run2_conservative
QLoRA adapter fine-tuned from Qwen/Qwen2.5-3B-Instruct on OpenAssistant/oasst1.
Final metrics
- train_loss: 1.3565049528865376
- eval_loss: 1.4962379932403564
- loss_gap: 0.13973304035381884
- eval_perplexity: 4.464860600017728
- train_runtime: 4330.0737
- samples_per_second: 4.183
- global_steps: 1134
- Downloads last month
- 19