yeray142's picture
End of training
22c77b5 verified
{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 9.9697933227345,
"eval_steps": 500,
"global_step": 3140,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.0,
"grad_norm": 112.57828521728516,
"learning_rate": 1e-05,
"loss": 32.1236,
"step": 315
},
{
"epoch": 1.0,
"eval_loss": 25.001630783081055,
"eval_map": 0.172,
"eval_map_50": 0.3235,
"eval_map_75": 0.1677,
"eval_map_car": 0.3376,
"eval_map_large": 0.8944,
"eval_map_medium": 0.3483,
"eval_map_person": 0.0065,
"eval_map_small": 0.0881,
"eval_mar_1": 0.0877,
"eval_mar_10": 0.2218,
"eval_mar_100": 0.2617,
"eval_mar_100_car": 0.4775,
"eval_mar_100_person": 0.0459,
"eval_mar_large": 0.9347,
"eval_mar_medium": 0.4549,
"eval_mar_small": 0.1935,
"eval_runtime": 2301.478,
"eval_samples_per_second": 1.295,
"eval_steps_per_second": 0.162,
"step": 315
},
{
"epoch": 2.0,
"grad_norm": 244.69192504882812,
"learning_rate": 1e-05,
"loss": 25.8006,
"step": 630
},
{
"epoch": 2.0,
"eval_loss": 23.8843936920166,
"eval_map": 0.1836,
"eval_map_50": 0.3505,
"eval_map_75": 0.1708,
"eval_map_car": 0.3541,
"eval_map_large": 0.9132,
"eval_map_medium": 0.3691,
"eval_map_person": 0.0131,
"eval_map_small": 0.0973,
"eval_mar_1": 0.09,
"eval_mar_10": 0.2324,
"eval_mar_100": 0.276,
"eval_mar_100_car": 0.4901,
"eval_mar_100_person": 0.0619,
"eval_mar_large": 0.9434,
"eval_mar_medium": 0.4765,
"eval_mar_small": 0.2074,
"eval_runtime": 2182.9105,
"eval_samples_per_second": 1.366,
"eval_steps_per_second": 0.171,
"step": 630
},
{
"epoch": 3.0,
"grad_norm": 116.00335693359375,
"learning_rate": 1e-05,
"loss": 24.098,
"step": 945
},
{
"epoch": 3.0,
"eval_loss": 23.18219757080078,
"eval_map": 0.1892,
"eval_map_50": 0.3583,
"eval_map_75": 0.1751,
"eval_map_car": 0.3616,
"eval_map_large": 0.9297,
"eval_map_medium": 0.3823,
"eval_map_person": 0.0168,
"eval_map_small": 0.0975,
"eval_mar_1": 0.0904,
"eval_mar_10": 0.2392,
"eval_mar_100": 0.283,
"eval_mar_100_car": 0.4967,
"eval_mar_100_person": 0.0693,
"eval_mar_large": 0.9495,
"eval_mar_medium": 0.4814,
"eval_mar_small": 0.215,
"eval_runtime": 2092.4247,
"eval_samples_per_second": 1.425,
"eval_steps_per_second": 0.178,
"step": 945
},
{
"epoch": 4.0,
"grad_norm": 203.1825714111328,
"learning_rate": 1e-05,
"loss": 23.0237,
"step": 1260
},
{
"epoch": 4.0,
"eval_loss": 22.712690353393555,
"eval_map": 0.1913,
"eval_map_50": 0.3692,
"eval_map_75": 0.1751,
"eval_map_car": 0.3635,
"eval_map_large": 0.9289,
"eval_map_medium": 0.3846,
"eval_map_person": 0.0191,
"eval_map_small": 0.1017,
"eval_mar_1": 0.0933,
"eval_mar_10": 0.2437,
"eval_mar_100": 0.289,
"eval_mar_100_car": 0.5003,
"eval_mar_100_person": 0.0778,
"eval_mar_large": 0.9486,
"eval_mar_medium": 0.4827,
"eval_mar_small": 0.2225,
"eval_runtime": 2068.7659,
"eval_samples_per_second": 1.441,
"eval_steps_per_second": 0.18,
"step": 1260
},
{
"epoch": 5.0,
"grad_norm": 295.3192443847656,
"learning_rate": 1e-05,
"loss": 22.25,
"step": 1575
},
{
"epoch": 5.0,
"eval_loss": 22.591793060302734,
"eval_map": 0.1933,
"eval_map_50": 0.3765,
"eval_map_75": 0.1754,
"eval_map_car": 0.3648,
"eval_map_large": 0.9267,
"eval_map_medium": 0.3951,
"eval_map_person": 0.0218,
"eval_map_small": 0.1053,
"eval_mar_1": 0.0934,
"eval_mar_10": 0.2477,
"eval_mar_100": 0.2916,
"eval_mar_100_car": 0.5,
"eval_mar_100_person": 0.0832,
"eval_mar_large": 0.9474,
"eval_mar_medium": 0.4829,
"eval_mar_small": 0.2253,
"eval_runtime": 2102.8614,
"eval_samples_per_second": 1.418,
"eval_steps_per_second": 0.177,
"step": 1575
},
{
"epoch": 6.0,
"grad_norm": 278.1381530761719,
"learning_rate": 1e-05,
"loss": 21.7056,
"step": 1890
},
{
"epoch": 6.0,
"eval_loss": 21.966611862182617,
"eval_map": 0.2019,
"eval_map_50": 0.3913,
"eval_map_75": 0.1833,
"eval_map_car": 0.3775,
"eval_map_large": 0.9311,
"eval_map_medium": 0.4037,
"eval_map_person": 0.0263,
"eval_map_small": 0.1101,
"eval_mar_1": 0.0965,
"eval_mar_10": 0.256,
"eval_mar_100": 0.2998,
"eval_mar_100_car": 0.5145,
"eval_mar_100_person": 0.0852,
"eval_mar_large": 0.9497,
"eval_mar_medium": 0.4911,
"eval_mar_small": 0.235,
"eval_runtime": 2206.8239,
"eval_samples_per_second": 1.351,
"eval_steps_per_second": 0.169,
"step": 1890
},
{
"epoch": 7.0,
"grad_norm": 256.54266357421875,
"learning_rate": 1e-05,
"loss": 21.218,
"step": 2205
},
{
"epoch": 7.0,
"eval_loss": 22.137611389160156,
"eval_map": 0.2002,
"eval_map_50": 0.3859,
"eval_map_75": 0.1841,
"eval_map_car": 0.3751,
"eval_map_large": 0.9299,
"eval_map_medium": 0.412,
"eval_map_person": 0.0254,
"eval_map_small": 0.1087,
"eval_mar_1": 0.0974,
"eval_mar_10": 0.255,
"eval_mar_100": 0.3003,
"eval_mar_100_car": 0.5113,
"eval_mar_100_person": 0.0892,
"eval_mar_large": 0.9524,
"eval_mar_medium": 0.5004,
"eval_mar_small": 0.2331,
"eval_runtime": 2224.5924,
"eval_samples_per_second": 1.34,
"eval_steps_per_second": 0.168,
"step": 2205
},
{
"epoch": 8.0,
"grad_norm": 234.42987060546875,
"learning_rate": 1e-05,
"loss": 20.7151,
"step": 2520
},
{
"epoch": 8.0,
"eval_loss": 21.743112564086914,
"eval_map": 0.2013,
"eval_map_50": 0.3953,
"eval_map_75": 0.1819,
"eval_map_car": 0.375,
"eval_map_large": 0.9349,
"eval_map_medium": 0.411,
"eval_map_person": 0.0277,
"eval_map_small": 0.1105,
"eval_mar_1": 0.0973,
"eval_mar_10": 0.2595,
"eval_mar_100": 0.3059,
"eval_mar_100_car": 0.5178,
"eval_mar_100_person": 0.094,
"eval_mar_large": 0.9533,
"eval_mar_medium": 0.5016,
"eval_mar_small": 0.2401,
"eval_runtime": 2188.5922,
"eval_samples_per_second": 1.362,
"eval_steps_per_second": 0.17,
"step": 2520
},
{
"epoch": 9.0,
"grad_norm": 200.98818969726562,
"learning_rate": 1e-05,
"loss": 20.4197,
"step": 2835
},
{
"epoch": 9.0,
"eval_loss": 21.8546199798584,
"eval_map": 0.2024,
"eval_map_50": 0.3925,
"eval_map_75": 0.184,
"eval_map_car": 0.3781,
"eval_map_large": 0.9325,
"eval_map_medium": 0.4136,
"eval_map_person": 0.0267,
"eval_map_small": 0.1112,
"eval_mar_1": 0.0971,
"eval_mar_10": 0.2589,
"eval_mar_100": 0.3044,
"eval_mar_100_car": 0.5164,
"eval_mar_100_person": 0.0925,
"eval_mar_large": 0.9531,
"eval_mar_medium": 0.4965,
"eval_mar_small": 0.2387,
"eval_runtime": 2264.5474,
"eval_samples_per_second": 1.316,
"eval_steps_per_second": 0.165,
"step": 2835
},
{
"epoch": 9.9697933227345,
"grad_norm": 1650.528076171875,
"learning_rate": 1e-05,
"loss": 20.1339,
"step": 3140
},
{
"epoch": 9.9697933227345,
"eval_loss": 21.84906578063965,
"eval_map": 0.2024,
"eval_map_50": 0.3976,
"eval_map_75": 0.1846,
"eval_map_car": 0.3761,
"eval_map_large": 0.9371,
"eval_map_medium": 0.4171,
"eval_map_person": 0.0288,
"eval_map_small": 0.1131,
"eval_mar_1": 0.098,
"eval_mar_10": 0.2621,
"eval_mar_100": 0.3113,
"eval_mar_100_car": 0.5206,
"eval_mar_100_person": 0.102,
"eval_mar_large": 0.9545,
"eval_mar_medium": 0.5068,
"eval_mar_small": 0.2456,
"eval_runtime": 2315.7847,
"eval_samples_per_second": 1.287,
"eval_steps_per_second": 0.161,
"step": 3140
},
{
"epoch": 9.9697933227345,
"step": 3140,
"total_flos": 5.249371936187105e+18,
"train_loss": 23.158433114798964,
"train_runtime": 31718.6696,
"train_samples_per_second": 1.585,
"train_steps_per_second": 0.099
}
],
"logging_steps": 500,
"max_steps": 3140,
"num_input_tokens_seen": 0,
"num_train_epochs": 10,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 5.249371936187105e+18,
"train_batch_size": 8,
"trial_name": null,
"trial_params": null
}