File size: 10,778 Bytes
7c08dc3 |
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 |
from dotenv import load_dotenv
from utils.src.utils import ppt_to_images, get_json_from_response
import json
import shutil
from camel.models import ModelFactory
from camel.agents import ChatAgent
from utils.wei_utils import *
from camel.messages import BaseMessage
from PIL import Image
import pickle as pkl
from utils.pptx_utils import *
from utils.critic_utils import *
import yaml
from jinja2 import Environment, StrictUndefined
from pdf2image import convert_from_path
import argparse
load_dotenv()
def poster_apply_theme(args, actor_config, critic_config):
total_input_token, total_output_token = 0, 0
extract_input_token, extract_output_token = 0, 0
gen_input_token, gen_output_token = 0, 0
non_overlap_ckpt = pkl.load(open(f'checkpoints/{args.model_name}_{args.poster_name}_non_overlap_ckpt_{args.index}.pkl', 'rb'))
non_overlap_code = non_overlap_ckpt['final_code_by_section']
sections = list(non_overlap_code.keys())
sections = [s for s in sections if s != 'meta']
template_img = convert_from_path(args.template_path)[0]
image_bytes = io.BytesIO()
template_img.save(image_bytes, format="PNG")
image_bytes.seek(0)
# Reload the image from memory as a standard PIL.Image.Image
template_img = Image.open(image_bytes)
title_actor_agent_name = 'theme_agent_title'
with open(f"prompt_templates/{title_actor_agent_name}.yaml", "r") as f:
title_theme_actor_config = yaml.safe_load(f)
section_actor_agent_name = 'theme_agent_section'
with open(f"prompt_templates/{section_actor_agent_name}.yaml", "r") as f:
section_theme_actor_config = yaml.safe_load(f)
title_actor_model = ModelFactory.create(
model_platform=actor_config['model_platform'],
model_type=actor_config['model_type'],
model_config_dict=actor_config['model_config'], # [Optional] the config for model
)
title_actor_sys_msg = title_theme_actor_config['system_prompt']
title_actor_agent = ChatAgent(
system_message=title_actor_sys_msg,
model=title_actor_model,
message_window_size=10, # [Optional] the length for chat memory
)
section_actor_model = ModelFactory.create(
model_platform=actor_config['model_platform'],
model_type=actor_config['model_type'],
model_config_dict=actor_config['model_config'], # [Optional] the config for model
)
section_actor_sys_msg = section_theme_actor_config['system_prompt']
section_actor_agent = ChatAgent(
system_message=section_actor_sys_msg,
model=section_actor_model,
message_window_size=10, # [Optional] the length for chat memory
)
critic_model = ModelFactory.create(
model_platform=critic_config['model_platform'],
model_type=critic_config['model_type'],
model_config_dict=critic_config['model_config'],
)
critic_sys_msg = 'You are a helpful assistant.'
critic_agent = ChatAgent(
system_message=critic_sys_msg,
model=critic_model,
message_window_size=None,
)
theme_aspects = {
'background': ['background'],
'title': ['title_author', 'title_author_border'],
'section': ['section_body', 'section_title', 'section_border']
}
theme_styles = {}
for aspect in theme_aspects.keys():
theme_styles[aspect] = {}
for aspect, prompt_types in theme_aspects.items():
for prompt_type in prompt_types:
print(f'Getting style for {prompt_type}')
with open(f"prompt_templates/theme_templates/theme_{prompt_type}.txt", "r") as f:
prompt = f.read()
msg = BaseMessage.make_user_message(
role_name="User",
content=prompt,
image_list=[template_img],
)
critic_agent.reset()
response = critic_agent.step(msg)
input_token, output_token = account_token(response)
total_input_token += input_token
total_output_token += output_token
extract_input_token += input_token
extract_output_token += output_token
theme_style = get_json_from_response(response.msgs[0].content)
theme_styles[aspect][prompt_type] = theme_style
if 'fontStyle' in theme_styles['section']['section_body']:
del theme_styles['section']['section_body']['fontStyle']
outline_path = f'outlines/{args.model_name}_{args.poster_name}_outline_{args.index}.json'
outline = json.load(open(outline_path, 'r'))
outline_skeleton = {}
for key, val in outline.items():
if key == 'meta':
continue
if not 'subsections' in val:
outline_skeleton[key] = {
'section': key
}
else:
for subsection_name, subsection_dict in val['subsections'].items():
outline_skeleton[subsection_dict['name']] = {
'section': key
}
for key in outline_skeleton.keys():
if 'title' in key.lower() or 'author' in key.lower():
outline_skeleton[key]['style'] = theme_styles['section']['section_title']
else:
outline_skeleton[key]['style'] = theme_styles['section']['section_body']
outline_skeleton_list = []
for section in sections[1:]:
# append all subsections whose section key is the current section
for key, val in outline_skeleton.items():
if val['section'] == section:
outline_skeleton_list.append({key: val})
theme_logs = {}
theme_code = {}
concatenated_code = {}
# Title
jinja_env = Environment(undefined=StrictUndefined)
title_actor_template = jinja_env.from_string(title_theme_actor_config["template"])
# Title section
print(f'Processing section {sections[0]}')
curr_title_code = non_overlap_code[sections[0]]
for style in ['background', 'title']:
for sub_style in theme_styles[style].keys():
print(f' Applying theme for {sub_style}')
jinja_args = {
'style_json': {sub_style: theme_styles[style][sub_style]},
'function_docs': documentation,
'existing_code': curr_title_code
}
actor_prompt = title_actor_template.render(**jinja_args)
log = apply_theme(title_actor_agent, actor_prompt, args.max_retry, existing_code='')
if log[-1]['error'] is not None:
raise Exception(log[-1]['error'])
input_token, output_token = log[-1]['cumulative_tokens']
total_input_token += input_token
total_output_token += output_token
gen_input_token += input_token
gen_output_token += output_token
shutil.copy('poster.pptx', f'tmp/theme_poster_<{sections[0]}>_<{style}>_<{sub_style}>.pptx')
if not style in theme_logs:
theme_logs[style] = {}
theme_logs[style][sub_style] = log
curr_title_code = log[-1]['code']
theme_code[sections[0]] = curr_title_code
concatenated_code[sections[0]] = log[-1]['concatenated_code']
# Remaining sections
jinja_env = Environment(undefined=StrictUndefined)
section_actor_template = jinja_env.from_string(section_theme_actor_config["template"])
prev_section = None
for style_dict in outline_skeleton_list:
curr_subsection = list(style_dict.keys())[0]
curr_section = style_dict[curr_subsection]['section']
section_index = sections.index(curr_section)
print(f'Processing section {curr_section}')
if prev_section != curr_section:
prev_section = curr_section
curr_section_code = non_overlap_code[curr_section]
print(f' Applying theme for {curr_subsection}')
jinja_args = {
'style_json': json.dumps({curr_subsection: style_dict[curr_subsection]['style']}, indent=4),
'function_docs': documentation,
'existing_code': curr_section_code
}
actor_prompt = section_actor_template.render(**jinja_args)
existing_code = concatenated_code[sections[section_index - 1]]
log = apply_theme(section_actor_agent, actor_prompt, args.max_retry, existing_code=existing_code)
if log[-1]['error'] is not None:
raise Exception(log[-1]['error'])
input_token, output_token = log[-1]['cumulative_tokens']
total_input_token += input_token
total_output_token += output_token
gen_input_token += input_token
gen_output_token += output_token
shutil.copy('poster.pptx', f'tmp/theme_poster_<{curr_section}>_<{curr_subsection}>.pptx')
if not style in theme_logs:
theme_logs[style] = {}
theme_logs[style][sub_style] = log
curr_section_code = log[-1]['code']
theme_code[curr_section] = curr_section_code
concatenated_code[curr_section] = log[-1]['concatenated_code']
ppt_to_images(f'poster.pptx', 'tmp/theme_preview')
result_dir = f'results/{args.poster_name}/{args.model_name}/{args.index}'
shutil.copy('poster.pptx', f'{result_dir}/theme_poster.pptx')
ppt_to_images(f'poster.pptx', f'{result_dir}/theme_poster_preview')
ckpt = {
'theme_styles': theme_styles,
'theme_logs': theme_logs,
'theme_code': theme_code,
'concatenated_code': concatenated_code,
'total_input_token': total_input_token,
'total_output_token': total_output_token,
'extract_input_token': extract_input_token,
'extract_output_token': extract_output_token,
'gen_input_token': gen_input_token,
'gen_output_token': gen_output_token
}
pkl.dump(ckpt, open(f'checkpoints/{args.model_name}_{args.poster_name}_theme_ckpt.pkl', 'wb'))
return total_input_token, total_output_token
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--poster_name', type=str, default=None)
parser.add_argument('--model_name', type=str, default='4o')
parser.add_argument('--poster_path', type=str, required=True)
parser.add_argument('--index', type=int, default=0)
parser.add_argument('--template_path', type=str)
parser.add_argument('--max_retry', type=int, default=3)
args = parser.parse_args()
actor_config = get_agent_config(args.model_name)
critic_config = get_agent_config(args.model_name)
if args.poster_name is None:
args.poster_name = args.poster_path.split('/')[-1].replace('.pdf', '').replace(' ', '_')
input_token, output_token = poster_apply_theme(args, actor_config, critic_config)
print(f'Token consumption: {input_token} -> {output_token}') |