoch": 1.95, |
|
"learning_rate": 9.823181972085325e-06, |
|
"loss": 6.3668, |
|
"step": 104000 |
|
}, |
|
{ |
|
"epoch": 1.96, |
|
"learning_rate": 9.822241450660247e-06, |
|
"loss": 6.363, |
|
"step": 104500 |
|
}, |
|
{ |
|
"epoch": 1.97, |
|
"learning_rate": 9.821300929235169e-06, |
|
"loss": 6.3603, |
|
"step": 105000 |
|
}, |
|
{ |
|
"epoch": 1.98, |
|
"learning_rate": 9.820360407810091e-06, |
|
"loss": 6.3599, |
|
"step": 105500 |
|
}, |
|
{ |
|
"epoch": 1.99, |
|
"learning_rate": 9.819419886385014e-06, |
|
"loss": 6.3578, |
|
"step": 106000 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"learning_rate": 9.818479364959934e-06, |
|
"loss": 6.3626, |
|
"step": 106500 |
|
}, |
|
{ |
|
"epoch": 2.01, |
|
"learning_rate": 9.817538843534856e-06, |
|
"loss": 6.3577, |
|
"step": 107000 |
|
}, |
|
{ |
|
"epoch": 2.02, |
|
"learning_rate": 9.816598322109779e-06, |
|
"loss": 6.3573, |
|
"step": 107500 |
|
}, |
|
{ |
|
"epoch": 2.03, |
|
"learning_rate": 9.815657800684701e-06, |
|
"loss": 6.3532, |
|
"step": 108000 |
|
}, |
|
{ |
|
"epoch": 2.04, |
|
"learning_rate": 9.814717279259623e-06, |
|
"loss": 6.3568, |
|
"step": 108500 |
|
}, |
|
{ |
|
"epoch": 2.05, |
|
"learning_rate": 9.813776757834544e-06, |
|
"loss": 6.3536, |
|
"step": 109000 |
|
}, |
|
{ |
|
"epoch": 2.06, |
|
"learning_rate": 9.812836236409466e-06, |
|
"loss": 6.3504, |
|
"step": 109500 |
|
}, |
|
{ |
|
"epoch": 2.07, |
|
"learning_rate": 9.811895714984388e-06, |
|
"loss": 6.3499, |
|
"step": 110000 |
|
}, |
|
{ |
|
"epoch": 2.07, |
|
"learning_rate": 9.81095519355931e-06, |
|
"loss": 6.3555, |
|
"step": 110500 |
|
}, |
|
{ |
|
"epoch": 2.08, |
|
"learning_rate": 9.810014672134231e-06, |
|
"loss": 6.3496, |
|
"step": 111000 |
|
}, |
|
{ |
|
"epoch": 2.09, |
|
"learning_rate": 9.809074150709153e-06, |
|
"loss": 6.3466, |
|
"step": 111500 |
|
}, |
|
{ |
|
"epoch": 2.1, |
|
"learning_rate": 9.808133629284075e-06, |
|
"loss": 6.3501, |
|
"step": 112000 |
|
}, |
|
{ |
|
"epoch": 2.11, |
|
"learning_rate": 9.807193107858998e-06, |
|
"loss": 6.3493, |
|
"step": 112500 |
|
}, |
|
{ |
|
"epoch": 2.12, |
|
"learning_rate": 9.80625258643392e-06, |
|
"loss": 6.3473, |
|
"step": 113000 |
|
}, |
|
{ |
|
"epoch": 2.13, |
|
"learning_rate": 9.805312065008842e-06, |
|
"loss": 6.3449, |
|
"step": 113500 |
|
}, |
|
{ |
|
"epoch": 2.14, |
|
"learning_rate": 9.804371543583764e-06, |
|
"loss": 6.3413, |
|
"step": 114000 |
|
}, |
|
{ |
|
"epoch": 2.15, |
|
"learning_rate": 9.803431022158687e-06, |
|
"loss": 6.3436, |
|
"step": 114500 |
|
}, |
|
{ |
|
"epoch": 2.16, |
|
"learning_rate": 9.802490500733607e-06, |
|
"loss": 6.3399, |
|
"step": 115000 |
|
}, |
|
{ |
|
"epoch": 2.17, |
|
"learning_rate": 9.80154997930853e-06, |
|
"loss": 6.345, |
|
"step": 115500 |
|
}, |
|
{ |
|
"epoch": 2.18, |
|
"learning_rate": 9.800609457883452e-06, |
|
"loss": 6.3404, |
|
"step": 116000 |
|
}, |
|
{ |
|
"epoch": 2.19, |
|
"learning_rate": 9.799668936458372e-06, |
|
"loss": 6.3432, |
|
"step": 116500 |
|
}, |
|
{ |
|
"epoch": 2.2, |
|
"learning_rate": 9.798728415033294e-06, |
|
"loss": 6.3413, |
|
"step": 117000 |
|
}, |
|
{ |
|
"epoch": 2.21, |
|
"learning_rate": 9.797787893608217e-06, |
|
"loss": 6.3359, |
|
"step": 117500 |
|
}, |
|
{ |
|
"epoch": 2.22, |
|
"learning_rate": 9.796847372183139e-06, |
|
"loss": 6.3363, |
|
"step": 118000 |
|
}, |
|
{ |
|
"epoch": 2.22, |
|
"learning_rate": 9.795906850758061e-06, |
|
"loss": 6.3419, |
|
"step": 118500 |
|
}, |
|
{ |
|
"epoch": 2.23, |
|
"learning_rate": 9.794966329332983e-06, |
|
"loss": 6.3328, |
|
"step": 119000 |
|
}, |
|
{ |
|
"epoch": 2.24, |
|
"learning_rate": 9.794025807907906e-06, |
|
"loss": 6.3309, |
|
"step": 119500 |
|
}, |
|
{ |
|
"epoch": 2.25, |
|
"learning_rate": 9.793085286482828e-06, |
|
"loss": 6.3343, |
|
"step": 120000 |
|
}, |
|
{ |
|
"epoch": 2.26, |
|
"learning_rate": 9.792144765057748e-06, |
|
"loss": 6.3309, |
|
"step": 120500 |
|
}, |
|
{ |
|
"epoch": 2.27, |
|
"learning_rate": 9.79120424363267e-06, |
|
"loss": 6.3298, |
|
"step": 121000 |
|
}, |
|
{ |
|
"epoch": 2.28, |
|
"learning_rate": 9.790263722207593e-06, |
|
"loss": 6.3373, |
|
"step": 121500 |
|
}, |
|
{ |
|
"epoch": 2.29, |
|
"learning_rate": 9.789323200782515e-06, |
|
"loss": 6.3302, |
|
"step": 122000 |
|
}, |
|
{ |
|
"epoch": 2.3, |
|
"learning_rate": 9.788382679357436e-06, |
|
"loss": 6.3315, |
|
"step": 122500 |
|
}, |
|
{ |
|
"epoch": 2.31, |
|
"learning_rate": 9.787442157932358e-06, |
|
"loss": 6.3292, |
|
"step": 123000 |
|
}, |
|
{ |
|
"epoch": 2.32, |
|
"learning_rate": 9.78650163650728e-06, |
|
"loss": 6.3275, |
|
"step": 123500 |
|
}, |
|
{ |
|
"epoch": 2.33, |
|
"learning_rate": 9.785561115082202e-06, |
|
"loss": 6.3269, |
|
"step": 124000 |
|
}, |
|
{ |
|
"epoch": 2.34, |
|
"learning_rate": 9.784620593657125e-06, |
|
"loss": 6.3281, |
|
"step": 124500 |
|
}, |
|
{ |
|
"epoch": 2.35, |
|
"learning_rate": 9.783680072232045e-06, |
|
"loss": 6.3264, |
|
"step": 125000 |
|
}, |
|
{ |
|
"epoch": 2.36, |
|
"learning_rate": 9.782739550806967e-06, |
|
"loss": 6.327, |
|
"step": 125500 |
|
}, |
|
{ |
|
"epoch": 2.37, |
|
"learning_rate": 9.78179902938189e-06, |
|
"loss": 6.3214, |
|
"step": 126000 |
|
}, |
|
{ |
|
"epoch": 2.38, |
|
"learning_rate": 9.780858507956812e-06, |
|
"loss": 6.3243, |
|
"step": 126500 |
|
}, |
|
{ |
|
"epoch": 2.38, |
|
"learning_rate": 9.779917986531734e-06, |
|
"loss": 6.322, |
|
"step": 127000 |
|
}, |
|
{ |
|
"epoch": 2.39, |
|
"learning_rate": 9.778977465106656e-06, |
|
"loss": 6.3192, |
|
"step": 127500 |
|
}, |
|
{ |
|
"epoch": 2.4, |
|
"learning_rate": 9.778036943681579e-06, |
|
"loss": 6.3238, |
|
"step": 128000 |
|
}, |
|
{ |
|
"epoch": 2.41, |
|
"learning_rate": 9.777096422256501e-06, |
|
"loss": 6.3184, |
|
"step": 128500 |
|
}, |
|
{ |
|
"epoch": 2.42, |
|
"learning_rate": 9.776155900831421e-06, |
|
"loss": 6.3198, |
|
"step": 129000 |
|
}, |
|
{ |
|
"epoch": 2.43, |
|
"learning_rate": 9.775215379406344e-06, |
|
"loss": 6.3166, |
|
"step": 129500 |
|
}, |
|
{ |
|
"epoch": 2.44, |
|
"learning_rate": 9.774274857981266e-06, |
|
"loss": 6.3161, |
|
"step": 130000 |
|
}, |
|
{ |
|
"epoch": 2.45, |
|
"learning_rate": 9.773334336556187e-06, |
|
"loss": 6.3125, |
|
"step": 130500 |
|
}, |
|
{ |
|
"epoch": 2.46, |
|
"learning_rate": 9.772393815131109e-06, |
|
"loss": 6.3153, |
|
"step": 131000 |
|
}, |
|
{ |
|
"epoch": 2.47, |
|
"learning_rate": 9.771453293706031e-06, |
|
"loss": 6.3137, |
|
"step": 131500 |
|
}, |
|
{ |
|
"epoch": 2.48, |
|
"learning_rate": 9.770512772280953e-06, |
|
"loss": 6.3155, |
|
"step": 132000 |
|
}, |
|
{ |
|
"epoch": 2.49, |
|
"learning_rate": 9.769572250855875e-06, |
|
"loss": 6.3127, |
|
"step": 132500 |
|
}, |
|
{ |
|
"epoch": 2.5, |
|
"learning_rate": 9.768631729430798e-06, |
|
"loss": 6.3152, |
|
"step": 133000 |
|
}, |
|
{ |
|
"epoch": 2.51, |
|
"learning_rate": 9.76769120800572e-06, |
|
"loss": 6.3098, |
|
"step": 133500 |
|
}, |
|
{ |
|
"epoch": 2.52, |
|
"learning_rate": 9.766750686580642e-06, |
|
"loss": 6.3083, |
|
"step": 134000 |
|
}, |
|
{ |
|
"epoch": 2.53, |
|
"learning_rate": 9.765810165155563e-06, |
|
"loss": 6.3084, |
|
"step": 134500 |
|
}, |
|
{ |
|
"epoch": 2.53, |
|
"learning_rate": 9.764869643730485e-06, |
|
"loss": 6.311, |
|
"step": 135000 |
|
}, |
|
{ |
|
"epoch": 2.54, |
|
"learning_rate": 9.763929122305407e-06, |
|
"loss": 6.308, |
|
"step": 135500 |
|
}, |
|
{ |
|
"epoch": 2.55, |
|
"learning_rate": 9.76298860088033e-06, |
|
"loss": 6.3076, |
|
"step": 136000 |
|
}, |
|
{ |
|
"epoch": 2.56, |
|
"learning_rate": 9.76204807945525e-06, |
|
"loss": 6.3032, |
|
"step": 136500 |
|
}, |
|
{ |
|
"epoch": 2.57, |
|
"learning_rate": 9.761107558030172e-06, |
|
"loss": 6.3094, |
|
"step": 137000 |
|
}, |
|
{ |
|
"epoch": 2.58, |
|
"learning_rate": 9.760167036605095e-06, |
|
"loss": 6.3002, |
|
"step": 137500 |
|
}, |
|
{ |
|
"epoch": 2.59, |
|
"learning_rate": 9.759226515180017e-06, |
|
"loss": 6.3038, |
|
"step": 138000 |
|
}, |
|
{ |
|
"epoch": 2.6, |
|
"learning_rate": 9.758285993754939e-06, |
|
"loss": 6.3021, |
|
"step": 138500 |
|
}, |
|
{ |
|
"epoch": 2.61, |
|
"learning_rate": 9.75734547232986e-06, |
|
"loss": 6.3039, |
|
"step": 139000 |
|
}, |
|
{ |
|
"epoch": 2.62, |
|
"learning_rate": 9.756404950904782e-06, |
|
"loss": 6.3041, |
|
"step": 139500 |
|
}, |
|
{ |
|
"epoch": 2.63, |
|
"learning_rate": 9.755464429479704e-06, |
|
"loss": 6.3053, |
|
"step": 140000 |
|
}, |
|
{ |
|
"epoch": 2.64, |
|
"learning_rate": 9.754523908054626e-06, |
|
"loss": 6.2985, |
|
"step": 140500 |
|
}, |
|
{ |
|
"epoch": 2.65, |
|
"learning_rate": 9.753583386629549e-06, |
|
"loss": 6.2997, |
|
"step": 141000 |
|
}, |
|
{ |
|
"epoch": 2.66, |
|
"learning_rate": 9.75264286520447e-06, |
|
"loss": 6.2939, |
|
"step": 141500 |
|
}, |
|
{ |
|
"epoch": 2.67, |
|
"learning_rate": 9.751702343779393e-06, |
|
"loss": 6.2984, |
|
"step": 142000 |
|
}, |
|
{ |
|
"epoch": 2.68, |
|
"learning_rate": 9.750761822354314e-06, |
|
"loss": 6.2987, |
|
"step": 142500 |
|
}, |
|
{ |
|
"epoch": 2.68, |
|
"learning_rate": 9.749821300929236e-06, |
|
"loss": 6.3006, |
|
"step": 143000 |
|
}, |
|
{ |
|
"epoch": 2.69, |
|
"learning_rate": 9.748880779504158e-06, |
|
"loss": 6.2966, |
|
"step": 143500 |
|
}, |
|
{ |
|
"epoch": 2.7, |
|
"learning_rate": 9.747940258079079e-06, |
|
"loss": 6.2977, |
|
"step": 144000 |
|
}, |
|
{ |
|
"epoch": 2.71, |
|
"learning_rate": 9.746999736654e-06, |
|
"loss": 6.2947, |
|
"step": 144500 |
|
}, |
|
{ |
|
"epoch": 2.72, |
|
"learning_rate": 9.746059215228923e-06, |
|
"loss": 6.2979, |
|
"step": 145000 |
|
}, |
|
{ |
|
"epoch": 2.73, |
|
"learning_rate": 9.745118693803845e-06, |
|
"loss": 6.293, |
|
"step": 145500 |
|
}, |
|
{ |
|
"epoch": 2.74, |
|
"learning_rate": 9.744178172378768e-06, |
|
"loss": 6.2913, |
|
"step": 146000 |
|
}, |
|
{ |
|
"epoch": 2.75, |
|
"learning_rate": 9.74323765095369e-06, |
|
"loss": 6.2961, |
|
"step": 146500 |
|
}, |
|
{ |
|
"epoch": 2.76, |
|
"learning_rate": 9.742297129528612e-06, |
|
"loss": 6.2927, |
|
"step": 147000 |
|
}, |
|
{ |
|
"epoch": 2.77, |
|
"learning_rate": 9.741356608103534e-06, |
|
"loss": 6.2878, |
|
"step": 147500 |
|
}, |
|
{ |
|
"epoch": 2.78, |
|
"learning_rate": 9.740416086678457e-06, |
|
"loss": 6.2965, |
|
"step": 148000 |
|
}, |
|
{ |
|
"epoch": 2.79, |
|
"learning_rate": 9.739475565253377e-06, |
|
"loss": 6.2902, |
|
"step": 148500 |
|
}, |
|
{ |
|
"epoch": 2.8, |
|
"learning_rate": 9.7385350438283e-06, |
|
"loss": 6.2892, |
|
"step": 149000 |
|
}, |
|
{ |
|
"epoch": 2.81, |
|
"learning_rate": 9.737594522403222e-06, |
|
"loss": 6.2873, |
|
"step": 149500 |
|
}, |
|
{ |
|
"epoch": 2.82, |
|
"learning_rate": 9.736654000978144e-06, |
|
"loss": 6.2881, |
|
"step": 150000 |
|
}, |
|
{ |
|
"epoch": 2.83, |
|
"learning_rate": 9.735713479553064e-06, |
|
"loss": 6.2923, |
|
"step": 150500 |
|
}, |
|
{ |
|
"epoch": 2.84, |
|
"learning_rate": 9.734772958127987e-06, |
|
"loss": 6.2906, |
|
"step": 151000 |
|
}, |
|
{ |
|
"epoch": 2.84, |
|
"learning_rate": 9.733832436702909e-06, |
|
"loss": 6.2881, |
|
"step": 151500 |
|
}, |
|
{ |
|
"epoch": 2.85, |
|
"learning_rate": 9.732891915277831e-06, |
|
"loss": 6.282, |
|
"step": 152000 |
|
}, |
|
{ |
|
"epoch": 2.86, |
|
"learning_rate": 9.731951393852753e-06, |
|
"loss": 6.2854, |
|
"step": 152500 |
|
}, |
|
{ |
|
"epoch": 2.87, |
|
"learning_rate": 9.731010872427674e-06, |
|
"loss": 6.2808, |
|
"step": 153000 |
|
}, |
|
{ |
|
"epoch": 2.88, |
|
"learning_rate": 9.730070351002596e-06, |
|
"loss": 6.2815, |
|
"step": 153500 |
|
}, |
|
{ |
|
"epoch": 2.89, |
|
"learning_rate": 9.729129829577518e-06, |
|
"loss": 6.2824, |
|
"step": 154000 |
|
}, |
|
{ |
|
"epoch": 2.9, |
|
"learning_rate": 9.72818930815244e-06, |
|
"loss": 6.2823, |
|
"step": 154500 |
|
}, |
|
{ |
|
"epoch": 2.91, |
|
"learning_rate": 9.727248786727363e-06, |
|
"loss": 6.2788, |
|
"step": 155000 |
|
}, |
|
{ |
|
"epoch": 2.92, |
|
"learning_rate": 9.726308265302285e-06, |
|
"loss": 6.279, |
|
"step": 155500 |
|
}, |
|
{ |
|
"epoch": 2.93, |
|
"learning_rate": 9.725367743877207e-06, |
|
"loss": 6.2841, |
|
"step": 156000 |
|
}, |
|
{ |
|
"epoch": 2.94, |
|
"learning_rate": 9.724427222452128e-06, |
|
"loss": 6.2804, |
|
"step": 156500 |
|
}, |
|
{ |
|
"epoch": 2.95, |
|
"learning_rate": 9.72348670102705e-06, |
|
"loss": 6.2806, |
|
"step": 157000 |
|
}, |
|
{ |
|
"epoch": 2.96, |
|
"learning_rate": 9.722546179601972e-06, |
|
"loss": 6.2798, |
|
"step": 157500 |
|
}, |
|
{ |
|
"epoch": 2.97, |
|
"learning_rate": 9.721605658176893e-06, |
|
"loss": 6.2818, |
|
"step": 158000 |
|
}, |
|
{ |
|
"epoch": 2.98, |
|
"learning_rate": 9.720665136751815e-06, |
|
"loss": 6.2802, |
|
"step": 158500 |
|
}, |
|
{ |
|
"epoch": 2.99, |
|
"learning_rate": 9.719724615326737e-06, |
|
"loss": 6.2785, |
|
"step": 159000 |
|
}, |
|
{ |
|
"epoch": 2.99, |
|
"learning_rate": 9.71878409390166e-06, |
|
"loss": 6.2756, |
|
"step": 159500 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"learning_rate": 9.717843572476582e-06, |
|
"loss": 6.2767, |
|
"step": 160000 |
|
}, |
|
{ |
|
"epoch": 3.01, |
|
"learning_rate": 9.716903051051504e-06, |
|
"loss": 6.277, |
|
"step": 160500 |
|
}, |
|
{ |
|
"epoch": 3.02, |
|
"learning_rate": 9.715962529626426e-06, |
|
"loss": 6.2768, |
|
"step": 161000 |
|
}, |
|
{ |
|
"epoch": 3.03, |
|
"learning_rate": 9.715022008201349e-06, |
|
"loss": 6.2788, |
|
"step": 161500 |
|
}, |
|
{ |
|
"epoch": 3.04, |
|
"learning_rate": 9.71408148677627e-06, |
|
"loss": 6.2786, |
|
"step": 162000 |
|
}, |
|
{ |
|
"epoch": 3.05, |
|
"learning_rate": 9.713140965351191e-06, |
|
"loss": 6.2729, |
|
"step": 162500 |
|
}, |
|
{ |
|
"epoch": 3.06, |
|
"learning_rate": 9.712200443926114e-06, |
|
"loss": 6.2753, |
|
"step": 163000 |
|
}, |
|
{ |
|
"epoch": 3.07, |
|
"learning_rate": 9.711259922501036e-06, |
|
"loss": 6.2783, |
|
"step": 163500 |
|
}, |
|
{ |
|
"epoch": 3.08, |
|
"learning_rate": 9.710319401075956e-06, |
|
"loss": 6.2725, |
|
"step": 164000 |
|
}, |
|
{ |
|
"epoch": 3.09, |
|
"learning_rate": 9.709378879650879e-06, |
|
"loss": 6.2676, |
|
"step": 164500 |
|
}, |
|
{ |
|
"epoch": 3.1, |
|
"learning_rate": 9.708438358225801e-06, |
|
"loss": 6.2762, |
|
"step": 165000 |
|
}, |
|
{ |
|
"epoch": 3.11, |
|
"learning_rate": 9.707497836800723e-06, |
|
"loss": 6.2684, |
|
"step": 165500 |
|
}, |
|
{ |
|
"epoch": 3.12, |
|
"learning_rate": 9.706557315375645e-06, |
|
"loss": 6.2777, |
|
"step": 166000 |
|
}, |
|
{ |
|
"epoch": 3.13, |
|
"learning_rate": 9.705616793950568e-06, |
|
"loss": 6.2714, |
|
"step": 166500 |
|
}, |
|
{ |
|
"epoch": 3.14, |
|
"learning_rate": 9.704676272525488e-06, |
|
"loss": 6.2692, |
|
"step": 167000 |
|
}, |
|
{ |
|
"epoch": 3.14, |
|
"learning_rate": 9.70373575110041e-06, |
|
"loss": 6.2673, |
|
"step": 167500 |
|
}, |
|
{ |
|
"epoch": 3.15, |
|
"learning_rate": 9.702795229675333e-06, |
|
"loss": 6.2678, |
|
"step": 168000 |
|
}, |
|
{ |
|
"epoch": 3.16, |
|
"learning_rate": 9.701854708250255e-06, |
|
"loss": 6.2732, |
|
"step": 168500 |
|
}, |
|
{ |
|
"epoch": 3.17, |
|
"learning_rate": 9.700914186825177e-06, |
|
"loss": 6.2658, |
|
"step": 169000 |
|
}, |
|
{ |
|
"epoch": 3.18, |
|
"learning_rate": 9.6999736654001e-06, |
|
"loss": 6.2724, |
|
"step": 169500 |
|
}, |
|
{ |
|
"epoch": 3.19, |
|
"learning_rate": 9.69903314397502e-06, |
|
"loss": 6.2711, |
|
"step": 170000 |
|
}, |
|
{ |
|
"epoch": 3.2, |
|
"learning_rate": 9.698092622549942e-06, |
|
"loss": 6.2709, |
|
"step": 170500 |
|
}, |
|
{ |
|
"epoch": 3.21, |
|
"learning_rate": 9.697152101124864e-06, |
|
"loss": 6.2694, |
|
"step": 171000 |
|
}, |
|
{ |
|
"epoch": 3.22, |
|
"learning_rate": 9.696211579699787e-06, |
|
"loss": 6.2659, |
|
"step": 171500 |
|
}, |
|
{ |
|
"epoch": 3.23, |
|
"learning_rate": 9.695271058274707e-06, |
|
"loss": 6.2597, |
|
"step": 172000 |
|
}, |
|
{ |
|
"epoch": 3.24, |
|
"learning_rate": 9.69433053684963e-06, |
|
"loss": 6.2664, |
|
"step": 172500 |
|
}, |
|
{ |
|
"epoch": 3.25, |
|
"learning_rate": 9.693390015424552e-06, |
|
"loss": 6.2696, |
|
"step": 173000 |
|
}, |
|
{ |
|
"epoch": 3.26, |
|
"learning_rate": 9.692449493999474e-06, |
|
"loss": 6.2655, |
|
"step": 173500 |
|
}, |
|
{ |
|
"epoch": 3.27, |
|
"learning_rate": 9.691508972574396e-06, |
|
"loss": 6.2656, |
|
"step": 174000 |
|
}, |
|
{ |
|
"epoch": 3.28, |
|
"learning_rate": 9.690568451149318e-06, |
|
"loss": 6.2596, |
|
"step": 174500 |
|
}, |
|
{ |
|
"epoch": 3.29, |
|
"learning_rate": 9.68962792972424e-06, |
|
"loss": 6.2598, |
|
"step": 175000 |
|
}, |
|
{ |
|
"epoch": 3.3, |
|
"learning_rate": 9.688687408299163e-06, |
|
"loss": 6.2655, |
|
"step": 175500 |
|
}, |
|
{ |
|
"epoch": 3.3, |
|
"learning_rate": 9.687746886874085e-06, |
|
"loss": 6.2618, |
|
"step": 176000 |
|
}, |
|
{ |
|
"epoch": 3.31, |
|
"learning_rate": 9.686806365449006e-06, |
|
"loss": 6.2601, |
|
"step": 176500 |
|
}, |
|
{ |
|
"epoch": 3.32, |
|
"learning_rate": 9.685865844023928e-06, |
|
"loss": 6.2605, |
|
"step": 177000 |
|
}, |
|
{ |
|
"epoch": 3.33, |
|
"learning_rate": 9.68492532259885e-06, |
|
"loss": 6.2608, |
|
"step": 177500 |
|
}, |
|
{ |
|
"epoch": 3.34, |
|
"learning_rate": 9.68398480117377e-06, |
|
"loss": 6.2594, |
|
"step": 178000 |
|
}, |
|
{ |
|
"epoch": 3.35, |
|
"learning_rate": 9.683044279748693e-06, |
|
"loss": 6.2587, |
|
"step": 178500 |
|
}, |
|
{ |
|
"epoch": 3.36, |
|
"learning_rate": 9.682103758323615e-06, |
|
"loss": 6.2589, |
|
"step": 179000 |
|
}, |
|
{ |
|
"epoch": 3.37, |
|
"learning_rate": 9.681163236898537e-06, |
|
"loss": 6.2571, |
|
"step": 179500 |
|
}, |
|
{ |
|
"epoch": 3.38, |
|
"learning_rate": 9.68022271547346e-06, |
|
"loss": 6.2576, |
|
"step": 180000 |
|
}, |
|
{ |
|
"epoch": 3.39, |
|
"learning_rate": 9.679282194048382e-06, |
|
"loss": 6.257, |
|
"step": 180500 |
|
}, |
|
{ |
|
"epoch": 3.4, |
|
"learning_rate": 9.678341672623302e-06, |
|
"loss": 6.2566, |
|
"step": 181000 |
|
}, |
|
{ |
|
"epoch": 3.41, |
|
"learning_rate": 9.677401151198225e-06, |
|
"loss": 6.261, |
|
"step": 181500 |
|
}, |
|
{ |
|
"epoch": 3.42, |
|
"learning_rate": 9.676460629773147e-06, |
|
"loss": 6.2549, |
|
"step": 182000 |
|
}, |
|
{ |
|
"epoch": 3.43, |
|
"learning_rate": 9.67552010834807e-06, |
|
"loss": 6.2556, |
|
"step": 182500 |
|
}, |
|
{ |
|
"epoch": 3.44, |
|
"learning_rate": 9.674579586922991e-06, |
|
"loss": 6.258, |
|
"step": 183000 |
|
}, |
|
{ |
|
"epoch": 3.45, |
|
"learning_rate": 9.673639065497914e-06, |
|
"loss": 6.2557, |
|
"step": 183500 |
|
}, |
|
{ |
|
"epoch": 3.45, |
|
"learning_rate": 9.672698544072834e-06, |
|
"loss": 6.257, |
|
"step": 184000 |
|
}, |
|
{ |
|
"epoch": 3.46, |
|
"learning_rate": 9.671758022647756e-06, |
|
"loss": 6.2509, |
|
"step": 184500 |
|
}, |
|
{ |
|
"epoch": 3.47, |
|
"learning_rate": 9.670817501222679e-06, |
|
"loss": 6.2549, |
|
"step": 185000 |
|
}, |
|
{ |
|
"epoch": 3.48, |
|
"learning_rate": 9.669876979797601e-06, |
|
"loss": 6.2534, |
|
"step": 185500 |
|
}, |
|
{ |
|
"epoch": 3.49, |
|
"learning_rate": 9.668936458372521e-06, |
|
"loss": 6.2543, |
|
"step": 186000 |
|
}, |
|
{ |
|
"epoch": 3.5, |
|
"learning_rate": 9.667995936947444e-06, |
|
"loss": 6.2518, |
|
"step": 186500 |
|
}, |
|
{ |
|
"epoch": 3.51, |
|
"learning_rate": 9.667055415522366e-06, |
|
"loss": 6.2518, |
|
"step": 187000 |
|
}, |
|
{ |
|
"epoch": 3.52, |
|
"learning_rate": 9.666114894097288e-06, |
|
"loss": 6.2544, |
|
"step": 187500 |
|
}, |
|
{ |
|
"epoch": 3.53, |
|
"learning_rate": 9.66517437267221e-06, |
|
"loss": 6.2503, |
|
"step": 188000 |
|
}, |
|
{ |
|
"epoch": 3.54, |
|
"learning_rate": 9.664233851247133e-06, |
|
"loss": 6.2498, |
|
"step": 188500 |
|
}, |
|
{ |
|
"epoch": 3.55, |
|
"learning_rate": 9.663293329822055e-06, |
|
"loss": 6.2513, |
|
"step": 189000 |
|
}, |
|
{ |
|
"epoch": 3.56, |
|
"learning_rate": 9.662352808396977e-06, |
|
"loss": 6.2496, |
|
"step": 189500 |
|
}, |
|
{ |
|
"epoch": 3.57, |
|
"learning_rate": 9.661412286971898e-06, |
|
"loss": 6.2523, |
|
"step": 190000 |
|
}, |
|
{ |
|
"epoch": 3.58, |
|
"learning_rate": 9.66047176554682e-06, |
|
"loss": 6.2504, |
|
"step": 190500 |
|
}, |
|
{ |
|
"epoch": 3.59, |
|
"learning_rate": 9.659531244121742e-06, |
|
"loss": 6.2525, |
|
"step": 191000 |
|
}, |
|
{ |
|
"epoch": 3.6, |
|
"learning_rate": 9.658590722696663e-06, |
|
"loss": 6.2454, |
|
"step": 191500 |
|
}, |
|
{ |
|
"epoch": 3.6, |
|
"learning_rate": 9.657650201271585e-06, |
|
"loss": 6.253, |
|
"step": 192000 |
|
}, |
|
{ |
|
"epoch": 3.61, |
|
"learning_rate": 9.656709679846507e-06, |
|
"loss": 6.2448, |
|
"step": 192500 |
|
}, |
|
{ |
|
"epoch": 3.62, |
|
"learning_rate": 9.65576915842143e-06, |
|
"loss": 6.2475, |
|
"step": 193000 |
|
}, |
|
{ |
|
"epoch": 3.63, |
|
"learning_rate": 9.654828636996352e-06, |
|
"loss": 6.25, |
|
"step": 193500 |
|
}, |
|
{ |
|
"epoch": 3.64, |
|
"learning_rate": 9.653888115571274e-06, |
|
"loss": 6.2475, |
|
"step": 194000 |
|
}, |
|
{ |
|
"epoch": 3.65, |
|
"learning_rate": 9.652947594146196e-06, |
|
"loss": 6.2479, |
|
"step": 194500 |
|
}, |
|
{ |
|
"epoch": 3.66, |
|
"learning_rate": 9.652007072721118e-06, |
|
"loss": 6.2471, |
|
"step": 195000 |
|
}, |
|
{ |
|
"epoch": 3.67, |
|
"learning_rate": 9.651066551296039e-06, |
|
"loss": 6.2467, |
|
"step": 195500 |
|
}, |
|
{ |
|
"epoch": 3.68, |
|
"learning_rate": 9.650126029870961e-06, |
|
"loss": 6.2447, |
|
"step": 196000 |
|
}, |
|
{ |
|
"epoch": 3.69, |
|
"learning_rate": 9.649185508445883e-06, |
|
"loss": 6.2441, |
|
"step": 196500 |
|
}, |
|
{ |
|
"epoch": 3.7, |
|
"learning_rate": 9.648244987020806e-06, |
|
"loss": 6.2414, |
|
"step": 197000 |
|
}, |
|
{ |
|
"epoch": 3.71, |
|
"learning_rate": 9.647304465595726e-06, |
|
"loss": 6.242, |
|
"step": 197500 |
|
}, |
|
{ |
|
"epoch": 3.72, |
|
"learning_rate": 9.646363944170649e-06, |
|
"loss": 6.2441, |
|
"step": 198000 |
|
}, |
|
{ |
|
"epoch": 3.73, |
|
"learning_rate": 9.64542342274557e-06, |
|
"loss": 6.2481, |
|
"step": 198500 |
|
}, |
|
{ |
|
"epoch": 3.74, |
|
"learning_rate": 9.644482901320493e-06, |
|
"loss": 6.2428, |
|
"step": 199000 |
|
}, |
|
{ |
|
"epoch": 3.75, |
|
"learning_rate": 9.643542379895415e-06, |
|
"loss": 6.2455, |
|
"step": 199500 |
|
}, |
|
{ |
|
"epoch": 3.76, |
|
"learning_rate": 9.642601858470336e-06, |
|
"loss": 6.2413, |
|
"step": 200000 |
|
}, |
|
{ |
|
"epoch": 3.76, |
|
"learning_rate": 9.641661337045258e-06, |
|
"loss": 6.2423, |
|
"step": 200500 |
|
}, |
|
{ |
|
"epoch": 3.77, |
|
"learning_rate": 9.64072081562018e-06, |
|
"loss": 6.2425, |
|
"step": 201000 |
|
}, |
|
{ |
|
"epoch": 3.78, |
|
"learning_rate": 9.639780294195102e-06, |
|
"loss": 6.245, |
|
"step": 201500 |
|
}, |
|
{ |
|
"epoch": 3.79, |
|
"learning_rate": 9.638839772770025e-06, |
|
"loss": 6.2426, |
|
"step": 202000 |
|
}, |
|
{ |
|
"epoch": 3.8, |
|
"learning_rate": 9.637899251344947e-06, |
|
"loss": 6.2372, |
|
"step": 202500 |
|
}, |
|
{ |
|
"epoch": 3.81, |
|
"learning_rate": 9.63695872991987e-06, |
|
"loss": 6.2397, |
|
"step": 203000 |
|
}, |
|
{ |
|
"epoch": 3.82, |
|
"learning_rate": 9.636018208494791e-06, |
|
"loss": 6.243, |
|
"step": 203500 |
|
}, |
|
{ |
|
"epoch": 3.83, |
|
"learning_rate": 9.635077687069712e-06, |
|
"loss": 6.2341, |
|
"step": 204000 |
|
}, |
|
{ |
|
"epoch": 3.84, |
|
"learning_rate": 9.634137165644634e-06, |
|
"loss": 6.2402, |
|
"step": 204500 |
|
}, |
|
{ |
|
"epoch": 3.85, |
|
"learning_rate": 9.633196644219556e-06, |
|
"loss": 6.2448, |
|
"step": 205000 |
|
}, |
|
{ |
|
"epoch": 3.86, |
|
"learning_rate": 9.632256122794477e-06, |
|
"loss": 6.2401, |
|
"step": 205500 |
|
}, |
|
{ |
|
"epoch": 3.87, |
|
"learning_rate": 9.6313156013694e-06, |
|
"loss": 6.2431, |
|
"step": 206000 |
|
}, |
|
{ |
|
"epoch": 3.88, |
|
"learning_rate": 9.630375079944322e-06, |
|
"loss": 6.2375, |
|
"step": 206500 |
|
}, |
|
{ |
|
"epoch": 3.89, |
|
"learning_rate": 9.629434558519244e-06, |
|
"loss": 6.2388, |
|
"step": 207000 |
|
}, |
|
{ |
|
"epoch": 3.9, |
|
"learning_rate": 9.628494037094166e-06, |
|
"loss": 6.2371, |
|
"step": 207500 |
|
}, |
|
{ |
|
"epoch": 3.91, |
|
"learning_rate": 9.627553515669088e-06, |
|
"loss": 6.2322, |
|
"step": 208000 |
|
}, |
|
{ |
|
"epoch": 3.91, |
|
"learning_rate": 9.62661299424401e-06, |
|
"loss": 6.238, |
|
"step": 208500 |
|
}, |
|
{ |
|
"epoch": 3.92, |
|
"learning_rate": 9.625672472818933e-06, |
|
"loss": 6.2322, |
|
"step": 209000 |
|
}, |
|
{ |
|
"epoch": 3.93, |
|
"learning_rate": 9.624731951393853e-06, |
|
"loss": 6.2393, |
|
"step": 209500 |
|
}, |
|
{ |
|
"epoch": 3.94, |
|
"learning_rate": 9.623791429968776e-06, |
|
"loss": 6.2377, |
|
"step": 210000 |
|
}, |
|
{ |
|
"epoch": 3.95, |
|
"learning_rate": 9.622850908543698e-06, |
|
"loss": 6.2374, |
|
"step": 210500 |
|
}, |
|
{ |
|
"epoch": 3.96, |
|
"learning_rate": 9.62191038711862e-06, |
|
"loss": 6.2347, |
|
"step": 211000 |
|
}, |
|
{ |
|
"epoch": 3.97, |
|
"learning_rate": 9.62096986569354e-06, |
|
"loss": 6.2391, |
|
"step": 211500 |
|
}, |
|
{ |
|
"epoch": 3.98, |
|
"learning_rate": 9.620029344268463e-06, |
|
"loss": 6.2377, |
|
"step": 212000 |
|
}, |
|
{ |
|
"epoch": 3.99, |
|
"learning_rate": 9.619088822843385e-06, |
|
"loss": 6.2348, |
|
"step": 212500 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"learning_rate": 9.618148301418307e-06, |
|
"loss": 6.2322, |
|
"step": 213000 |
|
}, |
|
{ |
|
"epoch": 4.01, |
|
"learning_rate": 9.61720777999323e-06, |
|
"loss": 6.2337, |
|
"step": 213500 |
|
}, |
|
{ |
|
"epoch": 4.02, |
|
"learning_rate": 9.61626725856815e-06, |
|
"loss": 6.235, |
|
"step": 214000 |
|
}, |
|
{ |
|
"epoch": 4.03, |
|
"learning_rate": 9.615326737143072e-06, |
|
"loss": 6.2321, |
|
"step": 214500 |
|
}, |
|
{ |
|
"epoch": 4.04, |
|
"learning_rate": 9.614386215717995e-06, |
|
"loss": 6.2383, |
|
"step": 215000 |
|
}, |
|
{ |
|
"epoch": 4.05, |
|
"learning_rate": 9.613445694292917e-06, |
|
"loss": 6.2296, |
|
"step": 215500 |
|
}, |
|
{ |
|
"epoch": 4.06, |
|
"learning_rate": 9.612505172867839e-06, |
|
"loss": 6.233, |
|
"step": 216000 |
|
}, |
|
{ |
|
"epoch": 4.06, |
|
"learning_rate": 9.611564651442761e-06, |
|
"loss": 6.2303, |
|
"step": 216500 |
|
}, |
|
{ |
|
"epoch": 4.07, |
|
"learning_rate": 9.610624130017684e-06, |
|
"loss": 6.2312, |
|
"step": 217000 |
|
}, |
|
{ |
|
"epoch": 4.08, |
|
"learning_rate": 9.609683608592604e-06, |
|
"loss": 6.2331, |
|
"step": 217500 |
|
}, |
|
{ |
|
"epoch": 4.09, |
|
"learning_rate": 9.608743087167526e-06, |
|
"loss": 6.2283, |
|
"step": 218000 |
|
}, |
|
{ |
|
"epoch": 4.1, |
|
"learning_rate": 9.607802565742449e-06, |
|
"loss": 6.2325, |
|
"step": 218500 |
|
}, |
|
{ |
|
"epoch": 4.11, |
|
"learning_rate": 9.606862044317369e-06, |
|
"loss": 6.2332, |
|
"step": 219000 |
|
}, |
|
{ |
|
"epoch": 4.12, |
|
"learning_rate": 9.605921522892291e-06, |
|
"loss": 6.2285, |
|
"step": 219500 |
|
}, |
|
{ |
|
"epoch": 4.13, |
|
"learning_rate": 9.604981001467214e-06, |
|
"loss": 6.2273, |
|
"step": 220000 |
|
}, |
|
{ |
|
"epoch": 4.14, |
|
"learning_rate": 9.604040480042136e-06, |
|
"loss": 6.2302, |
|
"step": 220500 |
|
}, |
|
{ |
|
"epoch": 4.15, |
|
"learning_rate": 9.603099958617058e-06, |
|
"loss": 6.2298, |
|
"step": 221000 |
|
}, |
|
{ |
|
"epoch": 4.16, |
|
"learning_rate": 9.60215943719198e-06, |
|
"loss": 6.2284, |
|
"step": 221500 |
|
}, |
|
{ |
|
"epoch": 4.17, |
|
"learning_rate": 9.601218915766903e-06, |
|
"loss": 6.226, |
|
"step": 222000 |
|
}, |
|
{ |
|
"epoch": 4.18, |
|
"learning_rate": 9.600278394341825e-06, |
|
"loss": 6.2283, |
|
"step": 222500 |
|
}, |
|
{ |
|
"epoch": 4.19, |
|
"learning_rate": 9.599337872916747e-06, |
|
"loss": 6.2273, |
|
"step": 223000 |
|
}, |
|
{ |
|
"epoch": 4.2, |
|
"learning_rate": 9.598397351491668e-06, |
|
"loss": 6.2253, |
|
"step": 223500 |
|
}, |
|
{ |
|
"epoch": 4.21, |
|
"learning_rate": 9.59745683006659e-06, |
|
"loss": 6.2288, |
|
"step": 224000 |
|
}, |
|
{ |
|
"epoch": 4.22, |
|
"learning_rate": 9.596516308641512e-06, |
|
"loss": 6.2241, |
|
"step": 224500 |
|
}, |
|
{ |
|
"epoch": 4.22, |
|
"learning_rate": 9.595575787216433e-06, |
|
"loss": 6.2246, |
|
"step": 225000 |
|
}, |
|
{ |
|
"epoch": 4.23, |
|
"learning_rate": 9.594635265791355e-06, |
|
"loss": 6.2296, |
|
"step": 225500 |
|
}, |
|
{ |
|
"epoch": 4.24, |
|
"learning_rate": 9.593694744366277e-06, |
|
"loss": 6.2273, |
|
"step": 226000 |
|
}, |
|
{ |
|
"epoch": 4.25, |
|
"learning_rate": 9.5927542229412e-06, |
|
"loss": 6.2261, |
|
"step": 226500 |
|
}, |
|
{ |
|
"epoch": 4.26, |
|
"learning_rate": 9.591813701516122e-06, |
|
"loss": 6.2267, |
|
"step": 227000 |
|
}, |
|
{ |
|
"epoch": 4.27, |
|
"learning_rate": 9.590873180091044e-06, |
|
"loss": 6.2229, |
|
"step": 227500 |
|
}, |
|
{ |
|
"epoch": 4.28, |
|
"learning_rate": 9.589932658665964e-06, |
|
"loss": 6.2232, |
|
"step": 228000 |
|
}, |
|
{ |
|
"epoch": 4.29, |
|
"learning_rate": 9.588992137240887e-06, |
|
"loss": 6.2252, |
|
"step": 228500 |
|
}, |
|
{ |
|
"epoch": 4.3, |
|
"learning_rate": 9.588051615815809e-06, |
|
"loss": 6.2284, |
|
"step": 229000 |
|
}, |
|
{ |
|
"epoch": 4.31, |
|
"learning_rate": 9.587111094390731e-06, |
|
"loss": 6.2219, |
|
"step": 229500 |
|
}, |
|
{ |
|
"epoch": 4.32, |
|
"learning_rate": 9.586170572965653e-06, |
|
"loss": 6.2253, |
|
"step": 230000 |
|
}, |
|
{ |
|
"epoch": 4.33, |
|
"learning_rate": 9.585230051540576e-06, |
|
"loss": 6.2289, |
|
"step": 230500 |
|
}, |
|
{ |
|
"epoch": 4.34, |
|
"learning_rate": 9.584289530115498e-06, |
|
"loss": 6.2187, |
|
"step": 231000 |
|
}, |
|
{ |
|
"epoch": 4.35, |
|
"learning_rate": 9.583349008690418e-06, |
|
"loss": 6.2268, |
|
"step": 231500 |
|
}, |
|
{ |
|
"epoch": 4.36, |
|
"learning_rate": 9.58240848726534e-06, |
|
"loss": 6.2255, |
|
"step": 232000 |
|
}, |
|
{ |
|
"epoch": 4.37, |
|
"learning_rate": 9.581467965840263e-06, |
|
"loss": 6.2229, |
|
"step": 232500 |
|
}, |
|
{ |
|
"epoch": 4.37, |
|
"learning_rate": 9.580527444415183e-06, |
|
"loss": 6.2237, |
|
"step": 233000 |
|
}, |
|
{ |
|
"epoch": 4.38, |
|
"learning_rate": 9.579586922990106e-06, |
|
"loss": 6.2195, |
|
"step": 233500 |
|
}, |
|
{ |
|
"epoch": 4.39, |
|
"learning_rate": 9.578646401565028e-06, |
|
"loss": 6.2179, |
|
"step": 234000 |
|
}, |
|
{ |
|
"epoch": 4.4, |
|
"learning_rate": 9.57770588013995e-06, |
|
"loss": 6.2245, |
|
"step": 234500 |
|
}, |
|
{ |
|
"epoch": 4.41, |
|
"learning_rate": 9.576765358714872e-06, |
|
"loss": 6.2255, |
|
"step": 235000 |
|
}, |
|
{ |
|
"epoch": 4.42, |
|
"learning_rate": 9.575824837289795e-06, |
|
"loss": 6.2243, |
|
"step": 235500 |
|
}, |
|
{ |
|
"epoch": 4.43, |
|
"learning_rate": 9.574884315864717e-06, |
|
"loss": 6.2165, |
|
"step": 236000 |
|
}, |
|
{ |
|
"epoch": 4.44, |
|
"learning_rate": 9.573943794439639e-06, |
|
"loss": 6.2206, |
|
"step": 236500 |
|
}, |
|
{ |
|
"epoch": 4.45, |
|
"learning_rate": 9.573003273014561e-06, |
|
"loss": 6.2231, |
|
"step": 237000 |
|
}, |
|
{ |
|
"epoch": 4.46, |
|
"learning_rate": 9.572062751589482e-06, |
|
"loss": 6.2192, |
|
"step": 237500 |
|
}, |
|
{ |
|
"epoch": 4.47, |
|
"learning_rate": 9.571122230164404e-06, |
|
"loss": 6.2207, |
|
"step": 238000 |
|
}, |
|
{ |
|
"epoch": 4.48, |
|
"learning_rate": 9.570181708739326e-06, |
|
"loss": 6.2221, |
|
"step": 238500 |
|
}, |
|
{ |
|
"epoch": 4.49, |
|
"learning_rate": 9.569241187314247e-06, |
|
"loss": 6.222, |
|
"step": 239000 |
|
}, |
|
{ |
|
"epoch": 4.5, |
|
"learning_rate": 9.568300665889169e-06, |
|
"loss": 6.2197, |
|
"step": 239500 |
|
}, |
|
{ |
|
"epoch": 4.51, |
|
"learning_rate": 9.567360144464091e-06, |
|
"loss": 6.2231, |
|
"step": 240000 |
|
}, |
|
{ |
|
"epoch": 4.52, |
|
"learning_rate": 9.566419623039014e-06, |
|
"loss": 6.2141, |
|
"step": 240500 |
|
}, |
|
{ |
|
"epoch": 4.52, |
|
"learning_rate": 9.565479101613936e-06, |
|
"loss": 6.2153, |
|
"step": 241000 |
|
}, |
|
{ |
|
"epoch": 4.53, |
|
"learning_rate": 9.564538580188858e-06, |
|
"loss": 6.2171, |
|
"step": 241500 |
|
}, |
|
{ |
|
"epoch": 4.54, |
|
"learning_rate": 9.563598058763779e-06, |
|
"loss": 6.2191, |
|
"step": 242000 |
|
}, |
|
{ |
|
"epoch": 4.55, |
|
"learning_rate": 9.562657537338701e-06, |
|
"loss": 6.2174, |
|
"step": 242500 |
|
}, |
|
{ |
|
"epoch": 4.56, |
|
"learning_rate": 9.561717015913623e-06, |
|
"loss": 6.2108, |
|
"step": 243000 |
|
}, |
|
{ |
|
"epoch": 4.57, |
|
"learning_rate": 9.560776494488545e-06, |
|
"loss": 6.2193, |
|
"step": 243500 |
|
}, |
|
{ |
|
"epoch": 4.58, |
|
"learning_rate": 9.559835973063468e-06, |
|
"loss": 6.2125, |
|
"step": 244000 |
|
}, |
|
{ |
|
"epoch": 4.59, |
|
"learning_rate": 9.55889545163839e-06, |
|
"loss": 6.2175, |
|
"step": 244500 |
|
}, |
|
{ |
|
"epoch": 4.6, |
|
"learning_rate": 9.55795493021331e-06, |
|
"loss": 6.2168, |
|
"step": 245000 |
|
}, |
|
{ |
|
"epoch": 4.61, |
|
"learning_rate": 9.557014408788233e-06, |
|
"loss": 6.2183, |
|
"step": 245500 |
|
}, |
|
{ |
|
"epoch": 4.62, |
|
"learning_rate": 9.556073887363155e-06, |
|
"loss": 6.2129, |
|
"step": 246000 |
|
}, |
|
{ |
|
"epoch": 4.63, |
|
"learning_rate": 9.555133365938077e-06, |
|
"loss": 6.2134, |
|
"step": 246500 |
|
}, |
|
{ |
|
"epoch": 4.64, |
|
"learning_rate": 9.554192844512998e-06, |
|
"loss": 6.2176, |
|
"step": 247000 |
|
}, |
|
{ |
|
"epoch": 4.65, |
|
"learning_rate": 9.55325232308792e-06, |
|
"loss": 6.2188, |
|
"step": 247500 |
|
}, |
|
{ |
|
"epoch": 4.66, |
|
"learning_rate": 9.552311801662842e-06, |
|
"loss": 6.2131, |
|
"step": 248000 |
|
}, |
|
{ |
|
"epoch": 4.67, |
|
"learning_rate": 9.551371280237764e-06, |
|
"loss": 6.2125, |
|
"step": 248500 |
|
}, |
|
{ |
|
"epoch": 4.68, |
|
"learning_rate": 9.550430758812687e-06, |
|
"loss": 6.216, |
|
"step": 249000 |
|
}, |
|
{ |
|
"epoch": 4.68, |
|
"learning_rate": 9.549490237387609e-06, |
|
"loss": 6.2123, |
|
"step": 249500 |
|
}, |
|
{ |
|
"epoch": 4.69, |
|
"learning_rate": 9.548549715962531e-06, |
|
"loss": 6.2077, |
|
"step": 250000 |
|
}, |
|
{ |
|
"epoch": 4.7, |
|
"learning_rate": 9.547609194537453e-06, |
|
"loss": 6.2127, |
|
"step": 250500 |
|
}, |
|
{ |
|
"epoch": 4.71, |
|
"learning_rate": 9.546668673112374e-06, |
|
"loss": 6.2136, |
|
"step": 251000 |
|
}, |
|
{ |
|
"epoch": 4.72, |
|
"learning_rate": 9.545728151687296e-06, |
|
"loss": 6.2123, |
|
"step": 251500 |
|
}, |
|
{ |
|
"epoch": 4.73, |
|
"learning_rate": 9.544787630262218e-06, |
|
"loss": 6.2109, |
|
"step": 252000 |
|
}, |
|
{ |
|
"epoch": 4.74, |
|
"learning_rate": 9.54384710883714e-06, |
|
"loss": 6.2139, |
|
"step": 252500 |
|
}, |
|
{ |
|
"epoch": 4.75, |
|
"learning_rate": 9.542906587412061e-06, |
|
"loss": 6.214, |
|
"step": 253000 |
|
}, |
|
{ |
|
"epoch": 4.76, |
|
"learning_rate": 9.541966065986983e-06, |
|
"loss": 6.2109, |
|
"step": 253500 |
|
}, |
|
{ |
|
"epoch": 4.77, |
|
"learning_rate": 9.541025544561906e-06, |
|
"loss": 6.2141, |
|
"step": 254000 |
|
}, |
|
{ |
|
"epoch": 4.78, |
|
"learning_rate": 9.540085023136828e-06, |
|
"loss": 6.2128, |
|
"step": 254500 |
|
}, |
|
{ |
|
"epoch": 4.79, |
|
"learning_rate": 9.53914450171175e-06, |
|
"loss": 6.2107, |
|
"step": 255000 |
|
}, |
|
{ |
|
"epoch": 4.8, |
|
"learning_rate": 9.538203980286672e-06, |
|
"loss": 6.2131, |
|
"step": 255500 |
|
}, |
|
{ |
|
"epoch": 4.81, |
|
"learning_rate": 9.537263458861593e-06, |
|
"loss": 6.2075, |
|
"step": 256000 |
|
}, |
|
{ |
|
"epoch": 4.82, |
|
"learning_rate": 9.536322937436515e-06, |
|
"loss": 6.213, |
|
"step": 256500 |
|
}, |
|
{ |
|
"epoch": 4.83, |
|
"learning_rate": 9.535382416011437e-06, |
|
"loss": 6.2076, |
|
"step": 257000 |
|
}, |
|
{ |
|
"epoch": 4.83, |
|
"learning_rate": 9.53444189458636e-06, |
|
"loss": 6.2117, |
|
"step": 257500 |
|
}, |
|
{ |
|
"epoch": 4.84, |
|
"learning_rate": 9.533501373161282e-06, |
|
"loss": 6.2084, |
|
"step": 258000 |
|
}, |
|
{ |
|
"epoch": 4.85, |
|
"learning_rate": 9.532560851736204e-06, |
|
"loss": 6.2139, |
|
"step": 258500 |
|
}, |
|
{ |
|
"epoch": 4.86, |
|
"learning_rate": 9.531620330311125e-06, |
|
"loss": 6.2048, |
|
"step": 259000 |
|
}, |
|
{ |
|
"epoch": 4.87, |
|
"learning_rate": 9.530679808886047e-06, |
|
"loss": 6.2067, |
|
"step": 259500 |
|
}, |
|
{ |
|
"epoch": 4.88, |
|
"learning_rate": 9.52973928746097e-06, |
|
"loss": 6.2074, |
|
"step": 260000 |
|
}, |
|
{ |
|
"epoch": 4.89, |
|
"learning_rate": 9.528798766035891e-06, |
|
"loss": 6.2015, |
|
"step": 260500 |
|
}, |
|
{ |
|
"epoch": 4.9, |
|
"learning_rate": 9.527858244610812e-06, |
|
"loss": 6.2127, |
|
"step": 261000 |
|
}, |
|
{ |
|
"epoch": 4.91, |
|
"learning_rate": 9.526917723185734e-06, |
|
"loss": 6.2059, |
|
"step": 261500 |
|
}, |
|
{ |
|
"epoch": 4.92, |
|
"learning_rate": 9.525977201760656e-06, |
|
"loss": 6.206, |
|
"step": 262000 |
|
}, |
|
{ |
|
"epoch": 4.93, |
|
"learning_rate": 9.525036680335579e-06, |
|
"loss": 6.2048, |
|
"step": 262500 |
|
}, |
|
{ |
|
"epoch": 4.94, |
|
"learning_rate": 9.524096158910501e-06, |
|
"loss": 6.2089, |
|
"step": 263000 |
|
}, |
|
{ |
|
"epoch": 4.95, |
|
"learning_rate": 9.523155637485423e-06, |
|
"loss": 6.2059, |
|
"step": 263500 |
|
}, |
|
{ |
|
"epoch": 4.96, |
|
"learning_rate": 9.522215116060345e-06, |
|
"loss": 6.2076, |
|
"step": 264000 |
|
}, |
|
{ |
|
"epoch": 4.97, |
|
"learning_rate": 9.521274594635268e-06, |
|
"loss": 6.2056, |
|
"step": 264500 |
|
}, |
|
{ |
|
"epoch": 4.98, |
|
"learning_rate": 9.520334073210188e-06, |
|
"loss": 6.2044, |
|
"step": 265000 |
|
}, |
|
{ |
|
"epoch": 4.98, |
|
"learning_rate": 9.51939355178511e-06, |
|
"loss": 6.2106, |
|
"step": 265500 |
|
}, |
|
{ |
|
"epoch": 4.99, |
|
"learning_rate": 9.518453030360033e-06, |
|
"loss": 6.2077, |
|
"step": 266000 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"learning_rate": 9.517512508934953e-06, |
|
"loss": 6.1984, |
|
"step": 266500 |
|
}, |
|
{ |
|
"epoch": 5.01, |
|
"learning_rate": 9.516571987509876e-06, |
|
"loss": 6.2051, |
|
"step": 267000 |
|
}, |
|
{ |
|
"epoch": 5.02, |
|
"learning_rate": 9.515631466084798e-06, |
|
"loss": 6.2031, |
|
"step": 267500 |
|
}, |
|
{ |
|
"epoch": 5.03, |
|
"learning_rate": 9.51469094465972e-06, |
|
"loss": 6.2033, |
|
"step": 268000 |
|
}, |
|
{ |
|
"epoch": 5.04, |
|
"learning_rate": 9.513750423234642e-06, |
|
"loss": 6.2019, |
|
"step": 268500 |
|
}, |
|
{ |
|
"epoch": 5.05, |
|
"learning_rate": 9.512809901809564e-06, |
|
"loss": 6.2076, |
|
"step": 269000 |
|
}, |
|
{ |
|
"epoch": 5.06, |
|
"learning_rate": 9.511869380384487e-06, |
|
"loss": 6.2033, |
|
"step": 269500 |
|
}, |
|
{ |
|
"epoch": 5.07, |
|
"learning_rate": 9.510928858959409e-06, |
|
"loss": 6.2032, |
|
"step": 270000 |
|
}, |
|
{ |
|
"epoch": 5.08, |
|
"learning_rate": 9.50998833753433e-06, |
|
"loss": 6.2037, |
|
"step": 270500 |
|
}, |
|
{ |
|
"epoch": 5.09, |
|
"learning_rate": 9.509047816109252e-06, |
|
"loss": 6.2017, |
|
"step": 271000 |
|
}, |
|
{ |
|
"epoch": 5.1, |
|
"learning_rate": 9.508107294684174e-06, |
|
"loss": 6.2041, |
|
"step": 271500 |
|
}, |
|
{ |
|
"epoch": 5.11, |
|
"learning_rate": 9.507166773259096e-06, |
|
"loss": 6.2001, |
|
"step": 272000 |
|
}, |
|
{ |
|
"epoch": 5.12, |
|
"learning_rate": 9.506226251834017e-06, |
|
"loss": 6.2079, |
|
"step": 272500 |
|
}, |
|
{ |
|
"epoch": 5.13, |
|
"learning_rate": 9.505285730408939e-06, |
|
"loss": 6.1997, |
|
"step": 273000 |
|
}, |
|
{ |
|
"epoch": 5.13, |
|
"learning_rate": 9.504345208983861e-06, |
|
"loss": 6.2004, |
|
"step": 273500 |
|
}, |
|
{ |
|
"epoch": 5.14, |
|
"learning_rate": 9.503404687558783e-06, |
|
"loss": 6.2013, |
|
"step": 274000 |
|
}, |
|
{ |
|
"epoch": 5.15, |
|
"learning_rate": 9.502464166133706e-06, |
|
"loss": 6.2036, |
|
"step": 274500 |
|
}, |
|
{ |
|
"epoch": 5.16, |
|
"learning_rate": 9.501523644708626e-06, |
|
"loss": 6.2037, |
|
"step": 275000 |
|
}, |
|
{ |
|
"epoch": 5.17, |
|
"learning_rate": 9.500583123283549e-06, |
|
"loss": 6.2, |
|
"step": 275500 |
|
}, |
|
{ |
|
"epoch": 5.18, |
|
"learning_rate": 9.49964260185847e-06, |
|
"loss": 6.2013, |
|
"step": 276000 |
|
}, |
|
{ |
|
"epoch": 5.19, |
|
"learning_rate": 9.498702080433393e-06, |
|
"loss": 6.2019, |
|
"step": 276500 |
|
}, |
|
{ |
|
"epoch": 5.2, |
|
"learning_rate": 9.497761559008315e-06, |
|
"loss": 6.2004, |
|
"step": 277000 |
|
}, |
|
{ |
|
"epoch": 5.21, |
|
"learning_rate": 9.496821037583237e-06, |
|
"loss": 6.2018, |
|
"step": 277500 |
|
}, |
|
{ |
|
"epoch": 5.22, |
|
"learning_rate": 9.49588051615816e-06, |
|
"loss": 6.1988, |
|
"step": 278000 |
|
}, |
|
{ |
|
"epoch": 5.23, |
|
"learning_rate": 9.494939994733082e-06, |
|
"loss": 6.2008, |
|
"step": 278500 |
|
}, |
|
{ |
|
"epoch": 5.24, |
|
"learning_rate": 9.493999473308003e-06, |
|
"loss": 6.1972, |
|
"step": 279000 |
|
}, |
|
{ |
|
"epoch": 5.25, |
|
"learning_rate": 9.493058951882925e-06, |
|
"loss": 6.1969, |
|
"step": 279500 |
|
}, |
|
{ |
|
"epoch": 5.26, |
|
"learning_rate": 9.492118430457847e-06, |
|
"loss": 6.1972, |
|
"step": 280000 |
|
}, |
|
{ |
|
"epoch": 5.27, |
|
"learning_rate": 9.491177909032768e-06, |
|
"loss": 6.1986, |
|
"step": 280500 |
|
}, |
|
{ |
|
"epoch": 5.28, |
|
"learning_rate": 9.49023738760769e-06, |
|
"loss": 6.1955, |
|
"step": 281000 |
|
}, |
|
{ |
|
"epoch": 5.29, |
|
"learning_rate": 9.489296866182612e-06, |
|
"loss": 6.2028, |
|
"step": 281500 |
|
}, |
|
{ |
|
"epoch": 5.29, |
|
"learning_rate": 9.488356344757534e-06, |
|
"loss": 6.2004, |
|
"step": 282000 |
|
}, |
|
{ |
|
"epoch": 5.3, |
|
"learning_rate": 9.487415823332457e-06, |
|
"loss": 6.1984, |
|
"step": 282500 |
|
}, |
|
{ |
|
"epoch": 5.31, |
|
"learning_rate": 9.486475301907379e-06, |
|
"loss": 6.1962, |
|
"step": 283000 |
|
}, |
|
{ |
|
"epoch": 5.32, |
|
"learning_rate": 9.485534780482301e-06, |
|
"loss": 6.2001, |
|
"step": 283500 |
|
}, |
|
{ |
|
"epoch": 5.33, |
|
"learning_rate": 9.484594259057223e-06, |
|
"loss": 6.1984, |
|
"step": 284000 |
|
}, |
|
{ |
|
"epoch": 5.34, |
|
"learning_rate": 9.483653737632144e-06, |
|
"loss": 6.2006, |
|
"step": 284500 |
|
}, |
|
{ |
|
"epoch": 5.35, |
|
"learning_rate": 9.482713216207066e-06, |
|
"loss": 6.1997, |
|
"step": 285000 |
|
}, |
|
{ |
|
"epoch": 5.36, |
|
"learning_rate": 9.481772694781988e-06, |
|
"loss": 6.1972, |
|
"step": 285500 |
|
}, |
|
{ |
|
"epoch": 5.37, |
|
"learning_rate": 9.48083217335691e-06, |
|
"loss": 6.1954, |
|
"step": 286000 |
|
}, |
|
{ |
|
"epoch": 5.38, |
|
"learning_rate": 9.479891651931831e-06, |
|
"loss": 6.1967, |
|
"step": 286500 |
|
}, |
|
{ |
|
"epoch": 5.39, |
|
"learning_rate": 9.478951130506753e-06, |
|
"loss": 6.195, |
|
"step": 287000 |
|
}, |
|
{ |
|
"epoch": 5.4, |
|
"learning_rate": 9.478010609081676e-06, |
|
"loss": 6.1948, |
|
"step": 287500 |
|
}, |
|
{ |
|
"epoch": 5.41, |
|
"learning_rate": 9.477070087656598e-06, |
|
"loss": 6.1964, |
|
"step": 288000 |
|
}, |
|
{ |
|
"epoch": 5.42, |
|
"learning_rate": 9.47612956623152e-06, |
|
"loss": 6.1917, |
|
"step": 288500 |
|
}, |
|
{ |
|
"epoch": 5.43, |
|
"learning_rate": 9.47518904480644e-06, |
|
"loss": 6.1963, |
|
"step": 289000 |
|
}, |
|
{ |
|
"epoch": 5.44, |
|
"learning_rate": 9.474248523381363e-06, |
|
"loss": 6.19, |
|
"step": 289500 |
|
}, |
|
{ |
|
"epoch": 5.44, |
|
"learning_rate": 9.473308001956285e-06, |
|
"loss": 6.1942, |
|
"step": 290000 |
|
}, |
|
{ |
|
"epoch": 5.45, |
|
"learning_rate": 9.472367480531207e-06, |
|
"loss": 6.1906, |
|
"step": 290500 |
|
}, |
|
{ |
|
"epoch": 5.46, |
|
"learning_rate": 9.47142695910613e-06, |
|
"loss": 6.1927, |
|
"step": 291000 |
|
}, |
|
{ |
|
"epoch": 5.47, |
|
"learning_rate": 9.470486437681052e-06, |
|
"loss": 6.1934, |
|
"step": 291500 |
|
}, |
|
{ |
|
"epoch": 5.48, |
|
"learning_rate": 9.469545916255974e-06, |
|
"loss": 6.1954, |
|
"step": 292000 |
|
}, |
|
{ |
|
"epoch": 5.49, |
|
"learning_rate": 9.468605394830895e-06, |
|
"loss": 6.1924, |
|
"step": 292500 |
|
}, |
|
{ |
|
"epoch": 5.5, |
|
"learning_rate": 9.467664873405817e-06, |
|
"loss": 6.193, |
|
"step": 293000 |
|
}, |
|
{ |
|
"epoch": 5.51, |
|
"learning_rate": 9.466724351980739e-06, |
|
"loss": 6.1953, |
|
"step": 293500 |
|
}, |
|
{ |
|
"epoch": 5.52, |
|
"learning_rate": 9.46578383055566e-06, |
|
"loss": 6.1924, |
|
"step": 294000 |
|
}, |
|
{ |
|
"epoch": 5.53, |
|
"learning_rate": 9.464843309130582e-06, |
|
"loss": 6.1949, |
|
"step": 294500 |
|
}, |
|
{ |
|
"epoch": 5.54, |
|
"learning_rate": 9.463902787705504e-06, |
|
"loss": 6.1913, |
|
"step": 295000 |
|
}, |
|
{ |
|
"epoch": 5.55, |
|
"learning_rate": 9.462962266280426e-06, |
|
"loss": 6.1928, |
|
"step": 295500 |
|
}, |
|
{ |
|
"epoch": 5.56, |
|
"learning_rate": 9.462021744855349e-06, |
|
"loss": 6.1908, |
|
"step": 296000 |
|
}, |
|
{ |
|
"epoch": 5.57, |
|
"learning_rate": 9.46108122343027e-06, |
|
"loss": 6.1914, |
|
"step": 296500 |
|
}, |
|
{ |
|
"epoch": 5.58, |
|
"learning_rate": 9.460140702005193e-06, |
|
"loss": 6.1938, |
|
"step": 297000 |
|
}, |
|
{ |
|
"epoch": 5.59, |
|
"learning_rate": 9.459200180580115e-06, |
|
"loss": 6.1932, |
|
"step": 297500 |
|
}, |
|
{ |
|
"epoch": 5.59, |
|
"learning_rate": 9.458259659155038e-06, |
|
"loss": 6.1959, |
|
"step": 298000 |
|
}, |
|
{ |
|
"epoch": 5.6, |
|
"learning_rate": 9.457319137729958e-06, |
|
"loss": 6.1929, |
|
"step": 298500 |
|
}, |
|
{ |
|
"epoch": 5.61, |
|
"learning_rate": 9.45637861630488e-06, |
|
"loss": 6.1939, |
|
"step": 299000 |
|
}, |
|
{ |
|
"epoch": 5.62, |
|
"learning_rate": 9.455438094879803e-06, |
|
"loss": 6.1963, |
|
"step": 299500 |
|
}, |
|
{ |
|
"epoch": 5.63, |
|
"learning_rate": 9.454497573454723e-06, |
|
"loss": 6.1852, |
|
"step": 300000 |
|
}, |
|
{ |
|
"epoch": 5.64, |
|
"learning_rate": 9.453557052029645e-06, |
|
"loss": 6.1891, |
|
"step": 300500 |
|
}, |
|
{ |
|
"epoch": 5.65, |
|
"learning_rate": 9.452616530604568e-06, |
|
"loss": 6.1862, |
|
"step": 301000 |
|
}, |
|
{ |
|
"epoch": 5.66, |
|
"learning_rate": 9.45167600917949e-06, |
|
"loss": 6.1914, |
|
"step": 301500 |
|
}, |
|
{ |
|
"epoch": 5.67, |
|
"learning_rate": 9.450735487754412e-06, |
|
"loss": 6.1902, |
|
"step": 302000 |
|
}, |
|
{ |
|
"epoch": 5.68, |
|
"learning_rate": 9.449794966329334e-06, |
|
"loss": 6.1893, |
|
"step": 302500 |
|
}, |
|
{ |
|
"epoch": 5.69, |
|
"learning_rate": 9.448854444904255e-06, |
|
"loss": 6.1882, |
|
"step": 303000 |
|
}, |
|
{ |
|
"epoch": 5.7, |
|
"learning_rate": 9.447913923479177e-06, |
|
"loss": 6.1876, |
|
"step": 303500 |
|
}, |
|
{ |
|
"epoch": 5.71, |
|
"learning_rate": 9.4469734020541e-06, |
|
"loss": 6.1907, |
|
"step": 304000 |
|
}, |
|
{ |
|
"epoch": 5.72, |
|
"learning_rate": 9.446032880629022e-06, |
|
"loss": 6.192, |
|
"step": 304500 |
|
}, |
|
{ |
|
"epoch": 5.73, |
|
"learning_rate": 9.445092359203944e-06, |
|
"loss": 6.187, |
|
"step": 305000 |
|
}, |
|
{ |
|
"epoch": 5.74, |
|
"learning_rate": 9.444151837778866e-06, |
|
"loss": 6.1816, |
|
"step": 305500 |
|
}, |
|
{ |
|
"epoch": 5.75, |
|
"learning_rate": 9.443211316353788e-06, |
|
"loss": 6.1847, |
|
"step": 306000 |
|
}, |
|
{ |
|
"epoch": 5.75, |
|
"learning_rate": 9.442270794928709e-06, |
|
"loss": 6.1874, |
|
"step": 306500 |
|
}, |
|
{ |
|
"epoch": 5.76, |
|
"learning_rate": 9.441330273503631e-06, |
|
"loss": 6.1919, |
|
"step": 307000 |
|
}, |
|
{ |
|
"epoch": 5.77, |
|
"learning_rate": 9.440389752078553e-06, |
|
"loss": 6.1901, |
|
"step": 307500 |
|
}, |
|
{ |
|
"epoch": 5.78, |
|
"learning_rate": 9.439449230653474e-06, |
|
"loss": 6.1871, |
|
"step": 308000 |
|
}, |
|
{ |
|
"epoch": 5.79, |
|
"learning_rate": 9.438508709228396e-06, |
|
"loss": 6.187, |
|
"step": 308500 |
|
}, |
|
{ |
|
"epoch": 5.8, |
|
"learning_rate": 9.437568187803318e-06, |
|
"loss": 6.1868, |
|
"step": 309000 |
|
}, |
|
{ |
|
"epoch": 5.81, |
|
"learning_rate": 9.43662766637824e-06, |
|
"loss": 6.1826, |
|
"step": 309500 |
|
}, |
|
{ |
|
"epoch": 5.82, |
|
"learning_rate": 9.435687144953163e-06, |
|
"loss": 6.1857, |
|
"step": 310000 |
|
}, |
|
{ |
|
"epoch": 5.83, |
|
"learning_rate": 9.434746623528085e-06, |
|
"loss": 6.1862, |
|
"step": 310500 |
|
}, |
|
{ |
|
"epoch": 5.84, |
|
"learning_rate": 9.433806102103007e-06, |
|
"loss": 6.1851, |
|
"step": 311000 |
|
}, |
|
{ |
|
"epoch": 5.85, |
|
"learning_rate": 9.43286558067793e-06, |
|
"loss": 6.1853, |
|
"step": 311500 |
|
}, |
|
{ |
|
"epoch": 5.86, |
|
"learning_rate": 9.431925059252852e-06, |
|
"loss": 6.1835, |
|
"step": 312000 |
|
}, |
|
{ |
|
"epoch": 5.87, |
|
"learning_rate": 9.430984537827772e-06, |
|
"loss": 6.1897, |
|
"step": 312500 |
|
}, |
|
{ |
|
"epoch": 5.88, |
|
"learning_rate": 9.430044016402695e-06, |
|
"loss": 6.1846, |
|
"step": 313000 |
|
}, |
|
{ |
|
"epoch": 5.89, |
|
"learning_rate": 9.429103494977617e-06, |
|
"loss": 6.184, |
|
"step": 313500 |
|
}, |
|
{ |
|
"epoch": 5.9, |
|
"learning_rate": 9.428162973552537e-06, |
|
"loss": 6.1859, |
|
"step": 314000 |
|
}, |
|
{ |
|
"epoch": 5.9, |
|
"learning_rate": 9.42722245212746e-06, |
|
"loss": 6.1876, |
|
"step": 314500 |
|
}, |
|
{ |
|
"epoch": 5.91, |
|
"learning_rate": 9.426281930702382e-06, |
|
"loss": 6.1839, |
|
"step": 315000 |
|
}, |
|
{ |
|
"epoch": 5.92, |
|
"learning_rate": 9.425341409277304e-06, |
|
"loss": 6.187, |
|
"step": 315500 |
|
}, |
|
{ |
|
"epoch": 5.93, |
|
"learning_rate": 9.424400887852226e-06, |
|
"loss": 6.1829, |
|
"step": 316000 |
|
}, |
|
{ |
|
"epoch": 5.94, |
|
"learning_rate": 9.423460366427149e-06, |
|
"loss": 6.1835, |
|
"step": 316500 |
|
}, |
|
{ |
|
"epoch": 5.95, |
|
"learning_rate": 9.42251984500207e-06, |
|
"loss": 6.184, |
|
"step": 317000 |
|
}, |
|
{ |
|
"epoch": 5.96, |
|
"learning_rate": 9.421579323576991e-06, |
|
"loss": 6.1802, |
|
"step": 317500 |
|
}, |
|
{ |
|
"epoch": 5.97, |
|
"learning_rate": 9.420638802151914e-06, |
|
"loss": 6.1834, |
|
"step": 318000 |
|
}, |
|
{ |
|
"epoch": 5.98, |
|
"learning_rate": 9.419698280726836e-06, |
|
"loss": 6.1829, |
|
"step": 318500 |
|
}, |
|
{ |
|
"epoch": 5.99, |
|
"learning_rate": 9.418757759301758e-06, |
|
"loss": 6.1836, |
|
"step": 319000 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"learning_rate": 9.41781723787668e-06, |
|
"loss": 6.1841, |
|
"step": 319500 |
|
}, |
|
{ |
|
"epoch": 6.01, |
|
"learning_rate": 9.416876716451601e-06, |
|
"loss": 6.1846, |
|
"step": 320000 |
|
}, |
|
{ |
|
"epoch": 6.02, |
|
"learning_rate": 9.415936195026523e-06, |
|
"loss": 6.1838, |
|
"step": 320500 |
|
}, |
|
{ |
|
"epoch": 6.03, |
|
"learning_rate": 9.414995673601445e-06, |
|
"loss": 6.1821, |
|
"step": 321000 |
|
}, |
|
{ |
|
"epoch": 6.04, |
|
"learning_rate": 9.414055152176368e-06, |
|
"loss": 6.183, |
|
"step": 321500 |
|
}, |
|
{ |
|
"epoch": 6.05, |
|
"learning_rate": 9.413114630751288e-06, |
|
"loss": 6.1828, |
|
"step": 322000 |
|
}, |
|
{ |
|
"epoch": 6.05, |
|
"learning_rate": 9.41217410932621e-06, |
|
"loss": 6.1819, |
|
"step": 322500 |
|
}, |
|
{ |
|
"epoch": 6.06, |
|
"learning_rate": 9.411233587901133e-06, |
|
"loss": 6.1863, |
|
"step": 323000 |
|
}, |
|
{ |
|
"epoch": 6.07, |
|
"learning_rate": 9.410293066476055e-06, |
|
"loss": 6.1835, |
|
"step": 323500 |
|
}, |
|
{ |
|
"epoch": 6.08, |
|
"learning_rate": 9.409352545050977e-06, |
|
"loss": 6.1814, |
|
"step": 324000 |
|
}, |
|
{ |
|
"epoch": 6.09, |
|
"learning_rate": 9.4084120236259e-06, |
|
"loss": 6.179, |
|
"step": 324500 |
|
}, |
|
{ |
|
"epoch": 6.1, |
|
"learning_rate": 9.407471502200822e-06, |
|
"loss": 6.1815, |
|
"step": 325000 |
|
}, |
|
{ |
|
"epoch": 6.11, |
|
"learning_rate": 9.406530980775744e-06, |
|
"loss": 6.1843, |
|
"step": 325500 |
|
}, |
|
{ |
|
"epoch": 6.12, |
|
"learning_rate": 9.405590459350664e-06, |
|
"loss": 6.1836, |
|
"step": 326000 |
|
}, |
|
{ |
|
"epoch": 6.13, |
|
"learning_rate": 9.404649937925587e-06, |
|
"loss": 6.1838, |
|
"step": 326500 |
|
}, |
|
{ |
|
"epoch": 6.14, |
|
"learning_rate": 9.403709416500509e-06, |
|
"loss": 6.1825, |
|
"step": 327000 |
|
}, |
|
{ |
|
"epoch": 6.15, |
|
"learning_rate": 9.402768895075431e-06, |
|
"loss": 6.1746, |
|
"step": 327500 |
|
}, |
|
{ |
|
"epoch": 6.16, |
|
"learning_rate": 9.401828373650352e-06, |
|
"loss": 6.1839, |
|
"step": 328000 |
|
}, |
|
{ |
|
"epoch": 6.17, |
|
"learning_rate": 9.400887852225274e-06, |
|
"loss": 6.1799, |
|
"step": 328500 |
|
}, |
|
{ |
|
"epoch": 6.18, |
|
"learning_rate": 9.399947330800196e-06, |
|
"loss": 6.177, |
|
"step": 329000 |
|
}, |
|
{ |
|
"epoch": 6.19, |
|
"learning_rate": 9.399006809375118e-06, |
|
"loss": 6.1802, |
|
"step": 329500 |
|
}, |
|
{ |
|
"epoch": 6.2, |
|
"learning_rate": 9.39806628795004e-06, |
|
"loss": 6.1836, |
|
"step": 330000 |
|
}, |
|
{ |
|
"epoch": 6.21, |
|
"learning_rate": 9.397125766524963e-06, |
|
"loss": 6.179, |
|
"step": 330500 |
|
}, |
|
{ |
|
"epoch": 6.21, |
|
"learning_rate": 9.396185245099883e-06, |
|
"loss": 6.181, |
|
"step": 331000 |
|
}, |
|
{ |
|
"epoch": 6.22, |
|
"learning_rate": 9.395244723674806e-06, |
|
"loss": 6.1817, |
|
"step": 331500 |
|
}, |
|
{ |
|
"epoch": 6.23, |
|
"learning_rate": 9.394304202249728e-06, |
|
"loss": 6.1796, |
|
"step": 332000 |
|
}, |
|
{ |
|
"epoch": 6.24, |
|
"learning_rate": 9.39336368082465e-06, |
|
"loss": 6.178, |
|
"step": 332500 |
|
}, |
|
{ |
|
"epoch": 6.25, |
|
"learning_rate": 9.392423159399572e-06, |
|
"loss": 6.1792, |
|
"step": 333000 |
|
}, |
|
{ |
|
"epoch": 6.26, |
|
"learning_rate": 9.391482637974495e-06, |
|
"loss": 6.1775, |
|
"step": 333500 |
|
}, |
|
{ |
|
"epoch": 6.27, |
|
"learning_rate": 9.390542116549415e-06, |
|
"loss": 6.1788, |
|
"step": 334000 |
|
}, |
|
{ |
|
"epoch": 6.28, |
|
"learning_rate": 9.389601595124337e-06, |
|
"loss": 6.1772, |
|
"step": 334500 |
|
}, |
|
{ |
|
"epoch": 6.29, |
|
"learning_rate": 9.38866107369926e-06, |
|
"loss": 6.1826, |
|
"step": 335000 |
|
}, |
|
{ |
|
"epoch": 6.3, |
|
"learning_rate": 9.387720552274182e-06, |
|
"loss": 6.1789, |
|
"step": 335500 |
|
}, |
|
{ |
|
"epoch": 6.31, |
|
"learning_rate": 9.386780030849103e-06, |
|
"loss": 6.1753, |
|
"step": 336000 |
|
}, |
|
{ |
|
"epoch": 6.32, |
|
"learning_rate": 9.385839509424025e-06, |
|
"loss": 6.1785, |
|
"step": 336500 |
|
}, |
|
{ |
|
"epoch": 6.33, |
|
"learning_rate": 9.384898987998947e-06, |
|
"loss": 6.1738, |
|
"step": 337000 |
|
}, |
|
{ |
|
"epoch": 6.34, |
|
"learning_rate": 9.38395846657387e-06, |
|
"loss": 6.1738, |
|
"step": 337500 |
|
}, |
|
{ |
|
"epoch": 6.35, |
|
"learning_rate": 9.383017945148791e-06, |
|
"loss": 6.1781, |
|
"step": 338000 |
|
}, |
|
{ |
|
"epoch": 6.36, |
|
"learning_rate": 9.382077423723714e-06, |
|
"loss": 6.1746, |
|
"step": 338500 |
|
}, |
|
{ |
|
"epoch": 6.36, |
|
"learning_rate": 9.381136902298636e-06, |
|
"loss": 6.179, |
|
"step": 339000 |
|
}, |
|
{ |
|
"epoch": 6.37, |
|
"learning_rate": 9.380196380873558e-06, |
|
"loss": 6.1779, |
|
"step": 339500 |
|
}, |
|
{ |
|
"epoch": 6.38, |
|
"learning_rate": 9.379255859448479e-06, |
|
"loss": 6.1741, |
|
"step": 340000 |
|
}, |
|
{ |
|
"epoch": 6.39, |
|
"learning_rate": 9.378315338023401e-06, |
|
"loss": 6.17, |
|
"step": 340500 |
|
}, |
|
{ |
|
"epoch": 6.4, |
|
"learning_rate": 9.377374816598323e-06, |
|
"loss": 6.1762, |
|
"step": 341000 |
|
}, |
|
{ |
|
"epoch": 6.41, |
|
"learning_rate": 9.376434295173244e-06, |
|
"loss": 6.171, |
|
"step": 341500 |
|
}, |
|
{ |
|
"epoch": 6.42, |
|
"learning_rate": 9.375493773748166e-06, |
|
"loss": 6.1739, |
|
"step": 342000 |
|
}, |
|
{ |
|
"epoch": 6.43, |
|
"learning_rate": 9.374553252323088e-06, |
|
"loss": 6.1735, |
|
"step": 342500 |
|
}, |
|
{ |
|
"epoch": 6.44, |
|
"learning_rate": 9.37361273089801e-06, |
|
"loss": 6.1766, |
|
"step": 343000 |
|
}, |
|
{ |
|
"epoch": 6.45, |
|
"learning_rate": 9.372672209472933e-06, |
|
"loss": 6.1741, |
|
"step": 343500 |
|
}, |
|
{ |
|
"epoch": 6.46, |
|
"learning_rate": 9.371731688047855e-06, |
|
"loss": 6.1716, |
|
"step": 344000 |
|
}, |
|
{ |
|
"epoch": 6.47, |
|
"learning_rate": 9.370791166622777e-06, |
|
"loss": 6.1749, |
|
"step": 344500 |
|
}, |
|
{ |
|
"epoch": 6.48, |
|
"learning_rate": 9.369850645197698e-06, |
|
"loss": 6.1709, |
|
"step": 345000 |
|
}, |
|
{ |
|
"epoch": 6.49, |
|
"learning_rate": 9.36891012377262e-06, |
|
"loss": 6.1736, |
|
"step": 345500 |
|
}, |
|
{ |
|
"epoch": 6.5, |
|
"learning_rate": 9.367969602347542e-06, |
|
"loss": 6.1758, |
|
"step": 346000 |
|
}, |
|
{ |
|
"epoch": 6.51, |
|
"learning_rate": 9.367029080922464e-06, |
|
"loss": 6.1762, |
|
"step": 346500 |
|
}, |
|
{ |
|
"epoch": 6.51, |
|
"learning_rate": 9.366088559497387e-06, |
|
"loss": 6.1756, |
|
"step": 347000 |
|
}, |
|
{ |
|
"epoch": 6.52, |
|
"learning_rate": 9.365148038072307e-06, |
|
"loss": 6.1707, |
|
"step": 347500 |
|
}, |
|
{ |
|
"epoch": 6.53, |
|
"learning_rate": 9.36420751664723e-06, |
|
"loss": 6.171, |
|
"step": 348000 |
|
}, |
|
{ |
|
"epoch": 6.54, |
|
"learning_rate": 9.363266995222152e-06, |
|
"loss": 6.1713, |
|
"step": 348500 |
|
}, |
|
{ |
|
"epoch": 6.55, |
|
"learning_rate": 9.362326473797074e-06, |
|
"loss": 6.169, |
|
"step": 349000 |
|
}, |
|
{ |
|
"epoch": 6.56, |
|
"learning_rate": 9.361385952371996e-06, |
|
"loss": 6.1732, |
|
"step": 349500 |
|
}, |
|
{ |
|
"epoch": 6.57, |
|
"learning_rate": 9.360445430946917e-06, |
|
"loss": 6.1728, |
|
"step": 350000 |
|
}, |
|
{ |
|
"epoch": 6.58, |
|
"learning_rate": 9.359504909521839e-06, |
|
"loss": 6.1728, |
|
"step": 350500 |
|
}, |
|
{ |
|
"epoch": 6.59, |
|
"learning_rate": 9.358564388096761e-06, |
|
"loss": 6.1732, |
|
"step": 351000 |
|
}, |
|
{ |
|
"epoch": 6.6, |
|
"learning_rate": 9.357623866671684e-06, |
|
"loss": 6.1684, |
|
"step": 351500 |
|
}, |
|
{ |
|
"epoch": 6.61, |
|
"learning_rate": 9.356683345246606e-06, |
|
"loss": 6.1731, |
|
"step": 352000 |
|
}, |
|
{ |
|
"epoch": 6.62, |
|
"learning_rate": 9.355742823821528e-06, |
|
"loss": 6.1717, |
|
"step": 352500 |
|
}, |
|
{ |
|
"epoch": 6.63, |
|
"learning_rate": 9.35480230239645e-06, |
|
"loss": 6.1694, |
|
"step": 353000 |
|
}, |
|
{ |
|
"epoch": 6.64, |
|
"learning_rate": 9.353861780971372e-06, |
|
"loss": 6.178, |
|
"step": 353500 |
|
}, |
|
{ |
|
"epoch": 6.65, |
|
"learning_rate": 9.352921259546293e-06, |
|
"loss": 6.1693, |
|
"step": 354000 |
|
}, |
|
{ |
|
"epoch": 6.66, |
|
"learning_rate": 9.351980738121215e-06, |
|
"loss": 6.1682, |
|
"step": 354500 |
|
}, |
|
{ |
|
"epoch": 6.67, |
|
"learning_rate": 9.351040216696138e-06, |
|
"loss": 6.1772, |
|
"step": 355000 |
|
}, |
|
{ |
|
"epoch": 6.67, |
|
"learning_rate": 9.350099695271058e-06, |
|
"loss": 6.1692, |
|
"step": 355500 |
|
}, |
|
{ |
|
"epoch": 6.68, |
|
"learning_rate": 9.34915917384598e-06, |
|
"loss": 6.1708, |
|
"step": 356000 |
|
}, |
|
{ |
|
"epoch": 6.69, |
|
"learning_rate": 9.348218652420903e-06, |
|
"loss": 6.1716, |
|
"step": 356500 |
|
}, |
|
{ |
|
"epoch": 6.7, |
|
"learning_rate": 9.347278130995825e-06, |
|
"loss": 6.1695, |
|
"step": 357000 |
|
}, |
|
{ |
|
"epoch": 6.71, |
|
"learning_rate": 9.346337609570747e-06, |
|
"loss": 6.1689, |
|
"step": 357500 |
|
}, |
|
{ |
|
"epoch": 6.72, |
|
"learning_rate": 9.34539708814567e-06, |
|
"loss": 6.1707, |
|
"step": 358000 |
|
}, |
|
{ |
|
"epoch": 6.73, |
|
"learning_rate": 9.344456566720592e-06, |
|
"loss": 6.1703, |
|
"step": 358500 |
|
}, |
|
{ |
|
"epoch": 6.74, |
|
"learning_rate": 9.343516045295514e-06, |
|
"loss": 6.1718, |
|
"step": 359000 |
|
}, |
|
{ |
|
"epoch": 6.75, |
|
"learning_rate": 9.342575523870434e-06, |
|
"loss": 6.167, |
|
"step": 359500 |
|
}, |
|
{ |
|
"epoch": 6.76, |
|
"learning_rate": 9.341635002445357e-06, |
|
"loss": 6.1638, |
|
"step": 360000 |
|
}, |
|
{ |
|
"epoch": 6.77, |
|
"learning_rate": 9.340694481020279e-06, |
|
"loss": 6.1711, |
|
"step": 360500 |
|
}, |
|
{ |
|
"epoch": 6.78, |
|
"learning_rate": 9.339753959595201e-06, |
|
"loss": 6.1744, |
|
"step": 361000 |
|
}, |
|
{ |
|
"epoch": 6.79, |
|
"learning_rate": 9.338813438170122e-06, |
|
"loss": 6.1665, |
|
"step": 361500 |
|
}, |
|
{ |
|
"epoch": 6.8, |
|
"learning_rate": 9.337872916745044e-06, |
|
"loss": 6.1681, |
|
"step": 362000 |
|
}, |
|
{ |
|
"epoch": 6.81, |
|
"learning_rate": 9.336932395319966e-06, |
|
"loss": 6.1679, |
|
"step": 362500 |
|
}, |
|
{ |
|
"epoch": 6.82, |
|
"learning_rate": 9.335991873894888e-06, |
|
"loss": 6.17, |
|
"step": 363000 |
|
}, |
|
{ |
|
"epoch": 6.82, |
|
"learning_rate": 9.33505135246981e-06, |
|
"loss": 6.1715, |
|
"step": 363500 |
|
}, |
|
{ |
|
"epoch": 6.83, |
|
"learning_rate": 9.334110831044731e-06, |
|
"loss": 6.1687, |
|
"step": 364000 |
|
}, |
|
{ |
|
"epoch": 6.84, |
|
"learning_rate": 9.333170309619653e-06, |
|
"loss": 6.1669, |
|
"step": 364500 |
|
}, |
|
{ |
|
"epoch": 6.85, |
|
"learning_rate": 9.332229788194576e-06, |
|
"loss": 6.1648, |
|
"step": 365000 |
|
}, |
|
{ |
|
"epoch": 6.86, |
|
"learning_rate": 9.331289266769498e-06, |
|
"loss": 6.1705, |
|
"step": 365500 |
|
}, |
|
{ |
|
"epoch": 6.87, |
|
"learning_rate": 9.33034874534442e-06, |
|
"loss": 6.1685, |
|
"step": 366000 |
|
}, |
|
{ |
|
"epoch": 6.88, |
|
"learning_rate": 9.329408223919342e-06, |
|
"loss": 6.1674, |
|
"step": 366500 |
|
}, |
|
{ |
|
"epoch": 6.89, |
|
"learning_rate": 9.328467702494265e-06, |
|
"loss": 6.163, |
|
"step": 367000 |
|
}, |
|
{ |
|
"epoch": 6.9, |
|
"learning_rate": 9.327527181069185e-06, |
|
"loss": 6.1622, |
|
"step": 367500 |
|
}, |
|
{ |
|
"epoch": 6.91, |
|
"learning_rate": 9.326586659644107e-06, |
|
"loss": 6.1638, |
|
"step": 368000 |
|
}, |
|
{ |
|
"epoch": 6.92, |
|
"learning_rate": 9.32564613821903e-06, |
|
"loss": 6.1645, |
|
"step": 368500 |
|
}, |
|
{ |
|
"epoch": 6.93, |
|
"learning_rate": 9.32470561679395e-06, |
|
"loss": 6.1641, |
|
"step": 369000 |
|
}, |
|
{ |
|
"epoch": 6.94, |
|
"learning_rate": 9.323765095368872e-06, |
|
"loss": 6.1633, |
|
"step": 369500 |
|
}, |
|
{ |
|
"epoch": 6.95, |
|
"learning_rate": 9.322824573943795e-06, |
|
"loss": 6.1723, |
|
"step": 370000 |
|
}, |
|
{ |
|
"epoch": 6.96, |
|
"learning_rate": 9.321884052518717e-06, |
|
"loss": 6.1626, |
|
"step": 370500 |
|
}, |
|
{ |
|
"epoch": 6.97, |
|
"learning_rate": 9.320943531093639e-06, |
|
"loss": 6.1688, |
|
"step": 371000 |
|
}, |
|
{ |
|
"epoch": 6.97, |
|
"learning_rate": 9.320003009668561e-06, |
|
"loss": 6.1688, |
|
"step": 371500 |
|
}, |
|
{ |
|
"epoch": 6.98, |
|
"learning_rate": 9.319062488243484e-06, |
|
"loss": 6.1649, |
|
"step": 372000 |
|
}, |
|
{ |
|
"epoch": 6.99, |
|
"learning_rate": 9.318121966818406e-06, |
|
"loss": 6.164, |
|
"step": 372500 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"learning_rate": 9.317181445393328e-06, |
|
"loss": 6.1653, |
|
"step": 373000 |
|
}, |
|
{ |
|
"epoch": 7.01, |
|
"learning_rate": 9.316240923968249e-06, |
|
"loss": 6.1656, |
|
"step": 373500 |
|
}, |
|
{ |
|
"epoch": 7.02, |
|
"learning_rate": 9.315300402543171e-06, |
|
"loss": 6.1627, |
|
"step": 374000 |
|
}, |
|
{ |
|
"epoch": 7.03, |
|
"learning_rate": 9.314359881118093e-06, |
|
"loss": 6.1658, |
|
"step": 374500 |
|
}, |
|
{ |
|
"epoch": 7.04, |
|
"learning_rate": 9.313419359693014e-06, |
|
"loss": 6.1616, |
|
"step": 375000 |
|
}, |
|
{ |
|
"epoch": 7.05, |
|
"learning_rate": 9.312478838267936e-06, |
|
"loss": 6.1684, |
|
"step": 375500 |
|
}, |
|
{ |
|
"epoch": 7.06, |
|
"learning_rate": 9.311538316842858e-06, |
|
"loss": 6.164, |
|
"step": 376000 |
|
}, |
|
{ |
|
"epoch": 7.07, |
|
"learning_rate": 9.31059779541778e-06, |
|
"loss": 6.1669, |
|
"step": 376500 |
|
}, |
|
{ |
|
"epoch": 7.08, |
|
"learning_rate": 9.309657273992703e-06, |
|
"loss": 6.1651, |
|
"step": 377000 |
|
}, |
|
{ |
|
"epoch": 7.09, |
|
"learning_rate": 9.308716752567625e-06, |
|
"loss": 6.1627, |
|
"step": 377500 |
|
}, |
|
{ |
|
"epoch": 7.1, |
|
"learning_rate": 9.307776231142545e-06, |
|
"loss": 6.169, |
|
"step": 378000 |
|
}, |
|
{ |
|
"epoch": 7.11, |
|
"learning_rate": 9.306835709717468e-06, |
|
"loss": 6.163, |
|
"step": 378500 |
|
}, |
|
{ |
|
"epoch": 7.12, |
|
"learning_rate": 9.30589518829239e-06, |
|
"loss": 6.1664, |
|
"step": 379000 |
|
}, |
|
{ |
|
"epoch": 7.13, |
|
"learning_rate": 9.304954666867312e-06, |
|
"loss": 6.1629, |
|
"step": 379500 |
|
}, |
|
{ |
|
"epoch": 7.13, |
|
"learning_rate": 9.304014145442234e-06, |
|
"loss": 6.1638, |
|
"step": 380000 |
|
}, |
|
{ |
|
"epoch": 7.14, |
|
"learning_rate": 9.303073624017157e-06, |
|
"loss": 6.162, |
|
"step": 380500 |
|
}, |
|
{ |
|
"epoch": 7.15, |
|
"learning_rate": 9.302133102592079e-06, |
|
"loss": 6.1678, |
|
"step": 381000 |
|
}, |
|
{ |
|
"epoch": 7.16, |
|
"learning_rate": 9.301192581167e-06, |
|
"loss": 6.1627, |
|
"step": 381500 |
|
}, |
|
{ |
|
"epoch": 7.17, |
|
"learning_rate": 9.300252059741922e-06, |
|
"loss": 6.1614, |
|
"step": 382000 |
|
}, |
|
{ |
|
"epoch": 7.18, |
|
"learning_rate": 9.299311538316844e-06, |
|
"loss": 6.1625, |
|
"step": 382500 |
|
}, |
|
{ |
|
"epoch": 7.19, |
|
"learning_rate": 9.298371016891764e-06, |
|
"loss": 6.1686, |
|
"step": 383000 |
|
}, |
|
{ |
|
"epoch": 7.2, |
|
"learning_rate": 9.297430495466687e-06, |
|
"loss": 6.1628, |
|
"step": 383500 |
|
}, |
|
{ |
|
"epoch": 7.21, |
|
"learning_rate": 9.296489974041609e-06, |
|
"loss": 6.1573, |
|
"step": 384000 |
|
}, |
|
{ |
|
"epoch": 7.22, |
|
"learning_rate": 9.295549452616531e-06, |
|
"loss": 6.1652, |
|
"step": 384500 |
|
}, |
|
{ |
|
"epoch": 7.23, |
|
"learning_rate": 9.294608931191453e-06, |
|
"loss": 6.1579, |
|
"step": 385000 |
|
}, |
|
{ |
|
"epoch": 7.24, |
|
"learning_rate": 9.293668409766376e-06, |
|
"loss": 6.1585, |
|
"step": 385500 |
|
}, |
|
{ |
|
"epoch": 7.25, |
|
"learning_rate": 9.292727888341298e-06, |
|
"loss": 6.1568, |
|
"step": 386000 |
|
}, |
|
{ |
|
"epoch": 7.26, |
|
"learning_rate": 9.29178736691622e-06, |
|
"loss": 6.1644, |
|
"step": 386500 |
|
}, |
|
{ |
|
"epoch": 7.27, |
|
"learning_rate": 9.290846845491142e-06, |
|
"loss": 6.162, |
|
"step": 387000 |
|
}, |
|
{ |
|
"epoch": 7.28, |
|
"learning_rate": 9.289906324066063e-06, |
|
"loss": 6.1583, |
|
"step": 387500 |
|
}, |
|
{ |
|
"epoch": 7.28, |
|
"learning_rate": 9.288965802640985e-06, |
|
"loss": 6.1562, |
|
"step": 388000 |
|
}, |
|
{ |
|
"epoch": 7.29, |
|
"learning_rate": 9.288025281215907e-06, |
|
"loss": 6.1567, |
|
"step": 388500 |
|
}, |
|
{ |
|
"epoch": 7.3, |
|
"learning_rate": 9.287084759790828e-06, |
|
"loss": 6.1636, |
|
"step": 389000 |
|
}, |
|
{ |
|
"epoch": 7.31, |
|
"learning_rate": 9.28614423836575e-06, |
|
"loss": 6.1597, |
|
"step": 389500 |
|
}, |
|
{ |
|
"epoch": 7.32, |
|
"learning_rate": 9.285203716940672e-06, |
|
"loss": 6.1563, |
|
"step": 390000 |
|
}, |
|
{ |
|
"epoch": 7.33, |
|
"learning_rate": 9.284263195515595e-06, |
|
"loss": 6.1592, |
|
"step": 390500 |
|
}, |
|
{ |
|
"epoch": 7.34, |
|
"learning_rate": 9.283322674090517e-06, |
|
"loss": 6.1576, |
|
"step": 391000 |
|
}, |
|
{ |
|
"epoch": 7.35, |
|
"learning_rate": 9.282382152665439e-06, |
|
"loss": 6.1574, |
|
"step": 391500 |
|
}, |
|
{ |
|
"epoch": 7.36, |
|
"learning_rate": 9.28144163124036e-06, |
|
"loss": 6.1588, |
|
"step": 392000 |
|
}, |
|
{ |
|
"epoch": 7.37, |
|
"learning_rate": 9.280501109815282e-06, |
|
"loss": 6.1603, |
|
"step": 392500 |
|
}, |
|
{ |
|
"epoch": 7.38, |
|
"learning_rate": 9.279560588390204e-06, |
|
"loss": 6.1582, |
|
"step": 393000 |
|
}, |
|
{ |
|
"epoch": 7.39, |
|
"learning_rate": 9.278620066965126e-06, |
|
"loss": 6.1591, |
|
"step": 393500 |
|
}, |
|
{ |
|
"epoch": 7.4, |
|
"learning_rate": 9.277679545540049e-06, |
|
"loss": 6.16, |
|
"step": 394000 |
|
}, |
|
{ |
|
"epoch": 7.41, |
|
"learning_rate": 9.276739024114971e-06, |
|
"loss": 6.1581, |
|
"step": 394500 |
|
}, |
|
{ |
|
"epoch": 7.42, |
|
"learning_rate": 9.275798502689891e-06, |
|
"loss": 6.1588, |
|
"step": 395000 |
|
}, |
|
{ |
|
"epoch": 7.43, |
|
"learning_rate": 9.274857981264814e-06, |
|
"loss": 6.1583, |
|
"step": 395500 |
|
}, |
|
{ |
|
"epoch": 7.43, |
|
"learning_rate": 9.273917459839736e-06, |
|
"loss": 6.1561, |
|
"step": 396000 |
|
}, |
|
{ |
|
"epoch": 7.44, |
|
"learning_rate": 9.272976938414656e-06, |
|
"loss": 6.1625, |
|
"step": 396500 |
|
}, |
|
{ |
|
"epoch": 7.45, |
|
"learning_rate": 9.272036416989579e-06, |
|
"loss": 6.1577, |
|
"step": 397000 |
|
}, |
|
{ |
|
"epoch": 7.46, |
|
"learning_rate": 9.271095895564501e-06, |
|
"loss": 6.1594, |
|
"step": 397500 |
|
}, |
|
{ |
|
"epoch": 7.47, |
|
"learning_rate": 9.270155374139423e-06, |
|
"loss": 6.1623, |
|
"step": 398000 |
|
}, |
|
{ |
|
"epoch": 7.48, |
|
"learning_rate": 9.269214852714345e-06, |
|
"loss": 6.1609, |
|
"step": 398500 |
|
}, |
|
{ |
|
"epoch": 7.49, |
|
"learning_rate": 9.268274331289268e-06, |
|
"loss": 6.1654, |
|
"step": 399000 |
|
}, |
|
{ |
|
"epoch": 7.5, |
|
"learning_rate": 9.26733380986419e-06, |
|
"loss": 6.1614, |
|
"step": 399500 |
|
}, |
|
{ |
|
"epoch": 7.51, |
|
"learning_rate": 9.266393288439112e-06, |
|
"loss": 6.1612, |
|
"step": 400000 |
|
}, |
|
{ |
|
"epoch": 7.52, |
|
"learning_rate": 9.265452767014034e-06, |
|
"loss": 6.1583, |
|
"step": 400500 |
|
}, |
|
{ |
|
"epoch": 7.53, |
|
"learning_rate": 9.264512245588955e-06, |
|
"loss": 6.1571, |
|
"step": 401000 |
|
}, |
|
{ |
|
"epoch": 7.54, |
|
"learning_rate": 9.263571724163877e-06, |
|
"loss": 6.1598, |
|
"step": 401500 |
|
}, |
|
{ |
|
"epoch": 7.55, |
|
"learning_rate": 9.2626312027388e-06, |
|
"loss": 6.1546, |
|
"step": 402000 |
|
}, |
|
{ |
|
"epoch": 7.56, |
|
"learning_rate": 9.261690681313722e-06, |
|
"loss": 6.1621, |
|
"step": 402500 |
|
}, |
|
{ |
|
"epoch": 7.57, |
|
"learning_rate": 9.260750159888642e-06, |
|
"loss": 6.1612, |
|
"step": 403000 |
|
}, |
|
{ |
|
"epoch": 7.58, |
|
"learning_rate": 9.259809638463564e-06, |
|
"loss": 6.1528, |
|
"step": 403500 |
|
}, |
|
{ |
|
"epoch": 7.59, |
|
"learning_rate": 9.258869117038487e-06, |
|
"loss": 6.1581, |
|
"step": 404000 |
|
}, |
|
{ |
|
"epoch": 7.59, |
|
"learning_rate": 9.257928595613409e-06, |
|
"loss": 6.1601, |
|
"step": 404500 |
|
}, |
|
{ |
|
"epoch": 7.6, |
|
"learning_rate": 9.256988074188331e-06, |
|
"loss": 6.1548, |
|
"step": 405000 |
|
}, |
|
{ |
|
"epoch": 7.61, |
|
"learning_rate": 9.256047552763253e-06, |
|
"loss": 6.1564, |
|
"step": 405500 |
|
}, |
|
{ |
|
"epoch": 7.62, |
|
"learning_rate": 9.255107031338174e-06, |
|
"loss": 6.1582, |
|
"step": 406000 |
|
}, |
|
{ |
|
"epoch": 7.63, |
|
"learning_rate": 9.254166509913096e-06, |
|
"loss": 6.155, |
|
"step": 406500 |
|
}, |
|
{ |
|
"epoch": 7.64, |
|
"learning_rate": 9.253225988488018e-06, |
|
"loss": 6.1566, |
|
"step": 407000 |
|
}, |
|
{ |
|
"epoch": 7.65, |
|
"learning_rate": 9.25228546706294e-06, |
|
"loss": 6.1548, |
|
"step": 407500 |
|
}, |
|
{ |
|
"epoch": 7.66, |
|
"learning_rate": 9.251344945637863e-06, |
|
"loss": 6.1546, |
|
"step": 408000 |
|
}, |
|
{ |
|
"epoch": 7.67, |
|
"learning_rate": 9.250404424212785e-06, |
|
"loss": 6.1538, |
|
"step": 408500 |
|
}, |
|
{ |
|
"epoch": 7.68, |
|
"learning_rate": 9.249463902787706e-06, |
|
"loss": 6.1559, |
|
"step": 409000 |
|
}, |
|
{ |
|
"epoch": 7.69, |
|
"learning_rate": 9.248523381362628e-06, |
|
"loss": 6.1543, |
|
"step": 409500 |
|
}, |
|
{ |
|
"epoch": 7.7, |
|
"learning_rate": 9.24758285993755e-06, |
|
"loss": 6.1521, |
|
"step": 410000 |
|
}, |
|
{ |
|
"epoch": 7.71, |
|
"learning_rate": 9.246642338512472e-06, |
|
"loss": 6.1579, |
|
"step": 410500 |
|
}, |
|
{ |
|
"epoch": 7.72, |
|
"learning_rate": 9.245701817087393e-06, |
|
"loss": 6.1532, |
|
"step": 411000 |
|
}, |
|
{ |
|
"epoch": 7.73, |
|
"learning_rate": 9.244761295662315e-06, |
|
"loss": 6.1536, |
|
"step": 411500 |
|
}, |
|
{ |
|
"epoch": 7.74, |
|
"learning_rate": 9.243820774237238e-06, |
|
"loss": 6.1526, |
|
"step": 412000 |
|
}, |
|
{ |
|
"epoch": 7.74, |
|
"learning_rate": 9.24288025281216e-06, |
|
"loss": 6.1536, |
|
"step": 412500 |
|
}, |
|
{ |
|
"epoch": 7.75, |
|
"learning_rate": 9.241939731387082e-06, |
|
"loss": 6.1536, |
|
"step": 413000 |
|
}, |
|
{ |
|
"epoch": 7.76, |
|
"learning_rate": 9.240999209962004e-06, |
|
"loss": 6.1555, |
|
"step": 413500 |
|
}, |
|
{ |
|
"epoch": 7.77, |
|
"learning_rate": 9.240058688536926e-06, |
|
"loss": 6.1544, |
|
"step": 414000 |
|
}, |
|
{ |
|
"epoch": 7.78, |
|
"learning_rate": 9.239118167111849e-06, |
|
"loss": 6.1504, |
|
"step": 414500 |
|
}, |
|
{ |
|
"epoch": 7.79, |
|
"learning_rate": 9.23817764568677e-06, |
|
"loss": 6.1553, |
|
"step": 415000 |
|
}, |
|
{ |
|
"epoch": 7.8, |
|
"learning_rate": 9.237237124261692e-06, |
|
"loss": 6.1534, |
|
"step": 415500 |
|
}, |
|
{ |
|
"epoch": 7.81, |
|
"learning_rate": 9.236296602836614e-06, |
|
"loss": 6.1536, |
|
"step": 416000 |
|
}, |
|
{ |
|
"epoch": 7.82, |
|
"learning_rate": 9.235356081411534e-06, |
|
"loss": 6.1532, |
|
"step": 416500 |
|
}, |
|
{ |
|
"epoch": 7.83, |
|
"learning_rate": 9.234415559986457e-06, |
|
"loss": 6.1522, |
|
"step": 417000 |
|
}, |
|
{ |
|
"epoch": 7.84, |
|
"learning_rate": 9.233475038561379e-06, |
|
"loss": 6.1517, |
|
"step": 417500 |
|
}, |
|
{ |
|
"epoch": 7.85, |
|
"learning_rate": 9.232534517136301e-06, |
|
"loss": 6.1531, |
|
"step": 418000 |
|
}, |
|
{ |
|
"epoch": 7.86, |
|
"learning_rate": 9.231593995711223e-06, |
|
"loss": 6.1557, |
|
"step": 418500 |
|
}, |
|
{ |
|
"epoch": 7.87, |
|
"learning_rate": 9.230653474286146e-06, |
|
"loss": 6.1479, |
|
"step": 419000 |
|
}, |
|
{ |
|
"epoch": 7.88, |
|
"learning_rate": 9.229712952861068e-06, |
|
"loss": 6.1535, |
|
"step": 419500 |
|
}, |
|
{ |
|
"epoch": 7.89, |
|
"learning_rate": 9.228772431435988e-06, |
|
"loss": 6.1514, |
|
"step": 420000 |
|
}, |
|
{ |
|
"epoch": 7.89, |
|
"learning_rate": 9.22783191001091e-06, |
|
"loss": 6.1483, |
|
"step": 420500 |
|
}, |
|
{ |
|
"epoch": 7.9, |
|
"learning_rate": 9.226891388585833e-06, |
|
"loss": 6.1559, |
|
"step": 421000 |
|
}, |
|
{ |
|
"epoch": 7.91, |
|
"learning_rate": 9.225950867160755e-06, |
|
"loss": 6.1473, |
|
"step": 421500 |
|
}, |
|
{ |
|
"epoch": 7.92, |
|
"learning_rate": 9.225010345735677e-06, |
|
"loss": 6.1497, |
|
"step": 422000 |
|
}, |
|
{ |
|
"epoch": 7.93, |
|
"learning_rate": 9.224069824310598e-06, |
|
"loss": 6.1549, |
|
"step": 422500 |
|
}, |
|
{ |
|
"epoch": 7.94, |
|
"learning_rate": 9.22312930288552e-06, |
|
"loss": 6.1512, |
|
"step": 423000 |
|
}, |
|
{ |
|
"epoch": 7.95, |
|
"learning_rate": 9.222188781460442e-06, |
|
"loss": 6.153, |
|
"step": 423500 |
|
}, |
|
{ |
|
"epoch": 7.96, |
|
"learning_rate": 9.221248260035365e-06, |
|
"loss": 6.1516, |
|
"step": 424000 |
|
}, |
|
{ |
|
"epoch": 7.97, |
|
"learning_rate": 9.220307738610287e-06, |
|
"loss": 6.147, |
|
"step": 424500 |
|
}, |
|
{ |
|
"epoch": 7.98, |
|
"learning_rate": 9.219367217185207e-06, |
|
"loss": 6.1489, |
|
"step": 425000 |
|
}, |
|
{ |
|
"epoch": 7.99, |
|
"learning_rate": 9.21842669576013e-06, |
|
"loss": 6.1521, |
|
"step": 425500 |
|
}, |
|
{ |
|
"epoch": 8.0, |
|
"learning_rate": 9.217486174335052e-06, |
|
"loss": 6.1533, |
|
"step": 426000 |
|
}, |
|
{ |
|
"epoch": 8.01, |
|
"learning_rate": 9.216545652909974e-06, |
|
"loss": 6.1516, |
|
"step": 426500 |
|
}, |
|
{ |
|
"epoch": 8.02, |
|
"learning_rate": 9.215605131484896e-06, |
|
"loss": 6.1505, |
|
"step": 427000 |
|
}, |
|
{ |
|
"epoch": 8.03, |
|
"learning_rate": 9.214664610059819e-06, |
|
"loss": 6.1499, |
|
"step": 427500 |
|
}, |
|
{ |
|
"epoch": 8.04, |
|
"learning_rate": 9.21372408863474e-06, |
|
"loss": 6.143, |
|
"step": 428000 |
|
}, |
|
{ |
|
"epoch": 8.05, |
|
"learning_rate": 9.212783567209661e-06, |
|
"loss": 6.1487, |
|
"step": 428500 |
|
}, |
|
{ |
|
"epoch": 8.05, |
|
"learning_rate": 9.211843045784584e-06, |
|
"loss": 6.1522, |
|
"step": 429000 |
|
}, |
|
{ |
|
"epoch": 8.06, |
|
"learning_rate": 9.210902524359506e-06, |
|
"loss": 6.15, |
|
"step": 429500 |
|
}, |
|
{ |
|
"epoch": 8.07, |
|
"learning_rate": 9.209962002934428e-06, |
|
"loss": 6.1469, |
|
"step": 430000 |
|
}, |
|
{ |
|
"epoch": 8.08, |
|
"learning_rate": 9.209021481509349e-06, |
|
"loss": 6.1498, |
|
"step": 430500 |
|
}, |
|
{ |
|
"epoch": 8.09, |
|
"learning_rate": 9.20808096008427e-06, |
|
"loss": 6.1491, |
|
"step": 431000 |
|
}, |
|
{ |
|
"epoch": 8.1, |
|
"learning_rate": 9.207140438659193e-06, |
|
"loss": 6.1495, |
|
"step": 431500 |
|
}, |
|
{ |
|
"epoch": 8.11, |
|
"learning_rate": 9.206199917234115e-06, |
|
"loss": 6.1479, |
|
"step": 432000 |
|
}, |
|
{ |
|
"epoch": 8.12, |
|
"learning_rate": 9.205259395809038e-06, |
|
"loss": 6.1484, |
|
"step": 432500 |
|
}, |
|
{ |
|
"epoch": 8.13, |
|
"learning_rate": 9.20431887438396e-06, |
|
"loss": 6.1467, |
|
"step": 433000 |
|
}, |
|
{ |
|
"epoch": 8.14, |
|
"learning_rate": 9.203378352958882e-06, |
|
"loss": 6.1451, |
|
"step": 433500 |
|
}, |
|
{ |
|
"epoch": 8.15, |
|
"learning_rate": 9.202437831533803e-06, |
|
"loss": 6.1487, |
|
"step": 434000 |
|
}, |
|
{ |
|
"epoch": 8.16, |
|
"learning_rate": 9.201497310108725e-06, |
|
"loss": 6.1486, |
|
"step": 434500 |
|
}, |
|
{ |
|
"epoch": 8.17, |
|
"learning_rate": 9.200556788683647e-06, |
|
"loss": 6.1503, |
|
"step": 435000 |
|
}, |
|
{ |
|
"epoch": 8.18, |
|
"learning_rate": 9.19961626725857e-06, |
|
"loss": 6.1497, |
|
"step": 435500 |
|
}, |
|
{ |
|
"epoch": 8.19, |
|
"learning_rate": 9.198675745833492e-06, |
|
"loss": 6.1512, |
|
"step": 436000 |
|
}, |
|
{ |
|
"epoch": 8.2, |
|
"learning_rate": 9.197735224408412e-06, |
|
"loss": 6.1482, |
|
"step": 436500 |
|
}, |
|
{ |
|
"epoch": 8.2, |
|
"learning_rate": 9.196794702983334e-06, |
|
"loss": 6.1475, |
|
"step": 437000 |
|
}, |
|
{ |
|
"epoch": 8.21, |
|
"learning_rate": 9.195854181558257e-06, |
|
"loss": 6.1522, |
|
"step": 437500 |
|
}, |
|
{ |
|
"epoch": 8.22, |
|
"learning_rate": 9.194913660133179e-06, |
|
"loss": 6.1484, |
|
"step": 438000 |
|
}, |
|
{ |
|
"epoch": 8.23, |
|
"learning_rate": 9.193973138708101e-06, |
|
"loss": 6.1472, |
|
"step": 438500 |
|
}, |
|
{ |
|
"epoch": 8.24, |
|
"learning_rate": 9.193032617283022e-06, |
|
"loss": 6.1496, |
|
"step": 439000 |
|
}, |
|
{ |
|
"epoch": 8.25, |
|
"learning_rate": 9.192092095857944e-06, |
|
"loss": 6.145, |
|
"step": 439500 |
|
}, |
|
{ |
|
"epoch": 8.26, |
|
"learning_rate": 9.191151574432866e-06, |
|
"loss": 6.1483, |
|
"step": 440000 |
|
}, |
|
{ |
|
"epoch": 8.27, |
|
"learning_rate": 9.190211053007788e-06, |
|
"loss": 6.1474, |
|
"step": 440500 |
|
}, |
|
{ |
|
"epoch": 8.28, |
|
"learning_rate": 9.18927053158271e-06, |
|
"loss": 6.1519, |
|
"step": 441000 |
|
}, |
|
{ |
|
"epoch": 8.29, |
|
"learning_rate": 9.188330010157633e-06, |
|
"loss": 6.1457, |
|
"step": 441500 |
|
}, |
|
{ |
|
"epoch": 8.3, |
|
"learning_rate": 9.187389488732555e-06, |
|
"loss": 6.1546, |
|
"step": 442000 |
|
}, |
|
{ |
|
"epoch": 8.31, |
|
"learning_rate": 9.186448967307476e-06, |
|
"loss": 6.1521, |
|
"step": 442500 |
|
}, |
|
{ |
|
"epoch": 8.32, |
|
"learning_rate": 9.185508445882398e-06, |
|
"loss": 6.1489, |
|
"step": 443000 |
|
}, |
|
{ |
|
"epoch": 8.33, |
|
"learning_rate": 9.18456792445732e-06, |
|
"loss": 6.144, |
|
"step": 443500 |
|
}, |
|
{ |
|
"epoch": 8.34, |
|
"learning_rate": 9.18362740303224e-06, |
|
"loss": 6.1488, |
|
"step": 444000 |
|
}, |
|
{ |
|
"epoch": 8.35, |
|
"learning_rate": 9.182686881607163e-06, |
|
"loss": 6.148, |
|
"step": 444500 |
|
}, |
|
{ |
|
"epoch": 8.35, |
|
"learning_rate": 9.181746360182085e-06, |
|
"loss": 6.1493, |
|
"step": 445000 |
|
}, |
|
{ |
|
"epoch": 8.36, |
|
"learning_rate": 9.180805838757007e-06, |
|
"loss": 6.149, |
|
"step": 445500 |
|
}, |
|
{ |
|
"epoch": 8.37, |
|
"learning_rate": 9.17986531733193e-06, |
|
"loss": 6.143, |
|
"step": 446000 |
|
}, |
|
{ |
|
"epoch": 8.38, |
|
"learning_rate": 9.178924795906852e-06, |
|
"loss": 6.1425, |
|
"step": 446500 |
|
}, |
|
{ |
|
"epoch": 8.39, |
|
"learning_rate": 9.177984274481774e-06, |
|
"loss": 6.1459, |
|
"step": 447000 |
|
}, |
|
{ |
|
"epoch": 8.4, |
|
"learning_rate": 9.177043753056696e-06, |
|
"loss": 6.1452, |
|
"step": 447500 |
|
}, |
|
{ |
|
"epoch": 8.41, |
|
"learning_rate": 9.176103231631619e-06, |
|
"loss": 6.1414, |
|
"step": 448000 |
|
}, |
|
{ |
|
"epoch": 8.42, |
|
"learning_rate": 9.175162710206539e-06, |
|
"loss": 6.1449, |
|
"step": 448500 |
|
}, |
|
{ |
|
"epoch": 8.43, |
|
"learning_rate": 9.174222188781461e-06, |
|
"loss": 6.1461, |
|
"step": 449000 |
|
}, |
|
{ |
|
"epoch": 8.44, |
|
"learning_rate": 9.173281667356384e-06, |
|
"loss": 6.141, |
|
"step": 449500 |
|
}, |
|
{ |
|
"epoch": 8.45, |
|
"learning_rate": 9.172341145931304e-06, |
|
"loss": 6.1457, |
|
"step": 450000 |
|
}, |
|
{ |
|
"epoch": 8.46, |
|
"learning_rate": 9.171400624506226e-06, |
|
"loss": 6.1451, |
|
"step": 450500 |
|
}, |
|
{ |
|
"epoch": 8.47, |
|
"learning_rate": 9.170460103081149e-06, |
|
"loss": 6.1451, |
|
"step": 451000 |
|
}, |
|
{ |
|
"epoch": 8.48, |
|
"learning_rate": 9.169519581656071e-06, |
|
"loss": 6.1481, |
|
"step": 451500 |
|
}, |
|
{ |
|
"epoch": 8.49, |
|
"learning_rate": 9.168579060230993e-06, |
|
"loss": 6.1489, |
|
"step": 452000 |
|
}, |
|
{ |
|
"epoch": 8.5, |
|
"learning_rate": 9.167638538805915e-06, |
|
"loss": 6.1411, |
|
"step": 452500 |
|
}, |
|
{ |
|
"epoch": 8.51, |
|
"learning_rate": 9.166698017380836e-06, |
|
"loss": 6.1481, |
|
"step": 453000 |
|
}, |
|
{ |
|
"epoch": 8.51, |
|
"learning_rate": 9.165757495955758e-06, |
|
"loss": 6.1418, |
|
"step": 453500 |
|
}, |
|
{ |
|
"epoch": 8.52, |
|
"learning_rate": 9.16481697453068e-06, |
|
"loss": 6.1406, |
|
"step": 454000 |
|
}, |
|
{ |
|
"epoch": 8.53, |
|
"learning_rate": 9.163876453105603e-06, |
|
"loss": 6.1398, |
|
"step": 454500 |
|
}, |
|
{ |
|
"epoch": 8.54, |
|
"learning_rate": 9.162935931680525e-06, |
|
"loss": 6.1475, |
|
"step": 455000 |
|
}, |
|
{ |
|
"epoch": 8.55, |
|
"learning_rate": 9.161995410255447e-06, |
|
"loss": 6.1464, |
|
"step": 455500 |
|
}, |
|
{ |
|
"epoch": 8.56, |
|
"learning_rate": 9.16105488883037e-06, |
|
"loss": 6.1431, |
|
"step": 456000 |
|
}, |
|
{ |
|
"epoch": 8.57, |
|
"learning_rate": 9.16011436740529e-06, |
|
"loss": 6.146, |
|
"step": 456500 |
|
}, |
|
{ |
|
"epoch": 8.58, |
|
"learning_rate": 9.159173845980212e-06, |
|
"loss": 6.144, |
|
"step": 457000 |
|
}, |
|
{ |
|
"epoch": 8.59, |
|
"learning_rate": 9.158233324555134e-06, |
|
"loss": 6.1384, |
|
"step": 457500 |
|
}, |
|
{ |
|
"epoch": 8.6, |
|
"learning_rate": 9.157292803130055e-06, |
|
"loss": 6.1442, |
|
"step": 458000 |
|
}, |
|
{ |
|
"epoch": 8.61, |
|
"learning_rate": 9.156352281704977e-06, |
|
"loss": 6.1382, |
|
"step": 458500 |
|
}, |
|
{ |
|
"epoch": 8.62, |
|
"learning_rate": 9.1554117602799e-06, |
|
"loss": 6.1435, |
|
"step": 459000 |
|
}, |
|
{ |
|
"epoch": 8.63, |
|
"learning_rate": 9.154471238854822e-06, |
|
"loss": 6.1398, |
|
"step": 459500 |
|
}, |
|
{ |
|
"epoch": 8.64, |
|
"learning_rate": 9.153530717429744e-06, |
|
"loss": 6.1453, |
|
"step": 460000 |
|
}, |
|
{ |
|
"epoch": 8.65, |
|
"learning_rate": 9.152590196004666e-06, |
|
"loss": 6.1452, |
|
"step": 460500 |
|
}, |
|
{ |
|
"epoch": 8.66, |
|
"learning_rate": 9.151649674579588e-06, |
|
"loss": 6.1494, |
|
"step": 461000 |
|
}, |
|
{ |
|
"epoch": 8.66, |
|
"learning_rate": 9.15070915315451e-06, |
|
"loss": 6.1421, |
|
"step": 461500 |
|
}, |
|
{ |
|
"epoch": 8.67, |
|
"learning_rate": 9.149768631729433e-06, |
|
"loss": 6.1379, |
|
"step": 462000 |
|
}, |
|
{ |
|
"epoch": 8.68, |
|
"learning_rate": 9.148828110304353e-06, |
|
"loss": 6.1457, |
|
"step": 462500 |
|
}, |
|
{ |
|
"epoch": 8.69, |
|
"learning_rate": 9.147887588879276e-06, |
|
"loss": 6.1413, |
|
"step": 463000 |
|
}, |
|
{ |
|
"epoch": 8.7, |
|
"learning_rate": 9.146947067454198e-06, |
|
"loss": 6.1447, |
|
"step": 463500 |
|
}, |
|
{ |
|
"epoch": 8.71, |
|
"learning_rate": 9.146006546029118e-06, |
|
"loss": 6.1414, |
|
"step": 464000 |
|
}, |
|
{ |
|
"epoch": 8.72, |
|
"learning_rate": 9.14506602460404e-06, |
|
"loss": 6.143, |
|
"step": 464500 |
|
}, |
|
{ |
|
"epoch": 8.73, |
|
"learning_rate": 9.144125503178963e-06, |
|
"loss": 6.1426, |
|
"step": 465000 |
|
}, |
|
{ |
|
"epoch": 8.74, |
|
"learning_rate": 9.143184981753885e-06, |
|
"loss": 6.1427, |
|
"step": 465500 |
|
}, |
|
{ |
|
"epoch": 8.75, |
|
"learning_rate": 9.142244460328807e-06, |
|
"loss": 6.1428, |
|
"step": 466000 |
|
}, |
|
{ |
|
"epoch": 8.76, |
|
"learning_rate": 9.14130393890373e-06, |
|
"loss": 6.1448, |
|
"step": 466500 |
|
}, |
|
{ |
|
"epoch": 8.77, |
|
"learning_rate": 9.14036341747865e-06, |
|
"loss": 6.1395, |
|
"step": 467000 |
|
}, |
|
{ |
|
"epoch": 8.78, |
|
"learning_rate": 9.139422896053572e-06, |
|
"loss": 6.1428, |
|
"step": 467500 |
|
}, |
|
{ |
|
"epoch": 8.79, |
|
"learning_rate": 9.138482374628495e-06, |
|
"loss": 6.1453, |
|
"step": 468000 |
|
}, |
|
{ |
|
"epoch": 8.8, |
|
"learning_rate": 9.137541853203417e-06, |
|
"loss": 6.1369, |
|
"step": 468500 |
|
}, |
|
{ |
|
"epoch": 8.81, |
|
"learning_rate": 9.13660133177834e-06, |
|
"loss": 6.141, |
|
"step": 469000 |
|
}, |
|
{ |
|
"epoch": 8.81, |
|
"learning_rate": 9.135660810353261e-06, |
|
"loss": 6.1419, |
|
"step": 469500 |
|
}, |
|
{ |
|
"epoch": 8.82, |
|
"learning_rate": 9.134720288928182e-06, |
|
"loss": 6.1391, |
|
"step": 470000 |
|
}, |
|
{ |
|
"epoch": 8.83, |
|
"learning_rate": 9.133779767503104e-06, |
|
"loss": 6.1419, |
|
"step": 470500 |
|
}, |
|
{ |
|
"epoch": 8.84, |
|
"learning_rate": 9.132839246078026e-06, |
|
"loss": 6.143, |
|
"step": 471000 |
|
}, |
|
{ |
|
"epoch": 8.85, |
|
"learning_rate": 9.131898724652947e-06, |
|
"loss": 6.1416, |
|
"step": 471500 |
|
}, |
|
{ |
|
"epoch": 8.86, |
|
"learning_rate": 9.13095820322787e-06, |
|
"loss": 6.1403, |
|
"step": 472000 |
|
}, |
|
{ |
|
"epoch": 8.87, |
|
"learning_rate": 9.130017681802791e-06, |
|
"loss": 6.1365, |
|
"step": 472500 |
|
}, |
|
{ |
|
"epoch": 8.88, |
|
"learning_rate": 9.129077160377714e-06, |
|
"loss": 6.1376, |
|
"step": 473000 |
|
}, |
|
{ |
|
"epoch": 8.89, |
|
"learning_rate": 9.128136638952636e-06, |
|
"loss": 6.1377, |
|
"step": 473500 |
|
}, |
|
{ |
|
"epoch": 8.9, |
|
"learning_rate": 9.127196117527558e-06, |
|
"loss": 6.1404, |
|
"step": 474000 |
|
}, |
|
{ |
|
"epoch": 8.91, |
|
"learning_rate": 9.12625559610248e-06, |
|
"loss": 6.1403, |
|
"step": 474500 |
|
}, |
|
{ |
|
"epoch": 8.92, |
|
"learning_rate": 9.125315074677403e-06, |
|
"loss": 6.1411, |
|
"step": 475000 |
|
}, |
|
{ |
|
"epoch": 8.93, |
|
"learning_rate": 9.124374553252325e-06, |
|
"loss": 6.1448, |
|
"step": 475500 |
|
}, |
|
{ |
|
"epoch": 8.94, |
|
"learning_rate": 9.123434031827245e-06, |
|
"loss": 6.1323, |
|
"step": 476000 |
|
}, |
|
{ |
|
"epoch": 8.95, |
|
"learning_rate": 9.122493510402168e-06, |
|
"loss": 6.1371, |
|
"step": 476500 |
|
}, |
|
{ |
|
"epoch": 8.96, |
|
"learning_rate": 9.12155298897709e-06, |
|
"loss": 6.1374, |
|
"step": 477000 |
|
}, |
|
{ |
|
"epoch": 8.97, |
|
"learning_rate": 9.12061246755201e-06, |
|
"loss": 6.1393, |
|
"step": 477500 |
|
}, |
|
{ |
|
"epoch": 8.97, |
|
"learning_rate": 9.119671946126933e-06, |
|
"loss": 6.1418, |
|
"step": 478000 |
|
}, |
|
{ |
|
"epoch": 8.98, |
|
"learning_rate": 9.118731424701855e-06, |
|
"loss": 6.1383, |
|
"step": 478500 |
|
}, |
|
{ |
|
"epoch": 8.99, |
|
"learning_rate": 9.117790903276777e-06, |
|
"loss": 6.1415, |
|
"step": 479000 |
|
}, |
|
{ |
|
"epoch": 9.0, |
|
"learning_rate": 9.1168503818517e-06, |
|
"loss": 6.1365, |
|
"step": 479500 |
|
}, |
|
{ |
|
"epoch": 9.01, |
|
"learning_rate": 9.115909860426622e-06, |
|
"loss": 6.1378, |
|
"step": 480000 |
|
}, |
|
{ |
|
"epoch": 9.02, |
|
"learning_rate": 9.114969339001544e-06, |
|
"loss": 6.1351, |
|
"step": 480500 |
|
}, |
|
{ |
|
"epoch": 9.03, |
|
"learning_rate": 9.114028817576465e-06, |
|
"loss": 6.1367, |
|
"step": 481000 |
|
}, |
|
{ |
|
"epoch": 9.04, |
|
"learning_rate": 9.113088296151387e-06, |
|
"loss": 6.1354, |
|
"step": 481500 |
|
}, |
|
{ |
|
"epoch": 9.05, |
|
"learning_rate": 9.112147774726309e-06, |
|
"loss": 6.1459, |
|
"step": 482000 |
|
}, |
|
{ |
|
"epoch": 9.06, |
|
"learning_rate": 9.111207253301231e-06, |
|
"loss": 6.1389, |
|
"step": 482500 |
|
}, |
|
{ |
|
"epoch": 9.07, |
|
"learning_rate": 9.110266731876153e-06, |
|
"loss": 6.1416, |
|
"step": 483000 |
|
}, |
|
{ |
|
"epoch": 9.08, |
|
"learning_rate": 9.109326210451076e-06, |
|
"loss": 6.1327, |
|
"step": 483500 |
|
}, |
|
{ |
|
"epoch": 9.09, |
|
"learning_rate": 9.108385689025996e-06, |
|
"loss": 6.1377, |
|
"step": 484000 |
|
}, |
|
{ |
|
"epoch": 9.1, |
|
"learning_rate": 9.107445167600919e-06, |
|
"loss": 6.1376, |
|
"step": 484500 |
|
}, |
|
{ |
|
"epoch": 9.11, |
|
"learning_rate": 9.10650464617584e-06, |
|
"loss": 6.137, |
|
"step": 485000 |
|
}, |
|
{ |
|
"epoch": 9.12, |
|
"learning_rate": 9.105564124750761e-06, |
|
"loss": 6.1372, |
|
"step": 485500 |
|
}, |
|
{ |
|
"epoch": 9.12, |
|
"learning_rate": 9.104623603325684e-06, |
|
"loss": 6.1305, |
|
"step": 486000 |
|
}, |
|
{ |
|
"epoch": 9.13, |
|
"learning_rate": 9.103683081900606e-06, |
|
"loss": 6.1363, |
|
"step": 486500 |
|
}, |
|
{ |
|
"epoch": 9.14, |
|
"learning_rate": 9.102742560475528e-06, |
|
"loss": 6.1375, |
|
"step": 487000 |
|
}, |
|
{ |
|
"epoch": 9.15, |
|
"learning_rate": 9.10180203905045e-06, |
|
"loss": 6.1366, |
|
"step": 487500 |
|
}, |
|
{ |
|
"epoch": 9.16, |
|
"learning_rate": 9.100861517625373e-06, |
|
"loss": 6.1383, |
|
"step": 488000 |
|
}, |
|
{ |
|
"epoch": 9.17, |
|
"learning_rate": 9.099920996200295e-06, |
|
"loss": 6.1381, |
|
"step": 488500 |
|
}, |
|
{ |
|
"epoch": 9.18, |
|
"learning_rate": 9.098980474775217e-06, |
|
"loss": 6.1432, |
|
"step": 489000 |
|
}, |
|
{ |
|
"epoch": 9.19, |
|
"learning_rate": 9.09803995335014e-06, |
|
"loss": 6.135, |
|
"step": 489500 |
|
}, |
|
{ |
|
"epoch": 9.2, |
|
"learning_rate": 9.09709943192506e-06, |
|
"loss": 6.1402, |
|
"step": 490000 |
|
}, |
|
{ |
|
"epoch": 9.21, |
|
"learning_rate": 9.096158910499982e-06, |
|
"loss": 6.1369, |
|
"step": 490500 |
|
}, |
|
{ |
|
"epoch": 9.22, |
|
"learning_rate": 9.095218389074904e-06, |
|
"loss": 6.1316, |
|
"step": 491000 |
|
}, |
|
{ |
|
"epoch": 9.23, |
|
"learning_rate": 9.094277867649825e-06, |
|
"loss": 6.1418, |
|
"step": 491500 |
|
}, |
|
{ |
|
"epoch": 9.24, |
|
"learning_rate": 9.093337346224747e-06, |
|
"loss": 6.1369, |
|
"step": 492000 |
|
}, |
|
{ |
|
"epoch": 9.25, |
|
"learning_rate": 9.09239682479967e-06, |
|
"loss": 6.1403, |
|
"step": 492500 |
|
}, |
|
{ |
|
"epoch": 9.26, |
|
"learning_rate": 9.091456303374592e-06, |
|
"loss": 6.135, |
|
"step": 493000 |
|
}, |
|
{ |
|
"epoch": 9.27, |
|
"learning_rate": 9.090515781949514e-06, |
|
"loss": 6.1362, |
|
"step": 493500 |
|
}, |
|
{ |
|
"epoch": 9.27, |
|
"learning_rate": 9.089575260524436e-06, |
|
"loss": 6.1321, |
|
"step": 494000 |
|
}, |
|
{ |
|
"epoch": 9.28, |
|
"learning_rate": 9.088634739099358e-06, |
|
"loss": 6.1388, |
|
"step": 494500 |
|
}, |
|
{ |
|
"epoch": 9.29, |
|
"learning_rate": 9.087694217674279e-06, |
|
"loss": 6.1391, |
|
"step": 495000 |
|
}, |
|
{ |
|
"epoch": 9.3, |
|
"learning_rate": 9.086753696249201e-06, |
|
"loss": 6.1341, |
|
"step": 495500 |
|
}, |
|
{ |
|
"epoch": 9.31, |
|
"learning_rate": 9.085813174824123e-06, |
|
"loss": 6.1355, |
|
"step": 496000 |
|
}, |
|
{ |
|
"epoch": 9.32, |
|
"learning_rate": 9.084872653399046e-06, |
|
"loss": 6.1354, |
|
"step": 496500 |
|
}, |
|
{ |
|
"epoch": 9.33, |
|
"learning_rate": 9.083932131973968e-06, |
|
"loss": 6.1366, |
|
"step": 497000 |
|
}, |
|
{ |
|
"epoch": 9.34, |
|
"learning_rate": 9.082991610548888e-06, |
|
"loss": 6.1378, |
|
"step": 497500 |
|
}, |
|
{ |
|
"epoch": 9.35, |
|
"learning_rate": 9.08205108912381e-06, |
|
"loss": 6.1355, |
|
"step": 498000 |
|
}, |
|
{ |
|
"epoch": 9.36, |
|
"learning_rate": 9.081110567698733e-06, |
|
"loss": 6.1326, |
|
"step": 498500 |
|
}, |
|
{ |
|
"epoch": 9.37, |
|
"learning_rate": 9.080170046273655e-06, |
|
"loss": 6.1337, |
|
"step": 499000 |
|
}, |
|
{ |
|
"epoch": 9.38, |
|
"learning_rate": 9.079229524848577e-06, |
|
"loss": 6.1319, |
|
"step": 499500 |
|
}, |
|
{ |
|
"epoch": 9.39, |
|
"learning_rate": 9.078289003423498e-06, |
|
"loss": 6.1328, |
|
"step": 500000 |
|
}, |
|
{ |
|
"epoch": 9.4, |
|
"learning_rate": 9.07734848199842e-06, |
|
"loss": 6.1324, |
|
"step": 500500 |
|
}, |
|
{ |
|
"epoch": 9.41, |
|
"learning_rate": 9.076407960573342e-06, |
|
"loss": 6.1334, |
|
"step": 501000 |
|
}, |
|
{ |
|
"epoch": 9.42, |
|
"learning_rate": 9.075467439148265e-06, |
|
"loss": 6.1304, |
|
"step": 501500 |
|
}, |
|
{ |
|
"epoch": 9.43, |
|
"learning_rate": 9.074526917723187e-06, |
|
"loss": 6.1331, |
|
"step": 502000 |
|
}, |
|
{ |
|
"epoch": 9.43, |
|
"learning_rate": 9.073586396298109e-06, |
|
"loss": 6.1355, |
|
"step": 502500 |
|
}, |
|
{ |
|
"epoch": 9.44, |
|
"learning_rate": 9.072645874873031e-06, |
|
"loss": 6.1401, |
|
"step": 503000 |
|
}, |
|
{ |
|
"epoch": 9.45, |
|
"learning_rate": 9.071705353447952e-06, |
|
"loss": 6.1353, |
|
"step": 503500 |
|
}, |
|
{ |
|
"epoch": 9.46, |
|
"learning_rate": 9.070764832022874e-06, |
|
"loss": 6.1353, |
|
"step": 504000 |
|
}, |
|
{ |
|
"epoch": 9.47, |
|
"learning_rate": 9.069824310597796e-06, |
|
"loss": 6.1296, |
|
"step": 504500 |
|
}, |
|
{ |
|
"epoch": 9.48, |
|
"learning_rate": 9.068883789172719e-06, |
|
"loss": 6.1312, |
|
"step": 505000 |
|
}, |
|
{ |
|
"epoch": 9.49, |
|
"learning_rate": 9.067943267747639e-06, |
|
"loss": 6.1392, |
|
"step": 505500 |
|
}, |
|
{ |
|
"epoch": 9.5, |
|
"learning_rate": 9.067002746322561e-06, |
|
"loss": 6.1361, |
|
"step": 506000 |
|
}, |
|
{ |
|
"epoch": 9.51, |
|
"learning_rate": 9.066062224897484e-06, |
|
"loss": 6.1309, |
|
"step": 506500 |
|
}, |
|
{ |
|
"epoch": 9.52, |
|
"learning_rate": 9.065121703472406e-06, |
|
"loss": 6.1328, |
|
"step": 507000 |
|
}, |
|
{ |
|
"epoch": 9.53, |
|
"learning_rate": 9.064181182047328e-06, |
|
"loss": 6.1335, |
|
"step": 507500 |
|
}, |
|
{ |
|
"epoch": 9.54, |
|
"learning_rate": 9.06324066062225e-06, |
|
"loss": 6.1334, |
|
"step": 508000 |
|
}, |
|
{ |
|
"epoch": 9.55, |
|
"learning_rate": 9.062300139197173e-06, |
|
"loss": 6.1349, |
|
"step": 508500 |
|
}, |
|
{ |
|
"epoch": 9.56, |
|
"learning_rate": 9.061359617772093e-06, |
|
"loss": 6.1369, |
|
"step": 509000 |
|
}, |
|
{ |
|
"epoch": 9.57, |
|
"learning_rate": 9.060419096347015e-06, |
|
"loss": 6.1345, |
|
"step": 509500 |
|
}, |
|
{ |
|
"epoch": 9.58, |
|
"learning_rate": 9.059478574921938e-06, |
|
"loss": 6.1326, |
|
"step": 510000 |
|
}, |
|
{ |
|
"epoch": 9.58, |
|
"learning_rate": 9.05853805349686e-06, |
|
"loss": 6.127, |
|
"step": 510500 |
|
}, |
|
{ |
|
"epoch": 9.59, |
|
"learning_rate": 9.057597532071782e-06, |
|
"loss": 6.1346, |
|
"step": 511000 |
|
}, |
|
{ |
|
"epoch": 9.6, |
|
"learning_rate": 9.056657010646703e-06, |
|
"loss": 6.1347, |
|
"step": 511500 |
|
}, |
|
{ |
|
"epoch": 9.61, |
|
"learning_rate": 9.055716489221625e-06, |
|
"loss": 6.1351, |
|
"step": 512000 |
|
}, |
|
{ |
|
"epoch": 9.62, |
|
"learning_rate": 9.054775967796547e-06, |
|
"loss": 6.1313, |
|
"step": 512500 |
|
}, |
|
{ |
|
"epoch": 9.63, |
|
"learning_rate": 9.05383544637147e-06, |
|
"loss": 6.1371, |
|
"step": 513000 |
|
}, |
|
{ |
|
"epoch": 9.64, |
|
"learning_rate": 9.052894924946392e-06, |
|
"loss": 6.1319, |
|
"step": 513500 |
|
}, |
|
{ |
|
"epoch": 9.65, |
|
"learning_rate": 9.051954403521312e-06, |
|
"loss": 6.1326, |
|
"step": 514000 |
|
}, |
|
{ |
|
"epoch": 9.66, |
|
"learning_rate": 9.051013882096234e-06, |
|
"loss": 6.1303, |
|
"step": 514500 |
|
}, |
|
{ |
|
"epoch": 9.67, |
|
"learning_rate": 9.050073360671157e-06, |
|
"loss": 6.1277, |
|
"step": 515000 |
|
}, |
|
{ |
|
"epoch": 9.68, |
|
"learning_rate": 9.049132839246079e-06, |
|
"loss": 6.1348, |
|
"step": 515500 |
|
}, |
|
{ |
|
"epoch": 9.69, |
|
"learning_rate": 9.048192317821001e-06, |
|
"loss": 6.1325, |
|
"step": 516000 |
|
}, |
|
{ |
|
"epoch": 9.7, |
|
"learning_rate": 9.047251796395923e-06, |
|
"loss": 6.1351, |
|
"step": 516500 |
|
}, |
|
{ |
|
"epoch": 9.71, |
|
"learning_rate": 9.046311274970846e-06, |
|
"loss": 6.1338, |
|
"step": 517000 |
|
}, |
|
{ |
|
"epoch": 9.72, |
|
"learning_rate": 9.045370753545766e-06, |
|
"loss": 6.1354, |
|
"step": 517500 |
|
}, |
|
{ |
|
"epoch": 9.73, |
|
"learning_rate": 9.044430232120688e-06, |
|
"loss": 6.1313, |
|
"step": 518000 |
|
}, |
|
{ |
|
"epoch": 9.73, |
|
"learning_rate": 9.04348971069561e-06, |
|
"loss": 6.1346, |
|
"step": 518500 |
|
}, |
|
{ |
|
"epoch": 9.74, |
|
"learning_rate": 9.042549189270531e-06, |
|
"loss": 6.1317, |
|
"step": 519000 |
|
}, |
|
{ |
|
"epoch": 9.75, |
|
"learning_rate": 9.041608667845453e-06, |
|
"loss": 6.1343, |
|
"step": 519500 |
|
}, |
|
{ |
|
"epoch": 9.76, |
|
"learning_rate": 9.040668146420376e-06, |
|
"loss": 6.1319, |
|
"step": 520000 |
|
}, |
|
{ |
|
"epoch": 9.77, |
|
"learning_rate": 9.039727624995298e-06, |
|
"loss": 6.1331, |
|
"step": 520500 |
|
}, |
|
{ |
|
"epoch": 9.78, |
|
"learning_rate": 9.03878710357022e-06, |
|
"loss": 6.1253, |
|
"step": 521000 |
|
}, |
|
{ |
|
"epoch": 9.79, |
|
"learning_rate": 9.037846582145142e-06, |
|
"loss": 6.1318, |
|
"step": 521500 |
|
}, |
|
{ |
|
"epoch": 9.8, |
|
"learning_rate": 9.036906060720065e-06, |
|
"loss": 6.1299, |
|
"step": 522000 |
|
}, |
|
{ |
|
"epoch": 9.81, |
|
"learning_rate": 9.035965539294987e-06, |
|
"loss": 6.1296, |
|
"step": 522500 |
|
}, |
|
{ |
|
"epoch": 9.82, |
|
"learning_rate": 9.035025017869907e-06, |
|
"loss": 6.132, |
|
"step": 523000 |
|
}, |
|
{ |
|
"epoch": 9.83, |
|
"learning_rate": 9.03408449644483e-06, |
|
"loss": 6.1362, |
|
"step": 523500 |
|
}, |
|
{ |
|
"epoch": 9.84, |
|
"learning_rate": 9.033143975019752e-06, |
|
"loss": 6.1293, |
|
"step": 524000 |
|
}, |
|
{ |
|
"epoch": 9.85, |
|
"learning_rate": 9.032203453594674e-06, |
|
"loss": 6.1268, |
|
"step": 524500 |
|
}, |
|
{ |
|
"epoch": 9.86, |
|
"learning_rate": 9.031262932169595e-06, |
|
"loss": 6.1332, |
|
"step": 525000 |
|
}, |
|
{ |
|
"epoch": 9.87, |
|
"learning_rate": 9.030322410744517e-06, |
|
"loss": 6.1305, |
|
"step": 525500 |
|
}, |
|
{ |
|
"epoch": 9.88, |
|
"learning_rate": 9.02938188931944e-06, |
|
"loss": 6.1353, |
|
"step": 526000 |
|
}, |
|
{ |
|
"epoch": 9.89, |
|
"learning_rate": 9.028441367894361e-06, |
|
"loss": 6.1299, |
|
"step": 526500 |
|
}, |
|
{ |
|
"epoch": 9.89, |
|
"learning_rate": 9.027500846469284e-06, |
|
"loss": 6.1268, |
|
"step": 527000 |
|
}, |
|
{ |
|
"epoch": 9.9, |
|
"learning_rate": 9.026560325044206e-06, |
|
"loss": 6.1304, |
|
"step": 527500 |
|
}, |
|
{ |
|
"epoch": 9.91, |
|
"learning_rate": 9.025619803619126e-06, |
|
"loss": 6.1318, |
|
"step": 528000 |
|
}, |
|
{ |
|
"epoch": 9.92, |
|
"learning_rate": 9.024679282194049e-06, |
|
"loss": 6.1305, |
|
"step": 528500 |
|
}, |
|
{ |
|
"epoch": 9.93, |
|
"learning_rate": 9.023738760768971e-06, |
|
"loss": 6.1268, |
|
"step": 529000 |
|
}, |
|
{ |
|
"epoch": 9.94, |
|
"learning_rate": 9.022798239343893e-06, |
|
"loss": 6.1344, |
|
"step": 529500 |
|
}, |
|
{ |
|
"epoch": 9.95, |
|
"learning_rate": 9.021857717918815e-06, |
|
"loss": 6.1324, |
|
"step": 530000 |
|
}, |
|
{ |
|
"epoch": 9.96, |
|
"learning_rate": 9.020917196493738e-06, |
|
"loss": 6.127, |
|
"step": 530500 |
|
}, |
|
{ |
|
"epoch": 9.97, |
|
"learning_rate": 9.01997667506866e-06, |
|
"loss": 6.1316, |
|
"step": 531000 |
|
}, |
|
{ |
|
"epoch": 9.98, |
|
"learning_rate": 9.01903615364358e-06, |
|
"loss": 6.1274, |
|
"step": 531500 |
|
}, |
|
{ |
|
"epoch": 9.99, |
|
"learning_rate": 9.018095632218503e-06, |
|
"loss": 6.1288, |
|
"step": 532000 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"learning_rate": 9.017155110793425e-06, |
|
"loss": 6.1301, |
|
"step": 532500 |
|
}, |
|
{ |
|
"epoch": 10.01, |
|
"learning_rate": 9.016214589368345e-06, |
|
"loss": 6.1322, |
|
"step": 533000 |
|
}, |
|
{ |
|
"epoch": 10.02, |
|
"learning_rate": 9.015274067943268e-06, |
|
"loss": 6.1291, |
|
"step": 533500 |
|
}, |
|
{ |
|
"epoch": 10.03, |
|
"learning_rate": 9.01433354651819e-06, |
|
"loss": 6.1331, |
|
"step": 534000 |
|
}, |
|
{ |
|
"epoch": 10.04, |
|
"learning_rate": 9.013393025093112e-06, |
|
"loss": 6.1286, |
|
"step": 534500 |
|
}, |
|
{ |
|
"epoch": 10.04, |
|
"learning_rate": 9.012452503668034e-06, |
|
"loss": 6.1286, |
|
"step": 535000 |
|
}, |
|
{ |
|
"epoch": 10.05, |
|
"learning_rate": 9.011511982242957e-06, |
|
"loss": 6.1311, |
|
"step": 535500 |
|
}, |
|
{ |
|
"epoch": 10.06, |
|
"learning_rate": 9.010571460817879e-06, |
|
"loss": 6.1271, |
|
"step": 536000 |
|
}, |
|
{ |
|
"epoch": 10.07, |
|
"learning_rate": 9.009630939392801e-06, |
|
"loss": 6.1267, |
|
"step": 536500 |
|
}, |
|
{ |
|
"epoch": 10.08, |
|
"learning_rate": 9.008690417967723e-06, |
|
"loss": 6.1292, |
|
"step": 537000 |
|
}, |
|
{ |
|
"epoch": 10.09, |
|
"learning_rate": 9.007749896542644e-06, |
|
"loss": 6.1299, |
|
"step": 537500 |
|
}, |
|
{ |
|
"epoch": 10.1, |
|
"learning_rate": 9.006809375117566e-06, |
|
"loss": 6.126, |
|
"step": 538000 |
|
}, |
|
{ |
|
"epoch": 10.11, |
|
"learning_rate": 9.005868853692488e-06, |
|
"loss": 6.1266, |
|
"step": 538500 |
|
}, |
|
{ |
|
"epoch": 10.12, |
|
"learning_rate": 9.004928332267409e-06, |
|
"loss": 6.1274, |
|
"step": 539000 |
|
}, |
|
{ |
|
"epoch": 10.13, |
|
"learning_rate": 9.003987810842331e-06, |
|
"loss": 6.1299, |
|
"step": 539500 |
|
}, |
|
{ |
|
"epoch": 10.14, |
|
"learning_rate": 9.003047289417253e-06, |
|
"loss": 6.1294, |
|
"step": 540000 |
|
}, |
|
{ |
|
"epoch": 10.15, |
|
"learning_rate": 9.002106767992176e-06, |
|
"loss": 6.129, |
|
"step": 540500 |
|
}, |
|
{ |
|
"epoch": 10.16, |
|
"learning_rate": 9.001166246567098e-06, |
|
"loss": 6.1311, |
|
"step": 541000 |
|
}, |
|
{ |
|
"epoch": 10.17, |
|
"learning_rate": 9.00022572514202e-06, |
|
"loss": 6.1255, |
|
"step": 541500 |
|
}, |
|
{ |
|
"epoch": 10.18, |
|
"learning_rate": 8.99928520371694e-06, |
|
"loss": 6.1279, |
|
"step": 542000 |
|
}, |
|
{ |
|
"epoch": 10.19, |
|
"learning_rate": 8.998344682291863e-06, |
|
"loss": 6.1265, |
|
"step": 542500 |
|
}, |
|
{ |
|
"epoch": 10.19, |
|
"learning_rate": 8.997404160866785e-06, |
|
"loss": 6.1278, |
|
"step": 543000 |
|
}, |
|
{ |
|
"epoch": 10.2, |
|
"learning_rate": 8.996463639441707e-06, |
|
"loss": 6.1272, |
|
"step": 543500 |
|
}, |
|
{ |
|
"epoch": 10.21, |
|
"learning_rate": 8.99552311801663e-06, |
|
"loss": 6.1263, |
|
"step": 544000 |
|
}, |
|
{ |
|
"epoch": 10.22, |
|
"learning_rate": 8.994582596591552e-06, |
|
"loss": 6.1274, |
|
"step": 544500 |
|
}, |
|
{ |
|
"epoch": 10.23, |
|
"learning_rate": 8.993642075166472e-06, |
|
"loss": 6.1254, |
|
"step": 545000 |
|
}, |
|
{ |
|
"epoch": 10.24, |
|
"learning_rate": 8.992701553741395e-06, |
|
"loss": 6.1245, |
|
"step": 545500 |
|
}, |
|
{ |
|
"epoch": 10.25, |
|
"learning_rate": 8.991761032316317e-06, |
|
"loss": 6.1298, |
|
"step": 546000 |
|
}, |
|
{ |
|
"epoch": 10.26, |
|
"learning_rate": 8.990820510891238e-06, |
|
"loss": 6.1265, |
|
"step": 546500 |
|
}, |
|
{ |
|
"epoch": 10.27, |
|
"learning_rate": 8.98987998946616e-06, |
|
"loss": 6.1217, |
|
"step": 547000 |
|
}, |
|
{ |
|
"epoch": 10.28, |
|
"learning_rate": 8.988939468041082e-06, |
|
"loss": 6.1313, |
|
"step": 547500 |
|
}, |
|
{ |
|
"epoch": 10.29, |
|
"learning_rate": 8.987998946616004e-06, |
|
"loss": 6.1282, |
|
"step": 548000 |
|
}, |
|
{ |
|
"epoch": 10.3, |
|
"learning_rate": 8.987058425190926e-06, |
|
"loss": 6.1264, |
|
"step": 548500 |
|
}, |
|
{ |
|
"epoch": 10.31, |
|
"learning_rate": 8.986117903765849e-06, |
|
"loss": 6.1274, |
|
"step": 549000 |
|
}, |
|
{ |
|
"epoch": 10.32, |
|
"learning_rate": 8.985177382340771e-06, |
|
"loss": 6.1236, |
|
"step": 549500 |
|
}, |
|
{ |
|
"epoch": 10.33, |
|
"learning_rate": 8.984236860915693e-06, |
|
"loss": 6.1269, |
|
"step": 550000 |
|
}, |
|
{ |
|
"epoch": 10.34, |
|
"learning_rate": 8.983296339490615e-06, |
|
"loss": 6.1279, |
|
"step": 550500 |
|
}, |
|
{ |
|
"epoch": 10.35, |
|
"learning_rate": 8.982355818065536e-06, |
|
"loss": 6.1272, |
|
"step": 551000 |
|
}, |
|
{ |
|
"epoch": 10.35, |
|
"learning_rate": 8.981415296640458e-06, |
|
"loss": 6.1269, |
|
"step": 551500 |
|
}, |
|
{ |
|
"epoch": 10.36, |
|
"learning_rate": 8.98047477521538e-06, |
|
"loss": 6.131, |
|
"step": 552000 |
|
}, |
|
{ |
|
"epoch": 10.37, |
|
"learning_rate": 8.979534253790301e-06, |
|
"loss": 6.1273, |
|
"step": 552500 |
|
}, |
|
{ |
|
"epoch": 10.38, |
|
"learning_rate": 8.978593732365223e-06, |
|
"loss": 6.1217, |
|
"step": 553000 |
|
}, |
|
{ |
|
"epoch": 10.39, |
|
"learning_rate": 8.977653210940146e-06, |
|
"loss": 6.1223, |
|
"step": 553500 |
|
}, |
|
{ |
|
"epoch": 10.4, |
|
"learning_rate": 8.976712689515068e-06, |
|
"loss": 6.1262, |
|
"step": 554000 |
|
}, |
|
{ |
|
"epoch": 10.41, |
|
"learning_rate": 8.97577216808999e-06, |
|
"loss": 6.1283, |
|
"step": 554500 |
|
}, |
|
{ |
|
"epoch": 10.42, |
|
"learning_rate": 8.974831646664912e-06, |
|
"loss": 6.1299, |
|
"step": 555000 |
|
}, |
|
{ |
|
"epoch": 10.43, |
|
"learning_rate": 8.973891125239834e-06, |
|
"loss": 6.1241, |
|
"step": 555500 |
|
}, |
|
{ |
|
"epoch": 10.44, |
|
"learning_rate": 8.972950603814755e-06, |
|
"loss": 6.1192, |
|
"step": 556000 |
|
}, |
|
{ |
|
"epoch": 10.45, |
|
"learning_rate": 8.972010082389677e-06, |
|
"loss": 6.1316, |
|
"step": 556500 |
|
}, |
|
{ |
|
"epoch": 10.46, |
|
"learning_rate": 8.9710695609646e-06, |
|
"loss": 6.126, |
|
"step": 557000 |
|
}, |
|
{ |
|
"epoch": 10.47, |
|
"learning_rate": 8.970129039539522e-06, |
|
"loss": 6.1315, |
|
"step": 557500 |
|
}, |
|
{ |
|
"epoch": 10.48, |
|
"learning_rate": 8.969188518114444e-06, |
|
"loss": 6.1253, |
|
"step": 558000 |
|
}, |
|
{ |
|
"epoch": 10.49, |
|
"learning_rate": 8.968247996689366e-06, |
|
"loss": 6.1214, |
|
"step": 558500 |
|
}, |
|
{ |
|
"epoch": 10.5, |
|
"learning_rate": 8.967307475264287e-06, |
|
"loss": 6.1207, |
|
"step": 559000 |
|
}, |
|
{ |
|
"epoch": 10.5, |
|
"learning_rate": 8.966366953839209e-06, |
|
"loss": 6.1214, |
|
"step": 559500 |
|
}, |
|
{ |
|
"epoch": 10.51, |
|
"learning_rate": 8.965426432414131e-06, |
|
"loss": 6.1271, |
|
"step": 560000 |
|
}, |
|
{ |
|
"epoch": 10.52, |
|
"learning_rate": 8.964485910989052e-06, |
|
"loss": 6.1177, |
|
"step": 560500 |
|
}, |
|
{ |
|
"epoch": 10.53, |
|
"learning_rate": 8.963545389563974e-06, |
|
"loss": 6.1274, |
|
"step": 561000 |
|
}, |
|
{ |
|
"epoch": 10.54, |
|
"learning_rate": 8.962604868138896e-06, |
|
"loss": 6.1238, |
|
"step": 561500 |
|
}, |
|
{ |
|
"epoch": 10.55, |
|
"learning_rate": 8.961664346713819e-06, |
|
"loss": 6.1237, |
|
"step": 562000 |
|
}, |
|
{ |
|
"epoch": 10.56, |
|
"learning_rate": 8.96072382528874e-06, |
|
"loss": 6.1235, |
|
"step": 562500 |
|
}, |
|
{ |
|
"epoch": 10.57, |
|
"learning_rate": 8.959783303863663e-06, |
|
"loss": 6.1251, |
|
"step": 563000 |
|
}, |
|
{ |
|
"epoch": 10.58, |
|
"learning_rate": 8.958842782438585e-06, |
|
"loss": 6.1231, |
|
"step": 563500 |
|
}, |
|
{ |
|
"epoch": 10.59, |
|
"learning_rate": 8.957902261013508e-06, |
|
"loss": 6.1189, |
|
"step": 564000 |
|
}, |
|
{ |
|
"epoch": 10.6, |
|
"learning_rate": 8.95696173958843e-06, |
|
"loss": 6.1284, |
|
"step": 564500 |
|
}, |
|
{ |
|
"epoch": 10.61, |
|
"learning_rate": 8.95602121816335e-06, |
|
"loss": 6.127, |
|
"step": 565000 |
|
}, |
|
{ |
|
"epoch": 10.62, |
|
"learning_rate": 8.955080696738273e-06, |
|
"loss": 6.1268, |
|
"step": 565500 |
|
}, |
|
{ |
|
"epoch": 10.63, |
|
"learning_rate": 8.954140175313195e-06, |
|
"loss": 6.1217, |
|
"step": 566000 |
|
}, |
|
{ |
|
"epoch": 10.64, |
|
"learning_rate": 8.953199653888115e-06, |
|
"loss": 6.1239, |
|
"step": 566500 |
|
}, |
|
{ |
|
"epoch": 10.65, |
|
"learning_rate": 8.952259132463038e-06, |
|
"loss": 6.1247, |
|
"step": 567000 |
|
}, |
|
{ |
|
"epoch": 10.65, |
|
"learning_rate": 8.95131861103796e-06, |
|
"loss": 6.1277, |
|
"step": 567500 |
|
}, |
|
{ |
|
"epoch": 10.66, |
|
"learning_rate": 8.950378089612882e-06, |
|
"loss": 6.1222, |
|
"step": 568000 |
|
}, |
|
{ |
|
"epoch": 10.67, |
|
"learning_rate": 8.949437568187804e-06, |
|
"loss": 6.1259, |
|
"step": 568500 |
|
}, |
|
{ |
|
"epoch": 10.68, |
|
"learning_rate": 8.948497046762727e-06, |
|
"loss": 6.1224, |
|
"step": 569000 |
|
}, |
|
{ |
|
"epoch": 10.69, |
|
"step": 569359, |
|
"total_flos": 1.1551410394182451e+17, |
|
"train_loss": 6.316021415674524, |
|
"train_runtime": 197999.0367, |
|
"train_samples_per_second": 2958.983, |
|
"train_steps_per_second": 26.9 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 5326200, |
|
"num_train_epochs": 100, |
|
"save_steps": 1000000, |
|
"total_flos": 1.1551410394182451e+17, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|