[xdoctest][task 109] Reformat example code with google style in python/paddle/optimizer/lr.py #56225
Conversation
|
你的PR提交成功，感谢你对开源项目的贡献! |
|
✅ This PR's description meets the template requirements! |
Candy2Tang
changed the title
python/paddle/optimizer/lr.py
Outdated
|
|
||
| >>> paddle.enable_static() | ||
| >>> base_lr = 0.1 | ||
| >>> sgd_optimizer = fluid.optimizer.SGD( |
There was a problem hiding this comment.
fluid API 即将退场，这里的 fluid.optimizer.SGD 可以使用 paddle.optimizer.SGD 代替，两者参数可能略有不同，可能需要稍微调整下，辛苦修改一下～后面的几个 API 同样～
另外，如果不需要开启 paddle.enable_static() 的话，就不要开启了
@Difers 迁移 API 时可以注意一下文档也修改下
python/paddle/optimizer/lr.py
Outdated
| .. code-block:: python | ||
|
|
||
| import paddle.fluid as fluid | ||
|
|
||
| boundaries = [100, 200] | ||
| lr_steps = [0.1, 0.01, 0.001] | ||
| learning_rate = fluid.layers.piecewise_decay(boundaries, lr_steps) #case1, 1D-Tensor | ||
| #learning_rate = 0.1 #case2, single-value | ||
| warmup_steps = 50 | ||
| start_lr = 1. / 3. | ||
| end_lr = 0.1 | ||
| decayed_lr = fluid.layers.linear_lr_warmup(learning_rate, | ||
| warmup_steps, start_lr, end_lr) | ||
|
|
||
| place = fluid.CPUPlace() | ||
| exe = fluid.Executor(place) | ||
| exe.run(fluid.default_startup_program()) | ||
| out, = exe.run(fetch_list=[decayed_lr.name]) | ||
| print(out) | ||
| # case1: [0.33333334] | ||
| # case2: [0.33333334] | ||
| >>> import paddle.fluid as fluid | ||
| >>> paddle.enable_static() | ||
| >>> boundaries = [100, 200] | ||
| >>> lr_steps = [0.1, 0.01, 0.001] | ||
| >>> learning_rate = fluid.layers.piecewise_decay(boundaries, lr_steps) #case1, 1D-Tensor | ||
| >>> # learning_rate = 0.1 # case2, single-value | ||
| >>> warmup_steps = 50 | ||
| >>> start_lr = 0.1 | ||
| >>> end_lr = 1. / 3. | ||
| >>> decayed_lr = fluid.layers.linear_lr_warmup( | ||
| ... learning_rate, | ||
| ... warmup_steps, | ||
| ... start_lr, | ||
| ... end_lr) | ||
| >>> place = fluid.CPUPlace() | ||
| >>> exe = fluid.Executor(place) | ||
| >>> exe.run(fluid.default_startup_program()) | ||
| >>> out, = exe.run(fetch_list=[decayed_lr.name]) | ||
| >>> print(out) | ||
| [0.1] |
There was a problem hiding this comment.
整体加一个缩进吧
以及 fluid API 需要替换
@Difers，确认一下这个 API 是否是动静统一？如果是的话，麻烦 @Candy2Tang 用动态图写法重写一下吧～应该比现在的会更简洁些～
@Candy2Tang 注意下这几个 API 是 @Difers 三天前刚从 fluid 迁移出来的（#55986），所以需要最新版 Nightly build 或者自行编译才能在本地测试运行
There was a problem hiding this comment.
这几个迁移过来的似乎主要都是用于静态图下的，在动态图下就是直接调用同文件上方的同功能动态图API了，将这几个统一成静态图的写法用于跟上方的API做区分会不会好些？
There was a problem hiding this comment.
那就麻烦 @Candy2Tang 统一用静态图写法来写吧，不需要改成动态图写法了
There was a problem hiding this comment.
@SigureMo 抱歉，今天刚有空统一来修复。想问下这个文档样例是要改成静态图的么？我有一些疑问，这个api 并不是放到 static 目录下，为什么要改成静态图形式呀？paddle不是主推的动态图么？
There was a problem hiding this comment.
@Candy2Tang 你好，这部分api(autoincreased_step_counter...linear_lr_warmup)是几天前从fluid目录迁移出来的#55986，当时考虑到这部分api与optimizer/lr.py中的api功能相似，因此迁移到了这里，同时可以看出这部分api在动态图时实际是调用原lr中的功能相似的api,主要是在静态图中发挥作用，因此建议改为静态图形式，且方便于原lr.py中动态图api区分。
…n/paddle/optimizer/lr.py (PaddlePaddle#56225) * [xdoctest][task 109] test=docs_preview * fix whitespace test=docs_preview * Apply suggestions from code review * fix indent and legacy fluid apis --------- Co-authored-by: Nyakku Shigure <sigure.qaq@gmail.com>
PR types
Others
PR changes
Docs
Description
对应tracking issue: #55629
预览链接：