From 347efcedb7ec86bb8c421a35b9a3cac4356f9dea Mon Sep 17 00:00:00 2001 From: ChenXin Date: Thu, 26 Mar 2020 23:01:18 +0800 Subject: [PATCH] =?UTF-8?q?=E5=A2=9E=E5=8A=A0=E4=BA=86=E6=95=99=E7=A8=8B?= =?UTF-8?q?=E7=9A=84=E4=BB=A3=E7=A0=81=E4=B8=8B=E8=BD=BD?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- docs/source/tutorials/extend_1_bert_embedding.rst | 5 +++++ docs/source/tutorials/tutorial_1_data_preprocess.rst | 6 ++++++ docs/source/tutorials/tutorial_2_vocabulary.rst | 9 ++++++++- docs/source/tutorials/tutorial_3_embedding.rst | 7 +++++++ docs/source/tutorials/tutorial_4_load_dataset.rst | 7 +++++++ docs/source/tutorials/tutorial_5_loss_optimizer.rst | 6 ++++++ docs/source/tutorials/tutorial_6_datasetiter.rst | 6 ++++++ docs/source/tutorials/tutorial_7_metrics.rst | 6 ++++++ docs/source/tutorials/tutorial_8_modules_models.rst | 7 +++++++ docs/source/tutorials/tutorial_9_callback.rst | 6 ++++++ docs/source/tutorials/序列标注.rst | 9 ++++++++- docs/source/tutorials/文本分类.rst | 6 ++++++ 12 files changed, 78 insertions(+), 2 deletions(-) diff --git a/docs/source/tutorials/extend_1_bert_embedding.rst b/docs/source/tutorials/extend_1_bert_embedding.rst index 07693097..929b3213 100644 --- a/docs/source/tutorials/extend_1_bert_embedding.rst +++ b/docs/source/tutorials/extend_1_bert_embedding.rst @@ -222,3 +222,8 @@ Bert自从在 `BERT: Pre-training of Deep Bidirectional Transformers for Languag CMRC2018Metric: f1=85.61, em=66.08 +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_1_data_preprocess.rst b/docs/source/tutorials/tutorial_1_data_preprocess.rst index 5104bfce..3bad7df0 100644 --- a/docs/source/tutorials/tutorial_1_data_preprocess.rst +++ b/docs/source/tutorials/tutorial_1_data_preprocess.rst @@ -162,3 +162,9 @@ fastNLP中field的命名习惯 - **chars**: 表示已经切分为单独的汉字的序列。例如["这", "是", "一", "个", "示", "例", "。"]。但由于神经网络不能识别汉字，所以一般该列会被转为int形式，如[3, 4, 5, 6, ...]。 - **target**: 表示目标值。分类场景下，只有一个值；序列标注场景下是一个序列 - **seq_len**: 表示输入序列的长度 + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_2_vocabulary.rst b/docs/source/tutorials/tutorial_2_vocabulary.rst index 130fadd2..ab825aa4 100644 --- a/docs/source/tutorials/tutorial_2_vocabulary.rst +++ b/docs/source/tutorials/tutorial_2_vocabulary.rst @@ -128,4 +128,11 @@ fastNLP中的Vocabulary 首先train和test都能够从预训练中找到对应的vector，所以它们是各自的vector表示; only_in_train在预训练中找不到，StaticEmbedding为它 新建了一个entry，所以它有一个单独的vector; 而only_in_test在预训练中找不到改词，因此被指向了unk的值(fastNLP用零向量初始化unk)，与最后一行unk的 -表示相同。 \ No newline at end of file +表示相同。 + + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_3_embedding.rst b/docs/source/tutorials/tutorial_3_embedding.rst index b719041e..796fbe94 100644 --- a/docs/source/tutorials/tutorial_3_embedding.rst +++ b/docs/source/tutorials/tutorial_3_embedding.rst @@ -451,3 +451,10 @@ fastNLP通过在 :class:`~fastNLP.embeddings.StaticEmbedding` 增加了一个min tensor([[ 0.6707, -0.5786, -0.6967, 0.0111, 0.1209]], grad_fn=) # unk 可以看到a不再和最后一行的unknown共享一个表示了，这是由于a与A都算入了a的词频，且A的表示也是a的表示。 + + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_4_load_dataset.rst b/docs/source/tutorials/tutorial_4_load_dataset.rst index 7a2f0877..1c6b214a 100644 --- a/docs/source/tutorials/tutorial_4_load_dataset.rst +++ b/docs/source/tutorials/tutorial_4_load_dataset.rst @@ -208,3 +208,10 @@ Part V: 不同格式类型的基础Loader "A person on a horse jumps over a broken down airplane.", "A person is training his horse for a competition.", "neutral" "A person on a horse jumps over a broken down airplane.", "A person is at a diner, ordering an omelette.", "contradiction" "A person on a horse jumps over a broken down airplane.", "A person is outdoors, on a horse.", "entailment" + + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_5_loss_optimizer.rst b/docs/source/tutorials/tutorial_5_loss_optimizer.rst index 968e7c92..f80ec746 100644 --- a/docs/source/tutorials/tutorial_5_loss_optimizer.rst +++ b/docs/source/tutorials/tutorial_5_loss_optimizer.rst @@ -238,3 +238,9 @@ Evaluate data in 0.43 seconds! [tester] AccuracyMetric: acc=0.773333 + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_6_datasetiter.rst b/docs/source/tutorials/tutorial_6_datasetiter.rst index d8e5c7c0..ff3b28ff 100644 --- a/docs/source/tutorials/tutorial_6_datasetiter.rst +++ b/docs/source/tutorials/tutorial_6_datasetiter.rst @@ -413,3 +413,9 @@ Dataset个性化padding AccuracyMetric: acc=0.786667 + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_7_metrics.rst b/docs/source/tutorials/tutorial_7_metrics.rst index 2731c023..c5002e45 100644 --- a/docs/source/tutorials/tutorial_7_metrics.rst +++ b/docs/source/tutorials/tutorial_7_metrics.rst @@ -125,3 +125,9 @@ self.evaluate将计算一个批次(batch)的评价指标，并累计。 没有返回值 self.get_metric将统计当前的评价指标并返回评价结果, 返回值需要是一个dict, key是指标名称，value是指标的值 + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_8_modules_models.rst b/docs/source/tutorials/tutorial_8_modules_models.rst index e3673d84..98d6234d 100644 --- a/docs/source/tutorials/tutorial_8_modules_models.rst +++ b/docs/source/tutorials/tutorial_8_modules_models.rst @@ -182,3 +182,10 @@ FastNLP 中包含的各种模块如下表，您可以点击具体的名称查看 :class:`~fastNLP.modules.viterbi_decode` , 给定一个特征矩阵以及转移分数矩阵，计算出最佳的路径以及对应的分数 （与 :class:`~fastNLP.modules.ConditionalRandomField` 配合使用） :class:`~fastNLP.modules.allowed_transitions` , 给定一个id到label的映射表，返回所有可以跳转的列表（与 :class:`~fastNLP.modules.ConditionalRandomField` 配合使用） :class:`~fastNLP.modules.TimestepDropout` , 简单包装过的Dropout 组件 + + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/tutorial_9_callback.rst b/docs/source/tutorials/tutorial_9_callback.rst index 4a51fdd9..4833f39d 100644 --- a/docs/source/tutorials/tutorial_9_callback.rst +++ b/docs/source/tutorials/tutorial_9_callback.rst @@ -130,3 +130,9 @@ fastNLP 中提供了很多常用的 Callback，如梯度裁剪，训练时早停 callbacks = [MyCallBack()] train_with_callback(callbacks) + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/序列标注.rst b/docs/source/tutorials/序列标注.rst index 17cbb298..e8e59cac 100644 --- a/docs/source/tutorials/序列标注.rst +++ b/docs/source/tutorials/序列标注.rst @@ -196,4 +196,11 @@ fastNLP的数据载入主要是由Loader与Pipe两个基类衔接完成的，您 [tester] SpanFPreRecMetric: f=0.641774, pre=0.626424, rec=0.657895 -可以看出通过使用Bert，效果有明显的提升，从48.2提升到了64.1。 \ No newline at end of file +可以看出通过使用Bert，效果有明显的提升，从48.2提升到了64.1。 + + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file diff --git a/docs/source/tutorials/文本分类.rst b/docs/source/tutorials/文本分类.rst index 997e35c8..583ddbfc 100644 --- a/docs/source/tutorials/文本分类.rst +++ b/docs/source/tutorials/文本分类.rst @@ -368,3 +368,9 @@ fastNLP提供了Trainer对象来组织训练过程，包括完成loss计算(所 {'AccuracyMetric': {'acc': 0.919167}} + +---------------------------------- +代码下载 +---------------------------------- + +`点击下载 IPython Notebook 文件 `_) \ No newline at end of file