Skip to content

modify default value for GradScaler #5937

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Jul 6, 2023
Merged

modify default value for GradScaler #5937

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
cb48f07
add amp models
zhangting2020 Dec 12, 2022
ebf7594
modify default value for GradScaler
zhangting2020 Jun 14, 2023
60f537f
Revert "add amp models"
zhangting2020 Jun 14, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions docs/api/paddle/amp/GradScaler_cn.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@
GradScaler
-------------------------------

.. py:class:: paddle.amp.GradScaler(enable=True, init_loss_scaling=32768.0, incr_ratio=2.0, decr_ratio=0.5, incr_every_n_steps=1000, decr_every_n_nan_or_inf=2, use_dynamic_loss_scaling=True)
.. py:class:: paddle.amp.GradScaler(enable=True, init_loss_scaling=65536.0, incr_ratio=2.0, decr_ratio=0.5, incr_every_n_steps=2000, decr_every_n_nan_or_inf=1, use_dynamic_loss_scaling=True)



Expand All @@ -21,11 +21,11 @@ GradScaler 用于动态图模式下的"自动混合精度"的训练。它控制
参数
:::::::::
- **enable** (bool，可选) - 是否使用 loss scaling。默认值为 True。
- **init_loss_scaling** (float，可选) - 初始 loss scaling 因子。默认值为 32768.0。
- **init_loss_scaling** (float，可选) - 初始 loss scaling 因子。默认值为 65536.0。
- **incr_ratio** (float，可选) - 增大 loss scaling 时使用的乘数。默认值为 2.0。
- **decr_ratio** (float，可选) - 减小 loss scaling 时使用的小于 1 的乘数。默认值为 0.5。
- **incr_every_n_steps** (int，可选) - 连续 n 个 steps 的梯度都是有限值时，增加 loss scaling。默认值为 1000。
- **decr_every_n_nan_or_inf** (int，可选) - 累计出现 n 个 steps 的梯度为 nan 或者 inf 时，减小 loss scaling。默认值为 2。
- **incr_every_n_steps** (int，可选) - 连续 n 个 steps 的梯度都是有限值时，增加 loss scaling。默认值为 2000。
- **decr_every_n_nan_or_inf** (int，可选) - 累计出现 n 个 steps 的梯度为 nan 或者 inf 时，减小 loss scaling。默认值为 1。
- **use_dynamic_loss_scaling** (bool，可选) - 是否使用动态的 loss scaling。如果不使用，则使用固定的 loss scaling；如果使用，则会动态更新 loss scaling。默认值为 True。

返回
Expand Down