Augusto777's picture
End of training
6b80442 verified
{
"best_metric": 0.10869565217391304,
"best_model_checkpoint": "swinv2-tiny-patch4-window8-256-DMAE-15e\\checkpoint-3",
"epoch": 34.285714285714285,
"eval_steps": 500,
"global_step": 120,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 7.923763275146484,
"eval_runtime": 0.7307,
"eval_samples_per_second": 62.956,
"eval_steps_per_second": 4.106,
"step": 3
},
{
"epoch": 2.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 7.874618053436279,
"eval_runtime": 0.7097,
"eval_samples_per_second": 64.819,
"eval_steps_per_second": 4.227,
"step": 7
},
{
"epoch": 2.86,
"learning_rate": 1.375e-05,
"loss": 7.9273,
"step": 10
},
{
"epoch": 2.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 7.818480014801025,
"eval_runtime": 0.7207,
"eval_samples_per_second": 63.83,
"eval_steps_per_second": 4.163,
"step": 10
},
{
"epoch": 4.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 7.699648857116699,
"eval_runtime": 0.7117,
"eval_samples_per_second": 64.637,
"eval_steps_per_second": 4.215,
"step": 14
},
{
"epoch": 4.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 7.5876078605651855,
"eval_runtime": 0.7162,
"eval_samples_per_second": 64.231,
"eval_steps_per_second": 4.189,
"step": 17
},
{
"epoch": 5.71,
"learning_rate": 1.25e-05,
"loss": 7.6529,
"step": 20
},
{
"epoch": 6.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 7.418884754180908,
"eval_runtime": 0.7127,
"eval_samples_per_second": 64.546,
"eval_steps_per_second": 4.21,
"step": 21
},
{
"epoch": 6.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 7.267844200134277,
"eval_runtime": 0.7112,
"eval_samples_per_second": 64.683,
"eval_steps_per_second": 4.218,
"step": 24
},
{
"epoch": 8.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 7.048880577087402,
"eval_runtime": 0.7262,
"eval_samples_per_second": 63.347,
"eval_steps_per_second": 4.131,
"step": 28
},
{
"epoch": 8.57,
"learning_rate": 1.125e-05,
"loss": 7.1057,
"step": 30
},
{
"epoch": 8.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 6.884642124176025,
"eval_runtime": 0.7182,
"eval_samples_per_second": 64.052,
"eval_steps_per_second": 4.177,
"step": 31
},
{
"epoch": 10.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 6.686830043792725,
"eval_runtime": 0.7417,
"eval_samples_per_second": 62.022,
"eval_steps_per_second": 4.045,
"step": 35
},
{
"epoch": 10.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 6.559473991394043,
"eval_runtime": 0.7132,
"eval_samples_per_second": 64.5,
"eval_steps_per_second": 4.207,
"step": 38
},
{
"epoch": 11.43,
"learning_rate": 9.999999999999999e-06,
"loss": 6.8483,
"step": 40
},
{
"epoch": 12.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 6.38260555267334,
"eval_runtime": 0.7272,
"eval_samples_per_second": 63.258,
"eval_steps_per_second": 4.126,
"step": 42
},
{
"epoch": 12.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 6.227572917938232,
"eval_runtime": 0.7182,
"eval_samples_per_second": 64.051,
"eval_steps_per_second": 4.177,
"step": 45
},
{
"epoch": 14.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 6.03657341003418,
"eval_runtime": 0.7132,
"eval_samples_per_second": 64.5,
"eval_steps_per_second": 4.207,
"step": 49
},
{
"epoch": 14.29,
"learning_rate": 8.750000000000001e-06,
"loss": 6.224,
"step": 50
},
{
"epoch": 14.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.904354095458984,
"eval_runtime": 0.7787,
"eval_samples_per_second": 59.074,
"eval_steps_per_second": 3.853,
"step": 52
},
{
"epoch": 16.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.738274574279785,
"eval_runtime": 0.7132,
"eval_samples_per_second": 64.502,
"eval_steps_per_second": 4.207,
"step": 56
},
{
"epoch": 16.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.626614570617676,
"eval_runtime": 0.7317,
"eval_samples_per_second": 62.87,
"eval_steps_per_second": 4.1,
"step": 59
},
{
"epoch": 17.14,
"learning_rate": 7.5e-06,
"loss": 5.8234,
"step": 60
},
{
"epoch": 18.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.487057209014893,
"eval_runtime": 0.7047,
"eval_samples_per_second": 65.279,
"eval_steps_per_second": 4.257,
"step": 63
},
{
"epoch": 18.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.389139652252197,
"eval_runtime": 0.7212,
"eval_samples_per_second": 63.786,
"eval_steps_per_second": 4.16,
"step": 66
},
{
"epoch": 20.0,
"learning_rate": 6.25e-06,
"loss": 5.5423,
"step": 70
},
{
"epoch": 20.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.267205238342285,
"eval_runtime": 0.7192,
"eval_samples_per_second": 63.963,
"eval_steps_per_second": 4.172,
"step": 70
},
{
"epoch": 20.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.180857181549072,
"eval_runtime": 0.7337,
"eval_samples_per_second": 62.698,
"eval_steps_per_second": 4.089,
"step": 73
},
{
"epoch": 22.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.0741496086120605,
"eval_runtime": 0.7262,
"eval_samples_per_second": 63.347,
"eval_steps_per_second": 4.131,
"step": 77
},
{
"epoch": 22.86,
"learning_rate": 4.9999999999999996e-06,
"loss": 5.2547,
"step": 80
},
{
"epoch": 22.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 5.000693321228027,
"eval_runtime": 0.6972,
"eval_samples_per_second": 65.982,
"eval_steps_per_second": 4.303,
"step": 80
},
{
"epoch": 24.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.911642551422119,
"eval_runtime": 0.7242,
"eval_samples_per_second": 63.522,
"eval_steps_per_second": 4.143,
"step": 84
},
{
"epoch": 24.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.850471019744873,
"eval_runtime": 0.7197,
"eval_samples_per_second": 63.918,
"eval_steps_per_second": 4.169,
"step": 87
},
{
"epoch": 25.71,
"learning_rate": 3.75e-06,
"loss": 5.0166,
"step": 90
},
{
"epoch": 26.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.776952266693115,
"eval_runtime": 0.7807,
"eval_samples_per_second": 58.923,
"eval_steps_per_second": 3.843,
"step": 91
},
{
"epoch": 26.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.728125095367432,
"eval_runtime": 0.7177,
"eval_samples_per_second": 64.096,
"eval_steps_per_second": 4.18,
"step": 94
},
{
"epoch": 28.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.6712117195129395,
"eval_runtime": 0.7762,
"eval_samples_per_second": 59.264,
"eval_steps_per_second": 3.865,
"step": 98
},
{
"epoch": 28.57,
"learning_rate": 2.4999999999999998e-06,
"loss": 4.7751,
"step": 100
},
{
"epoch": 28.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.634757041931152,
"eval_runtime": 0.7157,
"eval_samples_per_second": 64.275,
"eval_steps_per_second": 4.192,
"step": 101
},
{
"epoch": 30.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.594343185424805,
"eval_runtime": 0.7407,
"eval_samples_per_second": 62.106,
"eval_steps_per_second": 4.05,
"step": 105
},
{
"epoch": 30.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.570145130157471,
"eval_runtime": 0.7192,
"eval_samples_per_second": 63.963,
"eval_steps_per_second": 4.172,
"step": 108
},
{
"epoch": 31.43,
"learning_rate": 1.2499999999999999e-06,
"loss": 4.7321,
"step": 110
},
{
"epoch": 32.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.545791149139404,
"eval_runtime": 0.7197,
"eval_samples_per_second": 63.918,
"eval_steps_per_second": 4.169,
"step": 112
},
{
"epoch": 32.86,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.533587455749512,
"eval_runtime": 0.7117,
"eval_samples_per_second": 64.637,
"eval_steps_per_second": 4.215,
"step": 115
},
{
"epoch": 34.0,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.525488376617432,
"eval_runtime": 0.7257,
"eval_samples_per_second": 63.39,
"eval_steps_per_second": 4.134,
"step": 119
},
{
"epoch": 34.29,
"learning_rate": 0.0,
"loss": 4.6731,
"step": 120
},
{
"epoch": 34.29,
"eval_accuracy": 0.10869565217391304,
"eval_loss": 4.524907112121582,
"eval_runtime": 0.7682,
"eval_samples_per_second": 59.882,
"eval_steps_per_second": 3.905,
"step": 120
},
{
"epoch": 34.29,
"step": 120,
"total_flos": 2.3770905934823424e+17,
"train_loss": 5.964630603790283,
"train_runtime": 190.6471,
"train_samples_per_second": 44.69,
"train_steps_per_second": 0.629
}
],
"logging_steps": 10,
"max_steps": 120,
"num_input_tokens_seen": 0,
"num_train_epochs": 40,
"save_steps": 500,
"total_flos": 2.3770905934823424e+17,
"train_batch_size": 16,
"trial_name": null,
"trial_params": null
}