Update main.py
Browse files
main.py
CHANGED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import pickle
|
2 |
+
import random
|
3 |
+
import numpy as np
|
4 |
+
|
5 |
+
with open('models.pickle', 'rb')as f:
|
6 |
+
models = pickle.load(f)
|
7 |
+
|
8 |
+
LORA_TOKEN = ''#'<|>LORA_TOKEN<|>'
|
9 |
+
# WEIGHT_TOKEN = '<|>WEIGHT_TOKEN<|>'
|
10 |
+
NOT_SPLIT_TOKEN = '<|>NOT_SPLIT_TOKEN<|>'
|
11 |
+
|
12 |
+
def sample_next(ctx:str,model,k):
|
13 |
+
|
14 |
+
ctx = ', '.join(ctx.split(', ')[-k:])
|
15 |
+
if model.get(ctx) is None:
|
16 |
+
return " "
|
17 |
+
possible_Chars = list(model[ctx].keys())
|
18 |
+
possible_values = list(model[ctx].values())
|
19 |
+
|
20 |
+
# print(possible_Chars)
|
21 |
+
# print(possible_values)
|
22 |
+
|
23 |
+
return np.random.choice(possible_Chars,p=possible_values)
|
24 |
+
|
25 |
+
def generateText(model, minLen=100, size=5):
|
26 |
+
keys = list(model.keys())
|
27 |
+
starting_sent = random.choice(keys)
|
28 |
+
k = len(random.choice(keys).split(', '))
|
29 |
+
|
30 |
+
sentence = starting_sent
|
31 |
+
ctx = ', '.join(starting_sent.split(', ')[-k:])
|
32 |
+
|
33 |
+
while True:
|
34 |
+
next_prediction = sample_next(ctx,model,k)
|
35 |
+
sentence += f", {next_prediction}"
|
36 |
+
ctx = ', '.join(sentence.split(', ')[-k:])
|
37 |
+
# if sentence.count('\n')>size: break
|
38 |
+
if '\n' in sentence: break
|
39 |
+
sentence = sentence.replace(NOT_SPLIT_TOKEN, ', ')
|
40 |
+
# sentence = re.sub(WEIGHT_TOKEN.replace('|', '\|'), lambda match: f":{random.randint(0,2)}.{random.randint(0,9)}", sentence)
|
41 |
+
# sentence = sentence.replace(":0.0", ':0.1')
|
42 |
+
# return sentence
|
43 |
+
|
44 |
+
prompt = sentence.split('\n')[0]
|
45 |
+
if len(prompt)<minLen:
|
46 |
+
prompt = generateText(model, minLen, size=1)[0]
|
47 |
+
|
48 |
+
size = size-1
|
49 |
+
if size == 0: return [prompt]
|
50 |
+
output = []
|
51 |
+
for i in range(size+1):
|
52 |
+
prompt = generateText(model, minLen, size=1)[0]
|
53 |
+
output.append(prompt)
|
54 |
+
|
55 |
+
return output
|
56 |
+
if __name__ == "__main__":
|
57 |
+
for model in models: # models = [(model, neg_model), (nsfw, neg_nsfw), (sfw, neg_sfw)]
|
58 |
+
text = generateText(model[0], k=k, minLen=300, size=5)
|
59 |
+
text_neg = generateText(model[1], k=k, minLen=300, size=5)
|
60 |
+
|
61 |
+
# print('\n'.join(text))
|
62 |
+
for i in range(len(text)):
|
63 |
+
print(text[i])
|
64 |
+
# print('negativePrompt:')
|
65 |
+
print(text_neg[i])
|
66 |
+
print('----------------------------------------------------------------')
|
67 |
+
print('********************************************************************************************************************************************************')
|
68 |
+
|