rewrite the text classification demo. #83
Conversation
lcy-seso
force-pushed
the
rewrite_text_classification
branch
21 times, most recently
from
eeccb28
to
f884a61
Compare
| @@ -51,10 +37,10 @@ def main(): | |||
| learning_rate_schedule="discexp", ) | |||
|
|
|||
| train_reader = paddle.batch( | |||
| paddle.reader.shuffle(reader.test_reader("train.list"), buf_size=1000), | |||
| paddle.reader.shuffle(reader.train_reader("train.list"), buf_size=1000), | |||
There was a problem hiding this comment.
这处修改用来修改图像分类例子的bug，目前每个例子读取数据的方式确实不统一。后续提PR修改图像分类的例子。
| batch_size=BATCH_SIZE) | ||
| test_reader = paddle.batch( | ||
| reader.train_reader("test.list"), batch_size=BATCH_SIZE) | ||
| reader.test_reader("test.list"), batch_size=BATCH_SIZE) |
There was a problem hiding this comment.
这处修改用来修改图像分类例子的bug，目前每个例子读取数据的方式确实不统一。后续提PR修改图像分类的例子。
text_classification/index.html
Outdated
| ├── run.sh # 运行此脚本，可以以默认参数直接开始训练任务 | ||
| ├── train.py # 训练任务脚本 | ||
| └── utils.py # 定义通用的函数，例如：打印日志、解析命令行参数、构建字典、加载字典等 | ||
| ``` |
There was a problem hiding this comment.
缺少一个快速开始。其中包含上面的目录说明，及一个训练说明（训练过程输出日志里都是什么意思）、和一个预测说明（可以是一段code，参考：https://www.oschina.net/p/jieba/?fromerr=btIKdxHH -功能 1)分词的代码及output、当然也可以是一个gif图）
There was a problem hiding this comment.
接下来应该是之前写在末尾的『修改参数说明』，这样可以在新手运行出一个结果的前提下，来对比修改不同参数得到的结果不同。于是顺势引出CNN&DNN该如何选择
text_classification/index.html
Outdated
| cost = paddle.layer.classification_cost(input=output, label=lbl) | ||
|
|
||
| return cost, output, lbl | ||
| ``` |
There was a problem hiding this comment.
如果是面向初学者，数据格式的自定义要比其他信息重要得多，因为这可能是唯一"需要考虑"的事，所以优先级需要提到快速开始后面。而详解的这个位置可以对应加一个锚点链接。
| #!/bin/sh | ||
|
|
||
| python train.py \ | ||
| --nn_type="dnn" \ |
There was a problem hiding this comment.
这里的train的方式怎么又改成shell传参了，按照约定都应该写到train.py里？
There was a problem hiding this comment.
shell 里面为 train.py 指定的参数，直接运行shell 即可，否则需要在命令行敲长串的参数。
|
|
||
|
|
||
| def train(topology, | ||
| train_data_dir=None, |
There was a problem hiding this comment.
lm_rnn.py 里有个run_type=GRU #'or LSTM'，在这里是否可以增加一个方法的选择参数比如DNN OR CNN
There was a problem hiding this comment.
已有此参数，nn_type 用来指定选择何种模型。
lcy-seso
force-pushed
the
rewrite_text_classification
branch
from
f884a61
to
b1231a5
Compare
lcy-seso
force-pushed
the
rewrite_text_classification
branch
from
b1231a5
to
501ce21
Compare
| batch_size=BATCH_SIZE) | ||
| test_reader = paddle.batch( | ||
| reader.train_reader("test.list"), batch_size=BATCH_SIZE) | ||
| reader.test_reader("test.list"), batch_size=BATCH_SIZE) |
There was a problem hiding this comment.
这处修改用来修改图像分类例子的bug，目前每个例子读取数据的方式确实不统一。后续提PR修改图像分类的例子。
| @@ -51,10 +37,10 @@ def main(): | |||
| learning_rate_schedule="discexp", ) | |||
|
|
|||
| train_reader = paddle.batch( | |||
| paddle.reader.shuffle(reader.test_reader("train.list"), buf_size=1000), | |||
| paddle.reader.shuffle(reader.train_reader("train.list"), buf_size=1000), | |||
There was a problem hiding this comment.
这处修改用来修改图像分类例子的bug，目前每个例子读取数据的方式确实不统一。后续提PR修改图像分类的例子。
lcy-seso
force-pushed
the
rewrite_text_classification
branch
from
6068a69
to
a0529eb
Compare
|
Readme的目录顺序需要调整么？把模型介绍、模型详解放到最开头的地方？ |
lcy-seso
force-pushed
the
rewrite_text_classification
branch
4 times, most recently
from
74980fe
to
ba69ba5
Compare
lcy-seso
force-pushed
the
rewrite_text_classification
branch
from
ba69ba5
to
136a60d
Compare
rewrite the text classification example.