-
Notifications
You must be signed in to change notification settings - Fork 2
/
image.py
367 lines (291 loc) · 14.4 KB
/
image.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
from PIL import Image, ImageDraw, ImageFont
import requests
from io import BytesIO
import textwrap
import qrcode
def text_wrap(text, width):
"""将文本按指定宽度换行，中英文字符长度不同"""
wrapped_text = ""
line_len = 0
for char in text:
if line_len + (2 if ord(char) > 127 else 1) > width:
wrapped_text += '\n'
line_len = 0
wrapped_text += char
line_len += 2 if ord(char) > 127 else 1
return wrapped_text
def split_text(text, width, font):
lines = []
line = ''
for word in text:
if word == '\n':
lines.append(line + '\n')
line = ""
continue
if font.getlength(line + word) <= width:
line += word
else:
lines.append(line)
line = word
if line:
lines.append(line)
return lines
def conversation2png(title, conv_user, conv_robot, seed=42):
# 图片大小
img_width = 1000
img_height = 1200 + 150 * len(conv_user)
# 背景颜色
bg_color = (255, 255, 255)
bg_color = (32, 32, 32)
# 字体文件路径
font_path = 'font/MaoKenZhuYuanTi-2.ttf'
# 字体大小
font_size = 24
# 字体颜色
font_color = (0, 0, 0)
# 创建一张空白图片
image = Image.new('RGB', (img_width, img_height), bg_color)
# 获取头像图片并调整大小
user_url = f"https://api.dicebear.com/5.x/lorelei/png?seed={seed}"
robot_url = f"https://api.dicebear.com/5.x/bottts/png?seed={seed}"
avatar_default = Image.new(mode="RGBA", size=(50, 50), color="gray")
try:
avatar_user = Image.open(BytesIO(requests.get(user_url).content)).resize((50, 50))
avatar_robot = Image.open(BytesIO(requests.get(robot_url).content)).resize((50, 50))
except:
avatar_user = avatar_default
avatar_robot = avatar_default
# image.paste(avatar_a, (50, 50))
# image.paste(avatar_b, (img_width - 150, 50))
# 创建Draw对象
draw = ImageDraw.Draw(image)
# 定义字体
font = ImageFont.truetype(font_path, font_size)
# 定义对话内容
dialogue = []
for us, rs in zip(conv_user, conv_robot):
dialogue.append({'speaker': 'user', 'text': us})
dialogue.append({'speaker': 'robot', 'text': rs})
# dialogue = [
# {'speaker': 'A', 'text': '你好啊\nHello'},
# {'speaker': 'B', 'text': '你好，最近怎么样？'},
# {'speaker': 'A', 'text': '还不错，最近忙着写代码呢'},
# {'speaker': 'B', 'text': '哦，有什么好玩的项目吗？'},
# {'speaker': 'A', 'text': '有啊，最近在做一个图片渲染的项目 This is a wonderful project nice'},
# {'speaker': 'B', 'text': '听起来很有趣啊，能不能给我看看效果？听起来很有趣啊，能不能给我看看效果？有啊，最近在做一个图片渲染的项目'}
# ]
# 定义对话框位置和大小
dialogue_box_width = 780
dialogue_box_height = font_size + 10
dialogue_box_x = 110
dialogue_box_y = 100
dialogue_box_padding = 10
# 渲染标题文字
draw.text((dialogue_box_x // 2, dialogue_box_y - font_size), title, font=font, fill=(233, 233, 233))
# 渲染对话文字
box_y = dialogue_box_x
total_height = dialogue_box_y
for i, message in enumerate(dialogue):
speaker = message['speaker']
text = message['text']
# 计算对话框位置
box_x = dialogue_box_x
box_y += (dialogue_box_padding + font_size)
# 自动换行
line_arr = []
for txt in text.split('\n'):
line_arr.append(textwrap.fill(txt, width=31))
lines = '\n'.join(line_arr)
# lines = text_wrap(text, width=62)
lines = lines.split('\n')
# 计算对话框高度
box_height = len(lines) * (font_size + dialogue_box_padding)
# 更新总高度
total_height += box_height + dialogue_box_padding
# 绘制对话框
fill_color = (240, 240, 240) if speaker == 'robot' else (237, 211, 161)
draw.rectangle([(box_x, box_y), (box_x + dialogue_box_width, box_y + box_height + dialogue_box_padding)], fill=fill_color, outline=(0, 0, 0))
# 绘制头像
if speaker == 'user':
image.paste(avatar_user, (box_x - 64, box_y))
else:
image.paste(avatar_robot, (img_width - 96, box_y))
# 绘制文字
for line in lines:
draw.text((box_x + 10, box_y + 10), line, font=font, fill=font_color)
box_y += dialogue_box_height
# 调整图片高度
total_height = box_y
if total_height + 100 < img_height:
image = image.crop((0, 0, img_width, total_height + 100))
# 保存图片
image.save('dialogue.png')
return image
def draw_rounded_rectangle(draw, xy, radius, fill=None, outline=None, width=1):
"""
绘制圆角矩形
"""
# 计算圆角矩形的四个角的坐标
x1, y1, x2, y2 = xy
draw.rectangle([(x1+radius,y1),(x2-radius,y2)], fill=fill, outline=outline, width=width)
draw.rectangle([(x1,y1+radius),(x2,y2-radius)], fill=fill, outline=outline, width=width)
draw.pieslice([(x1,y1),(x1+radius*2,y1+radius*2)], 180, 270, fill=fill, outline=outline, width=width)
draw.pieslice([(x2-radius*2,y1),(x2,y1+radius*2)], 270, 360, fill=fill, outline=outline, width=width)
draw.pieslice([(x1,y2-radius*2),(x1+radius*2,y2)], 90, 180, fill=fill, outline=outline, width=width)
draw.pieslice([(x2-radius*2,y2-radius*2),(x2,y2)], 0, 90, fill=fill, outline=outline, width=width)
def generate_tag_image(text):
# 字体文件路径
font_path = 'font/MaoKenZhuYuanTi-2.ttf'
# font_path = 'font/NotoColorEmoji-Regular.ttf'
# 设置字体
font = ImageFont.truetype(font_path, size=16)
# 计算文本所需宽度和高度
text_width, text_height = font.getbbox(text)[-2:]
# 创建画布
img_width = text_width + 20
img_height = text_height + 20
img = Image.new("RGBA", (img_width, img_height), color=(0, 0, 0, 0))
draw = ImageDraw.Draw(img)
# 绘制圆角矩形
radius = text_height / 1.5
rectangle_xy = (0, 0, text_width + 20, text_height + 20)
rectangle_fill = (255, 255, 255, 255)
draw_rounded_rectangle(draw, rectangle_xy, radius, fill=rectangle_fill)
# 绘制文本
text_x = 10
text_y = 10
text_color = (0, 0, 0, 255)
draw.text((text_x, text_y), text, font=font, fill=text_color)
return img
def generate_article_image(title, summary, url, tag_list, words, author):
title = title.strip()
summary = summary.strip()
# 设置图片宽度
img_width = 1000
# 字体文件路径
font_path = 'font/MaoKenZhuYuanTi-2.ttf'
# 设置字体
title_font = ImageFont.truetype(font_path, size=30)
summary_font = ImageFont.truetype(font_path, size=24)
hint_font = ImageFont.truetype(font_path, size=16)
# 计算摘要所需高度
summary_line_padding = 10
# summary_lines = textwrap.wrap(summary, width=32)
summary_lines = split_text(summary, 860, summary_font)
summary_height = len(summary_lines) * (summary_font.getbbox(' ')[-1] + summary_line_padding) + 68 - 7 * 3
summary_height = len(summary_lines) * (summary_font.getbbox(' ')[-1] + summary_line_padding + 4) + 31
print('Summary height', summary_height)
# 创建画布
title_lines = split_text(title, 820, title_font) # 自动标题换行
title = "\n".join(title_lines)
title_text_height = len(title_lines) * (title_font.getbbox(title)[-1] + 0)
print("["+title+"]", title_font.getbbox(title), summary_font.getlength(title), '==', title_text_height)
img_height = 50 + title_text_height + 50 + summary_height + 50 + 150
img = Image.new("RGB", (img_width, img_height), color=(255, 192, 0))
draw = ImageDraw.Draw(img)
# 绘制标题
title_x = 50
title_y = 50
draw.text((title_x, title_y), title, font=title_font, fill=(0, 0, 0))
# 绘制摘要
summary_x = 50
summary_y = title_y + 50 + title_text_height
summary_bg_width = 900
summary_bg_height = summary_height
summary_bg_x = summary_x
summary_bg_y = summary_y
ofs = 8
draw_rounded_rectangle(draw, (summary_bg_x+ofs, summary_bg_y+ofs, summary_bg_x + summary_bg_width + ofs, summary_bg_y + summary_bg_height + ofs), radius=40, fill=(132, 60, 12), outline=(0, 192, 0), width=0)
draw_rounded_rectangle(draw, (summary_bg_x, summary_bg_y, summary_bg_x + summary_bg_width, summary_bg_y + summary_bg_height), radius=40, fill=(255, 255, 255), outline=(0, 192, 0), width=0)
# 在摘要上方绘制提示文字
read_words_per_minute = 380
draw.text((summary_x + 20, summary_y - 30), f"全文约 {words} 字 （{max(round(words / read_words_per_minute), 1)} 分钟阅读）", font=hint_font, fill=(0, 32, 192))
summary_x += 20
summary_y += 20
summary_lines = split_text(summary, 860, summary_font)
for line in summary_lines:
draw.text((summary_x, summary_y), line, font=summary_font, fill=(0, 0, 0))
summary_y += summary_font.getbbox(line)[-1] + summary_line_padding
# summary_x = 50
# summary_y = title_y + 50 + title_font.getbbox(title)[-1]
# summary_lines = textwrap.wrap(summary, width=32)
# for line in summary_lines:
# draw.text((summary_x, summary_y), line, font=summary_font, fill=(0, 0, 0))
# summary_y += summary_font.getbbox(line)[-1] + summary_line_padding
print("["+line+"]", summary_font.getbbox(line), summary_font.getlength(line), '==', summary_y)
# 绘制标签
tag_x = summary_x
tag_y = summary_y + 40
for tag_str in tag_list:
img_tag = generate_tag_image("· " + tag_str)
img.paste(img_tag, (tag_x, tag_y), img_tag)
tag_x += img_tag.size[0] + 20
print('img size', img_tag.size)
# 绘制二维码
qr_size = 150
qr_margin = 20
qr_x = img_width - qr_size - qr_margin
qr_y = img_height - qr_size - qr_margin
qr = qrcode.QRCode(version=1, error_correction=qrcode.constants.ERROR_CORRECT_L, box_size=10, border=4)
qr.add_data(url)
qr.make(fit=True)
qr_img = qr.make_image(fill_color="black", back_color="white")
qr_img = qr_img.resize((qr_size, qr_size)) # 缩放二维码
img.paste(qr_img, (qr_x, qr_y))
# 绘制来源信息及版权声明
source_x = summary_x
source_y = tag_y + 60
draw.text((source_x, source_y), f"[*] 本内容使用 ChatGPT 辅助生成\n\n原始来源：{author} >> 扫描右侧二维码查看原文", font=hint_font, fill=(0, 32, 192))
print((source_x, source_y), "["+author+"]")
# 裁剪图片
img = img.crop((0, 0, img_width, title_y + 50 + title_text_height + 50 + summary_height + qr_size))
print(img.size)
return img
if __name__ == '__main__':
title = "人手一个ChatGPT！微软DeepSpeed Chat震撼发布.吴恩达ChatGPT Prompt Engineering for De"
summary = "本文介绍了微软开源的DeepSpeed Chat系统框架，它可以帮助开发者实现类ChatGPT模型的端到端RLHF训练，并大幅降低训练成本。文章详细介绍了DeepSpeed Chat的三大核心功能和不同规模模型和硬件配置下的训练时间和成本。同时，本文还介绍了DeepSpeed-HE在Azure云上训练RLHF模型的优异性能，包括快速训练、卓越的扩展性和完整的训练流程。特别是，混合引擎的设计使得RLHF训练的经验生成阶段的推理执行过程中，DeepSpeed混合引擎使用轻量级的内存管理系统，来处理KV缓存和中间结果，同时使用高度优化的推理CUDA核和张量并行计算，与现有方案相比，实现了吞吐量（每秒token数）的大幅提升。推荐阅读此文，了解如何使用DeepSpeed Chat实现高质量类ChatGPT模型的训练，以及DeepSpeed-HE的优势和混合引擎的设计，对于从事RLHF模型训练的数据科学家和研究者具有重要参考价值。"
url = "https://example.com"
tag_list = ["人工智能", "NLP"]
words = 400
author = "hahah"
summary = "本文介绍了吴恩达和OpenAI工程师Isa Fulford推出的课程ChatGPT Prompt Engineering for Developers，该课程讲解了基于预训练模型的Prompt工程和指导原则。对于开发者来说，这是一个非常有价值的课程，建议关注该领域的读者可以深入了解。推荐阅读原版视频学习，以获取更多实用知识和技能。"
# summary = "a\nb\ncccdddd"
img = generate_article_image(title, summary, url, tag_list, words, author)
# img.show()
img.save('summary.png')
#TODO: 调整每行的字符数以匹配宽度 Done
#TODO: 加入类别 + emoji 的标签 tag
#TODO: 给标题加上 emoji 的诠释
# https://playgpt3.streamlit.app/?share=y0snzxcr
# from pilmoji import Pilmoji
# from PIL import Image, ImageFont
# im = Image.new('RGB', (1024, 768))
# drawer = Pilmoji(im).draw
# FONT_PATH = 'font/NotoColorEmoji.ttf'
# FONT_PATH = 'font/MaoKenZhuYuanTi-2.ttf'
# FONT_PATH = r'D:\\Workspace\\streamlit-gpt3\\font\\NotoColorEmoji-Regular.ttf'
# font_size = 30
# font = ImageFont.truetype(FONT_PATH, size=font_size)
# text = "Hello, world! 👋 Here are some emojis: 🎨 🌊 😎"
# drawer((10, 10), text, font=font)
# im.save('emoji.png')
# from PIL import Image, ImageDraw, ImageFont
# back_ground_color = (50, 50, 50)
# # unicode_text = "some test\U0001f602"
# unicode_text = "😎" # "\U0001f602"
# im = Image.new("RGB", (500, 500), back_ground_color)
# draw = ImageDraw.Draw(im)
# # noto emoji, bit-map based, src: https://github.com/googlefonts/noto-emoji/blob/main/fonts/NotoColorEmoji.ttf
# # apple emoji, bit-map based, src: https://github.com/samuelngs/apple-emoji-linux/releases
# # segoe ui emoji, vector-based, works with different font size, font from here: https://fontsdata.com/132714/segoeuiemoji.htm (for test only)
# font_info = { #'symbola': [r"E:\symbola-font\Symbola-AjYx.ttf", 30],
# 'notoemoji': [r"D:\\Workspace\\streamlit-gpt3\\font\\NotoColorEmoji-Regular.ttf", 30],
# #'appleemoji': [r"E:\AppleColorEmoji.ttf", 137],
# #'segoeuiemoji': [r"E:\segoeuiemoji\seguiemj.ttf", 50]
# }
# pos = [(100, 10), (50, 40), (50, 160), (100, 320)]
# for i, item in enumerate(font_info.items()):
# path, size = item[1]
# font = ImageFont.FreeTypeFont(path, size)
# draw.text(pos[i], unicode_text, font=font, embedded_color=True)
# im.save('emoji.png')