Files
llama-3-8b-base-ipo-ultrafe…/all_results.json
ModelHub XC 56cb9bbd99 初始化项目,由ModelHub XC社区提供模型
Model: jackf857/llama-3-8b-base-ipo-ultrafeedback-4xh200-batch-128-20260428-004616
Source: Original Platform
2026-05-10 14:41:48 +08:00

22 lines
784 B
JSON

{
"epoch": 0.9989528795811519,
"eval_logits/chosen": -0.1925041377544403,
"eval_logits/rejected": -0.21385852992534637,
"eval_logps/chosen": -5.949731349945068,
"eval_logps/rejected": -8.787737846374512,
"eval_loss": 2305.42529296875,
"eval_rewards/accuracies": 0.6759999990463257,
"eval_rewards/chosen": -0.04833168908953667,
"eval_rewards/margins": 0.026571419090032578,
"eval_rewards/rejected": -0.07490310817956924,
"eval_runtime": 78.3976,
"eval_samples": 2000,
"eval_samples_per_second": 25.511,
"eval_steps_per_second": 1.594,
"total_flos": 0.0,
"train_loss": 19137.84001572327,
"train_runtime": 6782.201,
"train_samples": 61135,
"train_samples_per_second": 9.014,
"train_steps_per_second": 0.07
}