学习率调度与预热
步进、余弦与预热策略
学习率调度与预热 是 CoddyKit 上的免费 Deep Learning Academy 课时。 这是第 4 节课，共 4 节。 你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。 这是 Deep Learning Academy 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 Deep Learning Academy 课程共包含 4 节课。
本课时的部分内容尚未翻译，以英文显示。
A Rate That Changes
One fixed learning rate is rarely ideal for the whole run. A schedule changes it over time, usually large early and small near the end.
Why Decay Helps
Big early steps cover ground fast. Shrinking the rate later lets the model settle precisely into a minimum instead of bouncing around it.
Step Decay
The simplest schedule is step decay: drop the rate by a fixed factor every set number of epochs, like halving it every thirty epochs.
sched = torch.optim.lr_scheduler.StepLR(opt, step_size=30, gamma=0.5)Cosine Annealing
Cosine schedules glide the rate down a smooth curve toward zero. The gentle, gradual decay is a favorite for training modern networks.
sched = torch.optim.lr_scheduler.CosineAnnealingLR(opt, T_max=100)The Cold-Start Problem
Fresh weights are fragile. A full-size rate on step one can blow up the loss, especially with large batches or deep transformers.
Warmup Eases In
Warmup ramps the rate up from near zero over the first few hundred steps. This gentle start keeps early training stable before full speed.
Warmup Then Decay
The classic recipe is warmup followed by decay: climb to the peak rate, then ride a cosine curve down. It is the go-to for big models.
Step the Scheduler
A scheduler does nothing until you call step() on it, usually once per epoch right after the optimizer updates the weights.
opt.step()
sched.step()Watch the Current Rate
Log the live learning rate while you train. Seeing it warm up and decay confirms the schedule fires when expected and helps you debug.
print(sched.get_last_lr())Plateau-Based Decay
Prefer reacting to results? ReduceLROnPlateau drops the rate only when validation loss stops improving, no fixed timetable needed.
sched = torch.optim.lr_scheduler.ReduceLROnPlateau(opt)Schedules Are Free Wins
A good schedule often boosts final accuracy with zero extra data. Warmup plus cosine decay is a strong, safe default to start from.
Quick Check
Confirm what warmup is for.
Recap
A schedule shrinks the learning rate over time so the model settles cleanly, while warmup ramps it up first for stable starts. Warmup plus cosine is a great default. 📉
常见问题解答
「学习率调度与预热」课时是免费的吗？
是的 — 「学习率调度与预热」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 Deep Learning Academy 课程的其余内容，请升级到 CoddyKit PRO。 Deep Learning Academy 课程共包含 4 节课。
「学习率调度与预热」这节课中我会学到什么？
步进、余弦与预热策略 你通过在浏览器中直接运行的动手代码来练习 Deep Learning Academy，全天候 AI 导师会在你学习这节课的过程中回答你的问题。
学习 Deep Learning Academy 需要有经验吗？
无需任何先前经验。CoddyKit 上的 Deep Learning Academy 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。 这是第 4 节课，共 4 节。
「学习率调度与预热」课时需要多长时间？
大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。
我能在这节 Deep Learning Academy 课中编写并运行代码吗？
能。每节 Deep Learning Academy 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。