Team Ai
Modelpublic

UI-Simulator/UI_Simulator_F_Android

sourceHugging Facemitupdated 1y agoView on Hugging Face
1likes23downloads
trainer_state.json317 linesDownload Raw Back to root
1{2  "best_global_step": null,3  "best_metric": null,4  "best_model_checkpoint": null,5  "epoch": 1.0,6  "eval_steps": 500,7  "global_step": 138,8  "is_hyper_param_search": false,9  "is_local_process_zero": true,10  "is_world_process_zero": true,11  "log_history": [12    {13      "entropy": 0.39333338141441343,14      "epoch": 0.03640776699029126,15      "grad_norm": 5.541357291512188,16      "learning_rate": 2.8571428571428573e-06,17      "loss": 0.7481,18      "mean_token_accuracy": 0.8530021011829376,19      "num_tokens": 634119.0,20      "step": 521    },22    {23      "entropy": 0.46669652859369914,24      "epoch": 0.07281553398058252,25      "grad_norm": 3.55874761602301,26      "learning_rate": 6.4285714285714295e-06,27      "loss": 0.4894,28      "mean_token_accuracy": 0.8847795287768047,29      "num_tokens": 1273023.0,30      "step": 1031    },32    {33      "entropy": 0.2333444893360138,34      "epoch": 0.10922330097087378,35      "grad_norm": 0.683356185791462,36      "learning_rate": 1e-05,37      "loss": 0.2045,38      "mean_token_accuracy": 0.9458525717258454,39      "num_tokens": 1901638.0,40      "step": 1541    },42    {43      "entropy": 0.14950522979100545,44      "epoch": 0.14563106796116504,45      "grad_norm": 0.4345966321170585,46      "learning_rate": 9.959935885253715e-06,47      "loss": 0.1443,48      "mean_token_accuracy": 0.9594493408997854,49      "num_tokens": 2543415.0,50      "step": 2051    },52    {53      "entropy": 0.1172327217956384,54      "epoch": 0.1820388349514563,55      "grad_norm": 0.4083364127472746,56      "learning_rate": 9.840385594331022e-06,57      "loss": 0.1135,58      "mean_token_accuracy": 0.9674430509408315,59      "num_tokens": 3177702.0,60      "step": 2561    },62    {63      "entropy": 0.10821170260508856,64      "epoch": 0.21844660194174756,65      "grad_norm": 0.3274112578074671,66      "learning_rate": 9.643264997861312e-06,67      "loss": 0.1053,68      "mean_token_accuracy": 0.969423896074295,69      "num_tokens": 3817129.0,70      "step": 3071    },72    {73      "entropy": 0.10363089988629023,74      "epoch": 0.25485436893203883,75      "grad_norm": 0.2651750049561702,76      "learning_rate": 9.371733080722911e-06,77      "loss": 0.101,78      "mean_token_accuracy": 0.9701913992563883,79      "num_tokens": 4450713.0,80      "step": 3581    },82    {83      "entropy": 0.09188476577401161,84      "epoch": 0.2912621359223301,85      "grad_norm": 0.3322177757328344,86      "learning_rate": 9.030141317270026e-06,87      "loss": 0.0896,88      "mean_token_accuracy": 0.9733767688274384,89      "num_tokens": 5099094.0,90      "step": 4091    },92    {93      "entropy": 0.09189487385253112,94      "epoch": 0.3276699029126214,95      "grad_norm": 0.27499471047406965,96      "learning_rate": 8.6239639361456e-06,97      "loss": 0.0892,98      "mean_token_accuracy": 0.9734424968560537,99      "num_tokens": 5746298.0,100      "step": 45101    },102    {103      "entropy": 0.09149620061119397,104      "epoch": 0.3640776699029126,105      "grad_norm": 0.2426196466051672,106      "learning_rate": 8.15971019223152e-06,107      "loss": 0.0888,108      "mean_token_accuracy": 0.9731526096661886,109      "num_tokens": 6396438.0,110      "step": 50111    },112    {113      "entropy": 0.08535861944158872,114      "epoch": 0.40048543689320387,115      "grad_norm": 0.25072345756012987,116      "learning_rate": 7.644820051634813e-06,117      "loss": 0.0833,118      "mean_token_accuracy": 0.9750010808308919,119      "num_tokens": 7045524.0,120      "step": 55121    },122    {123      "entropy": 0.08259956675271193,124      "epoch": 0.4368932038834951,125      "grad_norm": 0.2328355084383256,126      "learning_rate": 7.087544961425317e-06,127      "loss": 0.079,128      "mean_token_accuracy": 0.9756735742092133,129      "num_tokens": 7696316.0,130      "step": 60131    },132    {133      "entropy": 0.08420657614866893,134      "epoch": 0.4733009708737864,135      "grad_norm": 0.2627786396905739,136      "learning_rate": 6.496815614866792e-06,137      "loss": 0.0834,138      "mean_token_accuracy": 0.9745731969674428,139      "num_tokens": 8335632.0,140      "step": 65141    },142    {143      "entropy": 0.07922674703101316,144      "epoch": 0.5097087378640777,145      "grad_norm": 0.23198033388844116,146      "learning_rate": 5.882098831289044e-06,147      "loss": 0.0764,148      "mean_token_accuracy": 0.9763782580693563,149      "num_tokens": 8980788.0,150      "step": 70151    },152    {153      "entropy": 0.08260756817956766,154      "epoch": 0.5461165048543689,155      "grad_norm": 0.23246807001208797,156      "learning_rate": 5.253245844193564e-06,157      "loss": 0.0784,158      "mean_token_accuracy": 0.975836310784022,159      "num_tokens": 9615133.0,160      "step": 75161    },162    {163      "entropy": 0.0775965532908837,164      "epoch": 0.5825242718446602,165      "grad_norm": 0.23509755618479103,166      "learning_rate": 4.62033442887377e-06,167      "loss": 0.075,168      "mean_token_accuracy": 0.9768483221530915,169      "num_tokens": 10272323.0,170      "step": 80171    },172    {173      "entropy": 0.0796258964886268,174      "epoch": 0.6189320388349514,175      "grad_norm": 0.2398016429212428,176      "learning_rate": 3.993507399556699e-06,177      "loss": 0.0769,178      "mean_token_accuracy": 0.9760827680428823,179      "num_tokens": 10909479.0,180      "step": 85181    },182    {183      "entropy": 0.08220818725725015,184      "epoch": 0.6553398058252428,185      "grad_norm": 0.22572998880022216,186      "learning_rate": 3.3828100642538097e-06,187      "loss": 0.0779,188      "mean_token_accuracy": 0.9757848223050435,189      "num_tokens": 11539056.0,190      "step": 90191    },192    {193      "entropy": 0.07774231346944968,194      "epoch": 0.691747572815534,195      "grad_norm": 0.22123308445803502,196      "learning_rate": 2.7980292422118282e-06,197      "loss": 0.0739,198      "mean_token_accuracy": 0.9770070095856984,199      "num_tokens": 12175621.0,200      "step": 95201    },202    {203      "entropy": 0.07790198549628258,204      "epoch": 0.7281553398058253,205      "grad_norm": 0.22549132644894587,206      "learning_rate": 2.2485364238130435e-06,207      "loss": 0.0743,208      "mean_token_accuracy": 0.9770615140597025,209      "num_tokens": 12820187.0,210      "step": 100211    },212    {213      "entropy": 0.07972427035371463,214      "epoch": 0.7645631067961165,215      "grad_norm": 0.22395855789852193,216      "learning_rate": 1.74313758638889e-06,217      "loss": 0.0753,218      "mean_token_accuracy": 0.9765656272570292,219      "num_tokens": 13471214.0,220      "step": 105221    },222    {223      "entropy": 0.07861521604160468,224      "epoch": 0.8009708737864077,225      "grad_norm": 0.21707494419816833,226      "learning_rate": 1.2899320727454472e-06,227      "loss": 0.0758,228      "mean_token_accuracy": 0.9762462894121806,229      "num_tokens": 14104842.0,230      "step": 110231    },232    {233      "entropy": 0.07681530776123206,234      "epoch": 0.837378640776699,235      "grad_norm": 0.2048280957572369,236      "learning_rate": 8.961827939636198e-07,237      "loss": 0.0752,238      "mean_token_accuracy": 0.9767355759938557,239      "num_tokens": 14754145.0,240      "step": 115241    },242    {243      "entropy": 0.07755378099779288,244      "epoch": 0.8737864077669902,245      "grad_norm": 0.21592550407174824,246      "learning_rate": 5.681998365579594e-07,247      "loss": 0.0743,248      "mean_token_accuracy": 0.9771900673707327,249      "num_tokens": 15399256.0,250      "step": 120251    },252    {253      "entropy": 0.07576616617540519,254      "epoch": 0.9101941747572816,255      "grad_norm": 0.19062427853278816,256      "learning_rate": 3.112393392645985e-07,257      "loss": 0.0727,258      "mean_token_accuracy": 0.9775316794713338,259      "num_tokens": 16043949.0,260      "step": 125261    },262    {263      "entropy": 0.07121096017460028,264      "epoch": 0.9466019417475728,265      "grad_norm": 0.20849501542469207,266      "learning_rate": 1.2941926002306536e-07,267      "loss": 0.068,268      "mean_token_accuracy": 0.9787502566973368,269      "num_tokens": 16696773.0,270      "step": 130271    },272    {273      "entropy": 0.07767875120043755,274      "epoch": 0.9830097087378641,275      "grad_norm": 0.19245780653274294,276      "learning_rate": 2.5653383040524228e-08,277      "loss": 0.0731,278      "mean_token_accuracy": 0.9773090521494547,279      "num_tokens": 17351570.0,280      "step": 135281    },282    {283      "entropy": 0.07084919246179718,284      "epoch": 1.0,285      "mean_token_accuracy": 0.9790157548018864,286      "num_tokens": 17657223.0,287      "step": 138,288      "total_flos": 39414093316096.0,289      "train_loss": 0.1270361747862636,290      "train_runtime": 1974.4991,291      "train_samples_per_second": 3.338,292      "train_steps_per_second": 0.07293    }294  ],295  "logging_steps": 5,296  "max_steps": 138,297  "num_input_tokens_seen": 0,298  "num_train_epochs": 1,299  "save_steps": 500,300  "stateful_callbacks": {301    "TrainerControl": {302      "args": {303        "should_epoch_stop": false,304        "should_evaluate": false,305        "should_log": false,306        "should_save": true,307        "should_training_stop": true308      },309      "attributes": {}310    }311  },312  "total_flos": 39414093316096.0,313  "train_batch_size": 2,314  "trial_name": null,315  "trial_params": null316}317