From 26e64a8827f899d28d4087f481e16f3029e104cc Mon Sep 17 00:00:00 2001 From: Varuna Jayasiri Date: Mon, 24 Jun 2024 15:59:56 +0530 Subject: [PATCH] zh --- docs/zh/rl/ppo/gae.html | 26 ++++++++++++++++++-------- translate_cache/rl/ppo/gae.zh.json | 2 +- 2 files changed, 19 insertions(+), 9 deletions(-) diff --git a/docs/zh/rl/ppo/gae.html b/docs/zh/rl/ppo/gae.html index 4f24233b..f2d85de1 100644 --- a/docs/zh/rl/ppo/gae.html +++ b/docs/zh/rl/ppo/gae.html @@ -112,8 +112,8 @@ #

计算优势

-

是高偏差，低方差，而无偏差，高方差。

-

我们采用加权平均值来平衡偏差和方差。这称为广义优势估计。我们设置，这给出了干净的计算

+

是高偏差，低方差，而无偏差，高方差。

+

我们采用加权平均值来平衡偏差和方差。这称为广义优势估计。我们设置，这给出了干净的计算

@@ -190,15 +190,25 @@
#
-

请注意，我们正在按相反的顺序收集。我最初的代码被追加到一个列表中，后来我忘记反转它了。我花了大约 4 到 5 个小时才发现 bug。在初始运行期间，该模型的性能略有改善，这可能是因为样本相似。

+

- 
83            advantages[:, t] = last_advantage
-84
-85            last_value = values[:, t]
-86
-87        return advantages
+ 
77            advantages[:, t] = last_advantage
+78
+79            last_value = values[:, t]
+
+ +
+
+
+ # +
+

+ +
+
+ 
82        return advantages
diff --git a/translate_cache/rl/ppo/gae.zh.json b/translate_cache/rl/ppo/gae.zh.json index 9902b54e..025dfda8 100644 --- a/translate_cache/rl/ppo/gae.zh.json +++ b/translate_cache/rl/ppo/gae.zh.json @@ -1,10 +1,10 @@ { "

Generalized Advantage Estimation (GAE)

\n

This is a PyTorch implementation of paper Generalized Advantage Estimation.

\n

You can find an experiment that uses it here.

\n": "

\u5e7f\u4e49\u4f18\u52bf\u4f30\u8ba1 (GAE)

\n

\u8fd9\u662f\u8bba\u6587\u5e7f\u4e49\u4f18\u52bf\u4f30\u8ba1\u7684 PyTorch \u5b9e\u73b0\u3002

\n

\u4f60\u53ef\u4ee5\u5728\u8fd9\u91cc\u627e\u5230\u4e00\u4e2a\u4f7f\u7528\u5b83\u7684\u5b9e\u9a8c\u3002

\n", "

Calculate advantages

\n_^_0_^_

_^_1_^_ is high bias, low variance, whilst _^_2_^_ is unbiased, high variance.

\n

We take a weighted average of _^_3_^_ to balance bias and variance. This is called Generalized Advantage Estimation. _^_4_^_ We set _^_5_^_, this gives clean calculation for _^_6_^_

\n_^_7_^_": "

\u8ba1\u7b97\u4f18\u52bf

\n_^_0_^_

_^_1_^_\u662f\u9ad8\u504f\u5dee\uff0c\u4f4e\u65b9\u5dee\uff0c\u800c_^_2_^_\u65e0\u504f\u5dee\uff0c\u9ad8\u65b9\u5dee\u3002

\n

\u6211\u4eec\u91c7\u7528\u52a0\u6743\u5e73\u5747\u503c_^_3_^_\u6765\u5e73\u8861\u504f\u5dee\u548c\u65b9\u5dee\u3002\u8fd9\u79f0\u4e3a\u5e7f\u4e49\u4f18\u52bf\u4f30\u8ba1\u3002_^_4_^_\u6211\u4eec\u8bbe\u7f6e_^_5_^_\uff0c\u8fd9\u7ed9\u51fa\u4e86\u5e72\u51c0\u7684\u8ba1\u7b97_^_6_^_

\n_^_7_^_", + "

\n": "

\n", "

_^_0_^_

\n": "

_^_0_^_

\n", "

advantages table

\n": "

\u4f18\u52bf\u8868

\n", "

mask if episode completed after step _^_0_^_

\n": "

\u5982\u679c\u5267\u96c6\u5728\u6b65\u9aa4\u4e4b\u540e\u5b8c\u6210\uff0c\u8bf7\u63a9\u76d6_^_0_^_

\n", - "

note that we are collecting in reverse order. My initial code was appending to a list and I forgot to reverse it later. It took me around 4 to 5 hours to find the bug. The performance of the model was improving slightly during initial runs, probably because the samples are similar.

\n": "

\u8bf7\u6ce8\u610f\uff0c\u6211\u4eec\u6b63\u5728\u6309\u76f8\u53cd\u7684\u987a\u5e8f\u6536\u96c6\u3002\u6211\u6700\u521d\u7684\u4ee3\u7801\u88ab\u8ffd\u52a0\u5230\u4e00\u4e2a\u5217\u8868\u4e2d\uff0c\u540e\u6765\u6211\u5fd8\u8bb0\u53cd\u8f6c\u5b83\u4e86\u3002\u6211\u82b1\u4e86\u5927\u7ea6 4 \u5230 5 \u4e2a\u5c0f\u65f6\u624d\u53d1\u73b0 bug\u3002\u5728\u521d\u59cb\u8fd0\u884c\u671f\u95f4\uff0c\u8be5\u6a21\u578b\u7684\u6027\u80fd\u7565\u6709\u6539\u5584\uff0c\u8fd9\u53ef\u80fd\u662f\u56e0\u4e3a\u6837\u672c\u76f8\u4f3c\u3002

\n", "A PyTorch implementation/tutorial of Generalized Advantage Estimation (GAE).": "\u5e7f\u4e49\u4f18\u52bf\u4f30\u8ba1\uff08GAE\uff09\u7684 PyTorch \u5b9e\u73b0/\u6559\u7a0b\u3002", "Generalized Advantage Estimation (GAE)": "\u5e7f\u4e49\u4f18\u52bf\u4f30\u8ba1 (GAE)" } \ No newline at end of file