Files
Qwen3-4B_Paper_Impact_SFT_1ep/train_results.json

8 lines
210 B
JSON
Raw Permalink Normal View History

{
"epoch": 1.0,
"total_flos": 2.850613244928721e+18,
"train_loss": 0.048995194472157194,
"train_runtime": 23358.4539,
"train_samples_per_second": 3.789,
"train_steps_per_second": 0.059
}