backup_luganda_150 / trainer_state.json
azawahry
Initial commit
9fa26d2
{
"best_metric": 2.347782850265503,
"best_model_checkpoint": "/home/azawahry_sunbird_ai/models/m2e/mbart-luganda-peft-1681982976/checkpoint-150",
"epoch": 0.8625051750310502,
"global_step": 150,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.06,
"eval_BLEU_lug": 36.0264,
"eval_BLEU_mean": 36.0264,
"eval_loss": 2.8559389114379883,
"eval_runtime": 47.4669,
"eval_samples_per_second": 10.534,
"eval_steps_per_second": 1.327,
"step": 10
},
{
"epoch": 0.12,
"eval_BLEU_lug": 37.8444,
"eval_BLEU_mean": 37.8444,
"eval_loss": 2.4016377925872803,
"eval_runtime": 47.7157,
"eval_samples_per_second": 10.479,
"eval_steps_per_second": 1.32,
"step": 20
},
{
"epoch": 0.17,
"eval_BLEU_lug": 38.246,
"eval_BLEU_mean": 38.246,
"eval_loss": 2.3729116916656494,
"eval_runtime": 47.5616,
"eval_samples_per_second": 10.513,
"eval_steps_per_second": 1.325,
"step": 30
},
{
"epoch": 0.23,
"eval_BLEU_lug": 38.4637,
"eval_BLEU_mean": 38.4637,
"eval_loss": 2.367535352706909,
"eval_runtime": 47.7816,
"eval_samples_per_second": 10.464,
"eval_steps_per_second": 1.318,
"step": 40
},
{
"epoch": 0.29,
"eval_BLEU_lug": 38.4294,
"eval_BLEU_mean": 38.4294,
"eval_loss": 2.3612842559814453,
"eval_runtime": 48.137,
"eval_samples_per_second": 10.387,
"eval_steps_per_second": 1.309,
"step": 50
},
{
"epoch": 0.35,
"eval_BLEU_lug": 39.2099,
"eval_BLEU_mean": 39.2099,
"eval_loss": 2.3571784496307373,
"eval_runtime": 46.703,
"eval_samples_per_second": 10.706,
"eval_steps_per_second": 1.349,
"step": 60
},
{
"epoch": 0.4,
"eval_BLEU_lug": 38.6832,
"eval_BLEU_mean": 38.6832,
"eval_loss": 2.35516095161438,
"eval_runtime": 47.312,
"eval_samples_per_second": 10.568,
"eval_steps_per_second": 1.332,
"step": 70
},
{
"epoch": 0.46,
"eval_BLEU_lug": 39.1361,
"eval_BLEU_mean": 39.1361,
"eval_loss": 2.353118896484375,
"eval_runtime": 46.6434,
"eval_samples_per_second": 10.72,
"eval_steps_per_second": 1.351,
"step": 80
},
{
"epoch": 0.52,
"eval_BLEU_lug": 38.9296,
"eval_BLEU_mean": 38.9296,
"eval_loss": 2.352966785430908,
"eval_runtime": 46.5171,
"eval_samples_per_second": 10.749,
"eval_steps_per_second": 1.354,
"step": 90
},
{
"epoch": 0.58,
"eval_BLEU_lug": 39.1036,
"eval_BLEU_mean": 39.1036,
"eval_loss": 2.35172438621521,
"eval_runtime": 46.318,
"eval_samples_per_second": 10.795,
"eval_steps_per_second": 1.36,
"step": 100
},
{
"epoch": 0.63,
"eval_BLEU_lug": 39.2866,
"eval_BLEU_mean": 39.2866,
"eval_loss": 2.349985361099243,
"eval_runtime": 46.4059,
"eval_samples_per_second": 10.774,
"eval_steps_per_second": 1.358,
"step": 110
},
{
"epoch": 0.69,
"eval_BLEU_lug": 39.2961,
"eval_BLEU_mean": 39.2961,
"eval_loss": 2.348721742630005,
"eval_runtime": 46.3187,
"eval_samples_per_second": 10.795,
"eval_steps_per_second": 1.36,
"step": 120
},
{
"epoch": 0.75,
"eval_BLEU_lug": 39.2138,
"eval_BLEU_mean": 39.2138,
"eval_loss": 2.3493213653564453,
"eval_runtime": 46.9837,
"eval_samples_per_second": 10.642,
"eval_steps_per_second": 1.341,
"step": 130
},
{
"epoch": 0.81,
"eval_BLEU_lug": 39.2112,
"eval_BLEU_mean": 39.2112,
"eval_loss": 2.348573684692383,
"eval_runtime": 47.285,
"eval_samples_per_second": 10.574,
"eval_steps_per_second": 1.332,
"step": 140
},
{
"epoch": 0.86,
"eval_BLEU_lug": 39.3584,
"eval_BLEU_mean": 39.3584,
"eval_loss": 2.347782850265503,
"eval_runtime": 46.3784,
"eval_samples_per_second": 10.781,
"eval_steps_per_second": 1.358,
"step": 150
}
],
"max_steps": 865,
"num_train_epochs": 5,
"total_flos": 7.715115655849574e+16,
"trial_name": null,
"trial_params": null
}