Upload WarBot.py
Browse files
WarBot.py
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
from transformers import AutoTokenizer ,AutoModelForCausalLM
|
2 |
import re
|
3 |
# Speller and punctuation:
|
@@ -6,7 +8,7 @@ import yaml
|
|
6 |
import torch
|
7 |
from torch import package
|
8 |
# not very necessary
|
9 |
-
import textwrap
|
10 |
from textwrap3 import wrap
|
11 |
|
12 |
# util function to get expected len after tokenizing
|
@@ -65,7 +67,6 @@ def prepare_punct():
|
|
65 |
|
66 |
def initialize():
|
67 |
""" Loading the model """
|
68 |
-
torch.backends.quantized.engine = 'qnnpack' # Just for the specific machine architecture
|
69 |
fit_checkpoint = "WarBot"
|
70 |
tokenizer = AutoTokenizer.from_pretrained(fit_checkpoint)
|
71 |
model = AutoModelForCausalLM.from_pretrained(fit_checkpoint)
|
@@ -75,10 +76,13 @@ def initialize():
|
|
75 |
def split_string(string,n=256):
|
76 |
return [string[i:i+n] for i in range(0, len(string), n)]
|
77 |
|
78 |
-
def get_response(quote:str,model,tokenizer,model_punct):
|
79 |
# encode the input, add the eos_token and return a tensor in Pytorch
|
80 |
-
|
81 |
-
|
|
|
|
|
|
|
82 |
|
83 |
chat_history_ids = user_inpit_ids # To be changed
|
84 |
|
@@ -88,22 +92,22 @@ def get_response(quote:str,model,tokenizer,model_punct):
|
|
88 |
else:
|
89 |
no_repeat_ngram_size = 1
|
90 |
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
|
108 |
response = tokenizer.decode(output_id[0], skip_special_tokens=True)
|
109 |
response = removeSigns(response)
|
@@ -113,26 +117,35 @@ def get_response(quote:str,model,tokenizer,model_punct):
|
|
113 |
response = remove_duplicates(re.sub(r"\d{4,}", "", response)) # Remove the consequent numbers with 4 or more digits
|
114 |
response = re.sub(r'\.\.+', '', response) # Remove the "....." thing
|
115 |
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
resps = wrap(response,maxLen)
|
121 |
-
for i in range(len(resps)):
|
122 |
resps[i] = model_punct.enhance_text(resps[i], lan='ru')
|
123 |
response = ''.join(resps)
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
|
|
|
129 |
response = re.sub(r'[UNK]', '', response) # Remove the [UNK] thing
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
130 |
return response
|
131 |
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
|
|
|
|
|
1 |
+
# Main library for WarBot
|
2 |
+
|
3 |
from transformers import AutoTokenizer ,AutoModelForCausalLM
|
4 |
import re
|
5 |
# Speller and punctuation:
|
|
|
8 |
import torch
|
9 |
from torch import package
|
10 |
# not very necessary
|
11 |
+
#import textwrap
|
12 |
from textwrap3 import wrap
|
13 |
|
14 |
# util function to get expected len after tokenizing
|
|
|
67 |
|
68 |
def initialize():
|
69 |
""" Loading the model """
|
|
|
70 |
fit_checkpoint = "WarBot"
|
71 |
tokenizer = AutoTokenizer.from_pretrained(fit_checkpoint)
|
72 |
model = AutoModelForCausalLM.from_pretrained(fit_checkpoint)
|
|
|
76 |
def split_string(string,n=256):
|
77 |
return [string[i:i+n] for i in range(0, len(string), n)]
|
78 |
|
79 |
+
def get_response(quote:str,model,tokenizer,model_punct,temperature=0.2):
|
80 |
# encode the input, add the eos_token and return a tensor in Pytorch
|
81 |
+
try:
|
82 |
+
user_inpit_ids = tokenizer.encode(f"|0|{get_length_param(quote, tokenizer)}|" \
|
83 |
+
+ quote + tokenizer.eos_token, return_tensors="pt")
|
84 |
+
except:
|
85 |
+
return "" # Exception in tokenization
|
86 |
|
87 |
chat_history_ids = user_inpit_ids # To be changed
|
88 |
|
|
|
92 |
else:
|
93 |
no_repeat_ngram_size = 1
|
94 |
|
95 |
+
try:
|
96 |
+
output_id = model.generate(
|
97 |
+
chat_history_ids,
|
98 |
+
num_return_sequences=1, # use for more variants, but have to print [i]
|
99 |
+
max_length=200, #512
|
100 |
+
no_repeat_ngram_size=no_repeat_ngram_size, #3
|
101 |
+
do_sample=True, #True
|
102 |
+
top_k=50,#50
|
103 |
+
top_p=0.9, #0.9
|
104 |
+
temperature = temperature, # was 0.6, 0 for greedy
|
105 |
+
eos_token_id=tokenizer.eos_token_id,
|
106 |
+
pad_token_id=tokenizer.pad_token_id,
|
107 |
+
#device='cpu'
|
108 |
+
)
|
109 |
+
except:
|
110 |
+
return "" # Exception in generation
|
111 |
|
112 |
response = tokenizer.decode(output_id[0], skip_special_tokens=True)
|
113 |
response = removeSigns(response)
|
|
|
117 |
response = remove_duplicates(re.sub(r"\d{4,}", "", response)) # Remove the consequent numbers with 4 or more digits
|
118 |
response = re.sub(r'\.\.+', '', response) # Remove the "....." thing
|
119 |
|
120 |
+
if len(response)>200:
|
121 |
+
resps = wrap(response,200)
|
122 |
+
for i in range(len(resps)):
|
123 |
+
try:
|
|
|
|
|
124 |
resps[i] = model_punct.enhance_text(resps[i], lan='ru')
|
125 |
response = ''.join(resps)
|
126 |
+
except:
|
127 |
+
return "" # Excepion in punctuation
|
128 |
+
else:
|
129 |
+
response = model_punct.enhance_text(response, lan='ru')
|
130 |
|
131 |
+
# Immanent postprocessing of the response
|
132 |
response = re.sub(r'[UNK]', '', response) # Remove the [UNK] thing
|
133 |
+
response = re.sub(r',+', ',', response) # Replace multi-commas with single one
|
134 |
+
response = re.sub(r'-+', ',', response) # Replace multi-dashes with single one
|
135 |
+
response = re.sub(r'\.\?', '?', response) # Fix the .? issue
|
136 |
+
response = re.sub(r'\.\!', '!', response) # Fix the .! issue
|
137 |
+
response = re.sub(r'\.\,', ',', response) # Fix the ,. issue
|
138 |
+
response = re.sub(r'\.\)', '.', response) # Fix the .) issue
|
139 |
+
response = response.replace('[]', '') # Fix the [] issue
|
140 |
+
|
141 |
return response
|
142 |
|
143 |
+
if __name__ == '__main__':
|
144 |
+
"""
|
145 |
+
quote = "Здравствуй, Жопа, Новый Год, выходи на ёлку!"
|
146 |
+
model, tokenizer, model_punct = initialize()
|
147 |
+
response = ""
|
148 |
+
while not response:
|
149 |
+
response = get_response(quote, model, tokenizer, model_punct,temperature=0.2)
|
150 |
+
print(response)
|
151 |
+
"""
|