vit-artworkclassifier / trainer_state.json
oschamp's picture
End of training
a839692
raw
history blame
2.57 kB
{
"best_metric": 1.3362700939178467,
"best_model_checkpoint": "./vit-artworkclassifier\\checkpoint-200",
"epoch": 8.0,
"global_step": 448,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.89,
"learning_rate": 8.883928571428571e-05,
"loss": 1.9187,
"step": 50
},
{
"epoch": 1.79,
"learning_rate": 7.767857142857144e-05,
"loss": 1.4136,
"step": 100
},
{
"epoch": 1.79,
"eval_accuracy": 0.5112359550561798,
"eval_loss": 1.5092945098876953,
"eval_runtime": 1.5015,
"eval_samples_per_second": 118.548,
"eval_steps_per_second": 15.318,
"step": 100
},
{
"epoch": 2.68,
"learning_rate": 6.651785714285714e-05,
"loss": 1.0465,
"step": 150
},
{
"epoch": 3.57,
"learning_rate": 5.535714285714286e-05,
"loss": 0.7189,
"step": 200
},
{
"epoch": 3.57,
"eval_accuracy": 0.4887640449438202,
"eval_loss": 1.3362700939178467,
"eval_runtime": 1.475,
"eval_samples_per_second": 120.678,
"eval_steps_per_second": 15.593,
"step": 200
},
{
"epoch": 4.46,
"learning_rate": 4.419642857142857e-05,
"loss": 0.4243,
"step": 250
},
{
"epoch": 5.36,
"learning_rate": 3.303571428571429e-05,
"loss": 0.2717,
"step": 300
},
{
"epoch": 5.36,
"eval_accuracy": 0.5280898876404494,
"eval_loss": 1.490658164024353,
"eval_runtime": 1.4785,
"eval_samples_per_second": 120.393,
"eval_steps_per_second": 15.556,
"step": 300
},
{
"epoch": 6.25,
"learning_rate": 2.1875e-05,
"loss": 0.184,
"step": 350
},
{
"epoch": 7.14,
"learning_rate": 1.0714285714285714e-05,
"loss": 0.1227,
"step": 400
},
{
"epoch": 7.14,
"eval_accuracy": 0.5561797752808989,
"eval_loss": 1.4826292991638184,
"eval_runtime": 1.463,
"eval_samples_per_second": 121.668,
"eval_steps_per_second": 15.721,
"step": 400
},
{
"epoch": 8.0,
"step": 448,
"total_flos": 1.0992153398028042e+18,
"train_loss": 0.6919720236744199,
"train_runtime": 221.2781,
"train_samples_per_second": 64.1,
"train_steps_per_second": 2.025
}
],
"max_steps": 448,
"num_train_epochs": 8,
"total_flos": 1.0992153398028042e+18,
"trial_name": null,
"trial_params": null
}