Qwen3-8b-Lean-LoRA / train_results.json
Jforeverss's picture
Upload final Qwen3-8B Lean LoRA adapter
995fad2 verified
{
"epoch": 2.0,
"total_flos": 1.872336395754128e+20,
"train_loss": 0.11010749320663236,
"train_runtime": 58527.1588,
"train_samples_per_second": 20.164,
"train_steps_per_second": 0.079
}