Qwen-14B-lora-pretrain / all_results.json
ytcheng's picture
End of training
0102256 verified
raw
history blame contribute delete
399 Bytes
{
"epoch": 2.998264893001735,
"eval_loss": 2.0396180152893066,
"eval_runtime": 154.6591,
"eval_samples_per_second": 3.731,
"eval_steps_per_second": 3.731,
"perplexity": 7.687672064754986,
"total_flos": 1.2799386304118784e+18,
"train_loss": 2.14731617155389,
"train_runtime": 14043.9874,
"train_samples_per_second": 1.108,
"train_steps_per_second": 0.138
}