olss 调度器，对 预训练的train_steps inference_steps promp 是否敏感；

我看实例代码里用

train_steps = 100
inference_steps = 10
prompt = "亭台楼阁，曲径通幽，水墨绘画，中国风"

olss 调度器，对 预训练的train_steps inference_steps promp 是否敏感？

如果不同的train_steps ，inference_steps ，promp 差距很大，有无推荐；能复现论文的结果

Start with the instance code using train_steps=100, inference_steps=10, and the stated prompt, then read the OLS scheduler entry point. Determine whether pretrained results depend on these settings and prompts, and document recommended values and the steps needed to reproduce the paper's results.

Written by the indexing model from the issue text.