amburger66's picture
LoRA fine-tune on RobotSmith task03 after fixing dataset
aea53cd verified
raw
history blame contribute delete
407 Bytes
{
"step": 450,
"metrics": {
"eval_rew_align/success_auprc_robotsmith": 0.5288600712338849,
"eval_rew_align/positive_success_acc_robotsmith": 0.8125,
"eval_rew_align/negative_success_acc_robotsmith": 0.9348591549295775,
"eval_rew_align/loss_robotsmith": 4.137787556648254,
"eval_rew_align/pearson_robotsmith": 0.9885291064855337,
"time/custom_evaluations": 83.76057602092624
}
}