forked from aiwaves-cn/RecurrentGPT
-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
55 lines (44 loc) · 2.34 KB
/
main.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
from recurrentgpt import RecurrentGPT
from human_simulator import Human
import json
import argparse
from sentence_transformers import SentenceTransformer
from utils import get_init
if __name__ == '__main__':
parser = argparse.ArgumentParser(
description='ChatGPT-based automatic novel writing')
parser.add_argument('--iter', type=int, default=1)
parser.add_argument('--r_file', type=str, default='response.txt')
parser.add_argument('--init_prompt', type=str, default='init_prompt.json')
parser.add_argument('--type', type=str, default='science fiction')
parser.add_argument('--topic', type=str, default='')
args = parser.parse_args()
prompts = json.load(open(args.init_prompt,'r'))
init_prompt = prompts['init_prompt'].format(type=args.type,topic=args.topic)
# prepare first init(if there is no paragraph written)
init_paragraphs = get_init(init_text=None, text=init_prompt, response_file=args.r_file)
# print(init_paragraphs)
start_input_to__human = {
'output_paragraph': init_paragraphs['Paragraph 3'],
'input_paragraph': '\n'.join([init_paragraphs['Paragraph 1'], init_paragraphs['Paragraph 2']]),
'output_memory': init_paragraphs['Summary'],
"output_instruction": [init_paragraphs['Instruction 1'], init_paragraphs['Instruction 2'], init_paragraphs['Instruction 3']]
}
# Build the semantic search model
embedder = SentenceTransformer('multi-qa-mpnet-base-cos-v1')
human = Human(input=start_input_to__human, memory=None, embedder=embedder)
#select plan
human.input["output_instruction"] = human.select_plan(args.r_file)
print(human.input["output_instruction"])
human.step(args.r_file)
start_short_memory = init_paragraphs['Summary']
writer_start_input = human.output
# Init writerGPT
writer = RecurrentGPT(input=writer_start_input, short_memory=start_short_memory, long_memory=[
init_paragraphs['Paragraph 1'], init_paragraphs['Paragraph 2']], memory_index=None, embedder=embedder)
for i in range(args.iter):
writer.step(args.r_file) # write new paragraph and give instructions
human.input = writer.output # update human input
human.input["output_instruction"] = human.select_plan(args.r_file)
human.step(args.r_file)
writer.input = human.output # update writer input