Skip to content

Commit 86ebdf4

Browse files
ShenYuhanPeterPanZH
ShenYuhan
and
PeterPanZH
authoredJan 29, 2021
Text demo (#923)
* add docs for text * fix image * fix image * fix image * fix text demo * fix grammer errors of text doc * fix grammer errors of text doc * fix grammer errors of text doc * fix grammer errors of text doc Co-authored-by: Peter Pan <littlepanzh@gmail.com>
·
v3.0.0-betav2.1.1
1 parent d556a76 commit 86ebdf4

File tree

5 files changed

+249
-1
lines changed

5 files changed

+249
-1
lines changed
 

‎demo/components/image_matrix_test.py

+53
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,53 @@
1+
# Copyright (c) 2021 VisualDL Authors. All Rights Reserve.
2+
#
3+
# Licensed under the Apache License, Version 2.0 (the "License");
4+
# you may not use this file except in compliance with the License.
5+
# You may obtain a copy of the License at
6+
#
7+
# http://www.apache.org/licenses/LICENSE-2.0
8+
#
9+
# Unless required by applicable law or agreed to in writing, software
10+
# distributed under the License is distributed on an "AS IS" BASIS,
11+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12+
# See the License for the specific language governing permissions and
13+
# limitations under the License.
14+
# =======================================================================
15+
16+
# coding=utf-8
17+
18+
import numpy as np
19+
from PIL import Image
20+
from visualdl import LogWriter
21+
22+
23+
def random_crop(img):
24+
"""Get random block of img, which size is 100x100.
25+
"""
26+
img = Image.open(img)
27+
w, h = img.size
28+
random_w = np.random.randint(0, w - 100)
29+
random_h = np.random.randint(0, h - 100)
30+
r = img.crop((random_w, random_h, random_w + 100, random_h + 100))
31+
return np.asarray(r)
32+
33+
34+
if __name__ == '__main__':
35+
imgs = []
36+
# 获取8张图像
37+
for step in range(8):
38+
img = random_crop("../../docs/images/dog.jpg")
39+
imgs.append(img)
40+
imgs = np.array(imgs)
41+
42+
with LogWriter(logdir='./log/image_matrix_test/train') as writer:
43+
writer.add_image(tag='doges', step=0, img=imgs[0])
44+
# 合成长宽尽量接近的图形矩阵，本例生成3X3的矩阵
45+
writer.add_image_matrix(tag='doges', step=1, imgs=imgs, rows=-1)
46+
# 合成长为1的图形矩阵，本例生成1x8的矩阵
47+
writer.add_image_matrix(tag='doges', step=2, scale=10, imgs=imgs, rows=1)
48+
# 合成长为2的图形矩阵，本例生成2X4的矩阵
49+
writer.add_image_matrix(tag='doges', step=3, scale=10, imgs=imgs, rows=2)
50+
# 合成长为3的图形矩阵，本例生成3X3的矩阵
51+
writer.add_image_matrix(tag='doges', step=4, imgs=imgs, rows=3)
52+
# 合成长为4的图形矩阵，本例生成4X2的矩阵
53+
writer.add_image_matrix(tag='doges', step=5, imgs=imgs, rows=4)

‎demo/components/test.sh

+11
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,11 @@
1+
#!/bin/bash
2+
python scalar_test.py
3+
python image_test.py
4+
python image_matrix_test.py
5+
python histogram_test.py
6+
python high_dimensional_test.py
7+
python audio_test.py
8+
python text_test.py
9+
python pr_curve_test.py
10+
python roc_curve_test.py
11+

‎demo/components/text_test.py

+31
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,31 @@
1+
# Copyright (c) 2021 VisualDL Authors. All Rights Reserve.
2+
#
3+
# Licensed under the Apache License, Version 2.0 (the "License");
4+
# you may not use this file except in compliance with the License.
5+
# You may obtain a copy of the License at
6+
#
7+
# http://www.apache.org/licenses/LICENSE-2.0
8+
#
9+
# Unless required by applicable law or agreed to in writing, software
10+
# distributed under the License is distributed on an "AS IS" BASIS,
11+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12+
# See the License for the specific language governing permissions and
13+
# limitations under the License.
14+
# =======================================================================
15+
16+
# coding=utf-8
17+
18+
from visualdl import LogWriter
19+
20+
if __name__ == '__main__':
21+
texts = [
22+
'上联: 众 佛 群 灵 光 圣 地 下联: 众 生 一 念 证 菩 提',
23+
'上联: 乡 愁 何 处 解 下联: 故 事 几 时 休',
24+
'上联: 清 池 荷 试 墨 下联: 碧 水 柳 含 情',
25+
'上联: 既 近 浅 流 安 笔 砚 下联: 欲 将 直 气 定 乾 坤',
26+
'上联: 日 丽 萱 闱 祝 无 量 寿 下联: 月 明 桂 殿 祝 有 余 龄',
27+
'上联: 一 地 残 红 风 拾 起 下联: 半 窗 疏 影 月 窥 来'
28+
]
29+
with LogWriter(logdir="./log/text_test/train") as writer:
30+
for step in range(len(texts)):
31+
writer.add_text(tag="output", step=step, text_string=texts[step])

‎docs/components/README.md

+78-1
Original file line numberDiff line numberDiff line change
@@ -13,12 +13,14 @@ VisualDL 是一个面向深度学习任务设计的可视化工具。VisualDL
1313
| [ Scalar](#Scalar--标量组件) | 折线图 | 动态展示损失函数值、准确率等标量数据 |
1414
| [Image](#Image--图片可视化组件) | 图片可视化 | 显示图片，可显示输入图片和处理后的结果，便于查看中间过程的变化 |
1515
| [Audio](#Audio--音频播放组件) | 音频可视化 | 播放训练过程中的音频数据，监控语音识别与合成等任务的训练过程 |
16+
| [Text](#Text--文本组件) | 文本可视化 | 展示文本任务任意阶段的数据输出，对比不同阶段的文本变化，便于深入了解训练过程及效果。 |
1617
| [Graph](#Graph--网络结构组件) | 网络结构 | 展示网络结构、节点属性及数据流向，辅助学习、优化网络结构 |
1718
| [Histogram](#Histogram--直方图组件) | 直方图 | 展示训练过程中权重、梯度等张量的分布 |
1819
| [PR Curve](#PR-Curve--PR曲线组件) | 折线图 | 权衡精度与召回率之间的平衡关系 |
1920
| [ROC Curve](#ROC-Curve--ROC曲线组件) | 折线图 | 展示不同阈值下的模型表现 |
2021
| [High Dimensional](#High-Dimensional--数据降维组件) | 数据降维 | 将高维数据映射到 2D/3D 空间来可视化嵌入，便于观察不同数据的相关性 |
2122

23+
2224
同时，VisualDL提供可视化结果保存服务，通过 [VDL.service](#vdlservice) 生成链接，保存并分享可视化结果
2325

2426
## Scalar--标量组件
@@ -305,6 +307,8 @@ add_image_matrix(tag, imgs, step, rows=-1, scale=1, walltime=None, dataformats="
305307
| walltime | int | 记录数据的时间戳，默认为当前时间戳 |
306308
| dataformats| string | 传入的图片格式，包括`NCHW`、`HWC`、`HW`，默认为`HWC`|
307309

310+
**PS：当给定的子图像数量不足时，将用空白图像填充，以保证生成的图形为完整矩形**
311+
308312
#### Demo
309313
下面展示了使用 Image 组件合成并记录多张图片数据的示例，代码文件请见[Image组件](https://github.com/PaddlePaddle/VisualDL/blob/develop/demo/components/image_matrix_test.py)
310314
```python
@@ -438,6 +442,80 @@ visualdl --logdir ./log --port 8080
438442
<img src="https://user-images.githubusercontent.com/48054808/87661166-c277b880-c792-11ea-8ad7-5c60bb08379b.png" width="40%"/>
439443
</p>
440444

445+
## Text--文本组件
446+
447+
### 介绍
448+
449+
Text展示文本任务任意阶段的数据输出，对比不同阶段的文本变化，便于深入了解训练过程及效果。
450+
451+
### 记录接口
452+
453+
Text组件的记录接口如下：
454+
455+
```python
456+
add_text(self, tag, text_string, step=None, walltime=None)
457+
```
458+
459+
接口参数说明如下：
460+
461+
| 参数 | 格式 | 含义 |
462+
| -------------- | --------------------- | ------------------------------------------- |
463+
| tag | string | 记录指标的标志，如`train/loss`，不能含有`%` |
464+
| text_string | string | 文本字符串 |
465+
| step | int | 记录的步数 |
466+
| walltime | int | 记录数据的时间戳，默认为当前时间戳 |
467+
468+
### Demo
469+
470+
下面展示了使用 Text 组件记录数据的示例，代码见[Text组件](https://github.com/PaddlePaddle/VisualDL/blob/develop/demo/components/text_test.py)
471+
472+
```python
473+
from visualdl import LogWriter
474+
if __name__ == '__main__':
475+
texts = [
476+
'上联: 众 佛 群 灵 光 圣 地 下联: 众 生 一 念 证 菩 提',
477+
'上联: 乡 愁 何 处 解 下联: 故 事 几 时 休',
478+
'上联: 清 池 荷 试 墨 下联: 碧 水 柳 含 情',
479+
'上联: 既 近 浅 流 安 笔 砚 下联: 欲 将 直 气 定 乾 坤',
480+
'上联: 日 丽 萱 闱 祝 无 量 寿 下联: 月 明 桂 殿 祝 有 余 龄',
481+
'上联: 一 地 残 红 风 拾 起 下联: 半 窗 疏 影 月 窥 来'
482+
]
483+
with LogWriter(logdir="./log/text_test/train") as writer:
484+
for step in range(len(texts)):
485+
writer.add_text(tag="output", step=step, text_string=texts[step])
486+
```
487+
488+
运行上述程序后，在命令行执行
489+
490+
```shell
491+
visualdl --logdir ./log --port 8080
492+
```
493+
494+
接着在浏览器打开`http://127.0.0.1:8080`，即可查看Text
495+
496+
<p align="center">
497+
<img src="https://user-images.githubusercontent.com/28444161/106248340-cdd09400-624b-11eb-8ea9-5a07a239c365.png" width="95%"/>
498+
</p>
499+
500+
### 功能操作说明
501+
502+
- 可搜索文本标签显示对应文本数据
503+
504+
<p align="center">
505+
<img src="https://user-images.githubusercontent.com/48054808/86536503-baaa4f80-bf1a-11ea-80ab-cd988617d018.png" width="40%"/>
506+
</p>
507+
508+
- 可搜索数据流标签显示对应数据流数据
509+
510+
<p align="center">
511+
<img src="https://user-images.githubusercontent.com/28444161/106256983-f4e09300-6256-11eb-9acc-a24a2ac9b70c.png" width="40%"/>
512+
</p>
513+
514+
- 可折叠标签
515+
516+
<p align="center">
517+
<img src="https://user-images.githubusercontent.com/28444161/106252364-28202380-6251-11eb-934c-d8893c2eaeca.png" width="80%"/>
518+
</p>
441519

442520
## Graph--网络结构组件
443521

@@ -880,7 +958,6 @@ visualdl --logdir ./log --port 8080
880958
<img src="https://user-images.githubusercontent.com/48054808/103192766-d2d0b980-4914-11eb-871e-e4b31542c5e9.png" width="27%"/>
881959
</p>
882960

883-
884961
## VDL.service
885962

886963
### 简介

‎docs/components/UserGuide-en.md

+76
Original file line numberDiff line numberDiff line change
@@ -16,6 +16,7 @@ Currently, VisualDL provides seven components: scalar, image, audio, graph, hist
1616
| [ Scalar](#Scalar--Line-Chart) | Line Chart | Display scalar data such as loss and accuracy dynamically. |
1717
| [Image](#Image--Image-Visualization) | Image Visualization | Display images, visualizing the input and the output and making it easy to view the changes in the intermediate process. |
1818
| [Audio](#Audio--Audio-Play) | Audio Play | Play the audio during the training process, making it easy to monitor the process of speech recognition and text-to-speech. |
19+
| [Text](#Text) | Text Visualization | Visualizes the text output of NLP models within any stage, aiding developers to compare the changes of outputs so as to deeply understand the training process and simply evaluate the performance of the model. |
1920
| [Graph](#Graph--Network-Structure) | Network Structure | Visualize network structures, node attributes and data flow, assisting developers to learn and to optimize network structures. |
2021
| [Histogram](#Histogram--Distribution-of-Tensors) | Distribution of Tensors | Present the changes of distributions of tensors, such as weights/gradients/bias, during the training process. |
2122
| [PR Curve](#PR-Curve) | Precision & Recall Curve | Display precision-recall curves across training steps, clarifying the tradeoff between precision and recall when comparing models. |
@@ -370,6 +371,81 @@ Then, open the browser and enter the address: `http://127.0.0.1:8080`to view:
370371
<img src="https://user-images.githubusercontent.com/48054808/88755377-9a755580-d193-11ea-947e-4275b9d3aa54.png" width="40%"/>
371372
</p>
372373

374+
## Text
375+
376+
### Introduction
377+
378+
visualizes the text output of NLP models within any stage, aiding developers to compare the changes of outputs so as to deeply understand the training process and simply evaluate the performance of the model.
379+
380+
### Record Interface
381+
382+
The interface of the Text is shown as follows:
383+
384+
```python
385+
add_text(self, tag, text_string, step=None, walltime=None)
386+
```
387+
388+
The interface parameters are described as follows:
389+
390+
| parameter | format | meaning |
391+
| -------------- | --------------------- | ------------------------------------------- |
392+
| tag | string | Record the name of the text data，e.g.train/loss. Notice that the name cannot contain `%` |
393+
| text_string | string | Value of text |
394+
| step | int | Record the training steps |
395+
| walltime | int | Record the time-stamp of the data, and the default is the current time-stamp |
396+
397+
### Demo
398+
399+
The following shows an example of how to use Text component, and script can be found in [Text Demo](../../demo/components/text_test.py)
400+
401+
```python
402+
from visualdl import LogWriter
403+
if __name__ == '__main__':
404+
texts = [
405+
'上联: 众 佛 群 灵 光 圣 地 下联: 众 生 一 念 证 菩 提',
406+
'上联: 乡 愁 何 处 解 下联: 故 事 几 时 休',
407+
'上联: 清 池 荷 试 墨 下联: 碧 水 柳 含 情',
408+
'上联: 既 近 浅 流 安 笔 砚 下联: 欲 将 直 气 定 乾 坤',
409+
'上联: 日 丽 萱 闱 祝 无 量 寿 下联: 月 明 桂 殿 祝 有 余 龄',
410+
'上联: 一 地 残 红 风 拾 起 下联: 半 窗 疏 影 月 窥 来'
411+
]
412+
with LogWriter(logdir="./log/text_test/train") as writer:
413+
for step in range(len(texts)):
414+
writer.add_text(tag="output", step=step, text_string=texts[step])
415+
```
416+
417+
After running the above program, developers can launch the panel by:
418+
419+
```shell
420+
visualdl --logdir ./log --port 8080
421+
```
422+
423+
Then, open the browser and enter the address`http://127.0.0.1:8080` to view:
424+
425+
<p align="center">
426+
<img src="https://user-images.githubusercontent.com/28444161/106248340-cdd09400-624b-11eb-8ea9-5a07a239c365.png" width="95%"/>
427+
</p>
428+
429+
### Functional Instrucions
430+
431+
- Developers can find the target text by searching corresponded tags.
432+
433+
<p align="center">
434+
<img src="https://user-images.githubusercontent.com/48054808/86536503-baaa4f80-bf1a-11ea-80ab-cd988617d018.png" width="40%"/>
435+
</p>
436+
437+
- Developers can find the target runs by searching corresponded tags.
438+
439+
<p align="center">
440+
<img src="https://user-images.githubusercontent.com/28444161/106256983-f4e09300-6256-11eb-9acc-a24a2ac9b70c.png" width="40%"/>
441+
</p>
442+
443+
- Developers can fold the tab of text.
444+
445+
<p align="center">
446+
<img src="https://user-images.githubusercontent.com/28444161/106252364-28202380-6251-11eb-934c-d8893c2eaeca.png" width="80%"/>
447+
</p>
448+
373449

374450
## Graph--Network Structure
375451

0 commit comments

Comments
 (0)
Please sign in to comment.