|
{ |
|
"best_metric": 0.8789323285321096, |
|
"best_model_checkpoint": "OS_OUT/harish/PT-UP-xlmR-FalseTrue-0_0_BEST/FalseTrue-0/2/checkpoint-18", |
|
"epoch": 9.0, |
|
"global_step": 18, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.6336996555328369, |
|
"eval_f1": 0.6278853421710564, |
|
"eval_loss": 1.0849474668502808, |
|
"eval_runtime": 0.7985, |
|
"eval_samples_per_second": 341.901, |
|
"eval_steps_per_second": 43.834, |
|
"step": 2 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.6996337175369263, |
|
"eval_f1": 0.7003429567212002, |
|
"eval_loss": 0.5072219371795654, |
|
"eval_runtime": 0.7903, |
|
"eval_samples_per_second": 345.433, |
|
"eval_steps_per_second": 44.286, |
|
"step": 4 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.7655677795410156, |
|
"eval_f1": 0.7507642323997964, |
|
"eval_loss": 0.47925040125846863, |
|
"eval_runtime": 0.7908, |
|
"eval_samples_per_second": 345.204, |
|
"eval_steps_per_second": 44.257, |
|
"step": 6 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.8131868243217468, |
|
"eval_f1": 0.8096849683022662, |
|
"eval_loss": 0.45054349303245544, |
|
"eval_runtime": 0.791, |
|
"eval_samples_per_second": 345.149, |
|
"eval_steps_per_second": 44.25, |
|
"step": 8 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_accuracy": 0.8351648449897766, |
|
"eval_f1": 0.8353801154290916, |
|
"eval_loss": 0.4265427887439728, |
|
"eval_runtime": 0.7901, |
|
"eval_samples_per_second": 345.511, |
|
"eval_steps_per_second": 44.296, |
|
"step": 10 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"eval_accuracy": 0.8571428656578064, |
|
"eval_f1": 0.8573294333718795, |
|
"eval_loss": 0.38742759823799133, |
|
"eval_runtime": 0.7896, |
|
"eval_samples_per_second": 345.734, |
|
"eval_steps_per_second": 44.325, |
|
"step": 12 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"eval_accuracy": 0.8644688725471497, |
|
"eval_f1": 0.8640971847926557, |
|
"eval_loss": 0.35351312160491943, |
|
"eval_runtime": 0.7926, |
|
"eval_samples_per_second": 344.433, |
|
"eval_steps_per_second": 44.158, |
|
"step": 14 |
|
}, |
|
{ |
|
"epoch": 8.0, |
|
"eval_accuracy": 0.8754578828811646, |
|
"eval_f1": 0.8753319033800172, |
|
"eval_loss": 0.33853575587272644, |
|
"eval_runtime": 0.7916, |
|
"eval_samples_per_second": 344.883, |
|
"eval_steps_per_second": 44.216, |
|
"step": 16 |
|
}, |
|
{ |
|
"epoch": 9.0, |
|
"eval_accuracy": 0.8791208863258362, |
|
"eval_f1": 0.8789323285321096, |
|
"eval_loss": 0.3337554633617401, |
|
"eval_runtime": 0.7905, |
|
"eval_samples_per_second": 345.343, |
|
"eval_steps_per_second": 44.275, |
|
"step": 18 |
|
} |
|
], |
|
"max_steps": 18, |
|
"num_train_epochs": 9, |
|
"total_flos": 101857961258496.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|