Model: jackf857/llama-3-8b-base-robust-dpo-ultrafeedback-8xh200 Source: Original Platform
22 lines
780 B
JSON
22 lines
780 B
JSON
{
|
|
"epoch": 0.9989528795811519,
|
|
"eval_logits/chosen": -0.790234386920929,
|
|
"eval_logits/rejected": -0.7727676630020142,
|
|
"eval_logps/chosen": -301.2696838378906,
|
|
"eval_logps/rejected": -287.254638671875,
|
|
"eval_loss": 0.347857266664505,
|
|
"eval_rewards/accuracies": 0.7580645084381104,
|
|
"eval_rewards/chosen": -0.33950960636138916,
|
|
"eval_rewards/margins": 1.651003122329712,
|
|
"eval_rewards/rejected": -1.990512728691101,
|
|
"eval_runtime": 44.5446,
|
|
"eval_samples": 2000,
|
|
"eval_samples_per_second": 44.899,
|
|
"eval_steps_per_second": 1.414,
|
|
"total_flos": 0.0,
|
|
"train_loss": 1.6700679651096433,
|
|
"train_runtime": 3395.1223,
|
|
"train_samples": 61135,
|
|
"train_samples_per_second": 18.007,
|
|
"train_steps_per_second": 0.14
|
|
} |