JinNian0072/Qwen2.5-Coder-7B-lora-java-abdiff-v2

1{2  "best_global_step": 244,3  "best_metric": 0.241361603140831,4  "best_model_checkpoint": "lora_qwen7b_java_abdiff_v2/checkpoint-244",5  "epoch": 1.0,6  "eval_steps": 500,7  "global_step": 244,8  "is_hyper_param_search": false,9  "is_local_process_zero": true,10  "is_world_process_zero": true,11  "log_history": [12    {13      "epoch": 0.041109969167523124,14      "grad_norm": 0.17571452260017395,15      "learning_rate": 3.648648648648649e-06,16      "loss": 0.5537,17      "step": 1018    },19    {20      "epoch": 0.08221993833504625,21      "grad_norm": 0.12462153285741806,22      "learning_rate": 7.702702702702703e-06,23      "loss": 0.5427,24      "step": 2025    },26    {27      "epoch": 0.12332990750256938,28      "grad_norm": 0.13460873067378998,29      "learning_rate": 1.1756756756756757e-05,30      "loss": 0.5255,31      "step": 3032    },33    {34      "epoch": 0.1644398766700925,35      "grad_norm": 0.2371131032705307,36      "learning_rate": 1.5810810810810808e-05,37      "loss": 0.4849,38      "step": 4039    },40    {41      "epoch": 0.20554984583761562,42      "grad_norm": 0.2611907720565796,43      "learning_rate": 1.9864864864864866e-05,44      "loss": 0.3981,45      "step": 5046    },47    {48      "epoch": 0.24665981500513876,49      "grad_norm": 0.10762301832437515,50      "learning_rate": 2.3918918918918917e-05,51      "loss": 0.3275,52      "step": 6053    },54    {55      "epoch": 0.28776978417266186,56      "grad_norm": 0.06910783052444458,57      "learning_rate": 2.7972972972972975e-05,58      "loss": 0.2976,59      "step": 7060    },61    {62      "epoch": 0.328879753340185,63      "grad_norm": 0.10020674765110016,64      "learning_rate": 2.9772036474164135e-05,65      "loss": 0.2958,66      "step": 8067    },68    {69      "epoch": 0.3699897225077081,70      "grad_norm": 0.06338030099868774,71      "learning_rate": 2.9316109422492404e-05,72      "loss": 0.2775,73      "step": 9074    },75    {76      "epoch": 0.41109969167523125,77      "grad_norm": 0.06356582045555115,78      "learning_rate": 2.886018237082067e-05,79      "loss": 0.2671,80      "step": 10081    },82    {83      "epoch": 0.4522096608427544,84      "grad_norm": 0.06176166608929634,85      "learning_rate": 2.8404255319148935e-05,86      "loss": 0.273,87      "step": 11088    },89    {90      "epoch": 0.4933196300102775,91      "grad_norm": 0.0641980841755867,92      "learning_rate": 2.7948328267477204e-05,93      "loss": 0.2708,94      "step": 12095    },96    {97      "epoch": 0.5344295991778006,98      "grad_norm": 0.060578711330890656,99      "learning_rate": 2.7492401215805473e-05,100      "loss": 0.2492,101      "step": 130102    },103    {104      "epoch": 0.5755395683453237,105      "grad_norm": 0.07500634342432022,106      "learning_rate": 2.7036474164133738e-05,107      "loss": 0.2516,108      "step": 140109    },110    {111      "epoch": 0.6166495375128469,112      "grad_norm": 0.06346568465232849,113      "learning_rate": 2.6580547112462007e-05,114      "loss": 0.2423,115      "step": 150116    },117    {118      "epoch": 0.65775950668037,119      "grad_norm": 0.214496448636055,120      "learning_rate": 2.6124620060790272e-05,121      "loss": 0.245,122      "step": 160123    },124    {125      "epoch": 0.6988694758478932,126      "grad_norm": 0.07425080984830856,127      "learning_rate": 2.566869300911854e-05,128      "loss": 0.2455,129      "step": 170130    },131    {132      "epoch": 0.7399794450154162,133      "grad_norm": 0.08862721920013428,134      "learning_rate": 2.521276595744681e-05,135      "loss": 0.2377,136      "step": 180137    },138    {139      "epoch": 0.7810894141829393,140      "grad_norm": 0.09345080703496933,141      "learning_rate": 2.4756838905775076e-05,142      "loss": 0.2493,143      "step": 190144    },145    {146      "epoch": 0.8221993833504625,147      "grad_norm": 0.12085441499948502,148      "learning_rate": 2.4300911854103345e-05,149      "loss": 0.243,150      "step": 200151    },152    {153      "epoch": 0.8633093525179856,154      "grad_norm": 0.09944415092468262,155      "learning_rate": 2.3844984802431613e-05,156      "loss": 0.2276,157      "step": 210158    },159    {160      "epoch": 0.9044193216855088,161      "grad_norm": 0.11600342392921448,162      "learning_rate": 2.338905775075988e-05,163      "loss": 0.2288,164      "step": 220165    },166    {167      "epoch": 0.9455292908530318,168      "grad_norm": 0.11351948231458664,169      "learning_rate": 2.2933130699088144e-05,170      "loss": 0.2136,171      "step": 230172    },173    {174      "epoch": 0.986639260020555,175      "grad_norm": 0.12985852360725403,176      "learning_rate": 2.2477203647416413e-05,177      "loss": 0.2166,178      "step": 240179    },180    {181      "epoch": 1.0,182      "eval_loss": 0.241361603140831,183      "eval_runtime": 48.3711,184      "eval_samples_per_second": 13.479,185      "eval_steps_per_second": 3.37,186      "step": 244187    }188  ],189  "logging_steps": 10,190  "max_steps": 732,191  "num_input_tokens_seen": 0,192  "num_train_epochs": 3,193  "save_steps": 500,194  "stateful_callbacks": {195    "TrainerControl": {196      "args": {197        "should_epoch_stop": false,198        "should_evaluate": false,199        "should_log": false,200        "should_save": true,201        "should_training_stop": false202      },203      "attributes": {}204    }205  },206  "total_flos": 3.369854600141783e+17,207  "train_batch_size": 4,208  "trial_name": null,209  "trial_params": null210}211