Update app.py
Browse files
app.py
CHANGED
@@ -1,4 +1,211 @@
|
|
|
|
|
|
1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
def load_html():
|
4 |
with open('index.html','r') as h:
|
|
|
1 |
+
from huggingface_hub import InferenceClient, HfApi, upload_file
|
2 |
+
import datetime
|
3 |
import gradio as gr
|
4 |
+
import random
|
5 |
+
import prompts
|
6 |
+
import json
|
7 |
+
import uuid
|
8 |
+
import os
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
token=os.environ.get("HF_TOKEN")
|
13 |
+
username="omnibus"
|
14 |
+
dataset_name="tmp2"
|
15 |
+
api=HfApi(token="")
|
16 |
+
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
|
17 |
+
|
18 |
+
history = []
|
19 |
+
hist_out= []
|
20 |
+
summary =[]
|
21 |
+
main_point=[]
|
22 |
+
summary.append("")
|
23 |
+
main_point.append("")
|
24 |
+
|
25 |
+
def format_prompt(message, history):
|
26 |
+
prompt = "<s>"
|
27 |
+
for user_prompt, bot_response in history:
|
28 |
+
prompt += f"[INST] {user_prompt} [/INST]"
|
29 |
+
prompt += f" {bot_response}</s> "
|
30 |
+
prompt += f"[INST] {message} [/INST]"
|
31 |
+
return prompt
|
32 |
+
|
33 |
+
agents =[
|
34 |
+
"COMMENTER",
|
35 |
+
"BLOG_POSTER",
|
36 |
+
"COMPRESS_HISTORY_PROMPT"
|
37 |
+
]
|
38 |
+
|
39 |
+
temperature=0.9
|
40 |
+
max_new_tokens=256
|
41 |
+
max_new_tokens2=1048
|
42 |
+
top_p=0.95
|
43 |
+
repetition_penalty=1.0,
|
44 |
+
|
45 |
+
def compress_history(formatted_prompt):
|
46 |
+
|
47 |
+
seed = random.randint(1,1111111111111111)
|
48 |
+
agent=prompts.COMPRESS_HISTORY_PROMPT.format(history=summary[0],focus=main_point[0])
|
49 |
+
|
50 |
+
system_prompt=agent
|
51 |
+
temperature = 0.9
|
52 |
+
if temperature < 1e-2:
|
53 |
+
temperature = 1e-2
|
54 |
+
|
55 |
+
generate_kwargs = dict(
|
56 |
+
temperature=temperature,
|
57 |
+
max_new_tokens=30480,
|
58 |
+
top_p=0.95,
|
59 |
+
repetition_penalty=1.0,
|
60 |
+
do_sample=True,
|
61 |
+
seed=seed,
|
62 |
+
)
|
63 |
+
#history.append((prompt,""))
|
64 |
+
#formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
|
65 |
+
formatted_prompt = formatted_prompt
|
66 |
+
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
67 |
+
output = ""
|
68 |
+
|
69 |
+
for response in stream:
|
70 |
+
output += response.token.text
|
71 |
+
#history.append((output,history))
|
72 |
+
print(output)
|
73 |
+
print(main_point[0])
|
74 |
+
return output
|
75 |
+
|
76 |
+
|
77 |
+
def question_generate(prompt, history, agent_name=agents[0], sys_prompt="", temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,):
|
78 |
+
#def question_generate(prompt, history):
|
79 |
+
seed = random.randint(1,1111111111111111)
|
80 |
+
agent=prompts.COMMENTER.format(focus=main_point[0])
|
81 |
+
system_prompt=agent
|
82 |
+
temperature = float(temperature)
|
83 |
+
if temperature < 1e-2:
|
84 |
+
temperature = 1e-2
|
85 |
+
top_p = float(top_p)
|
86 |
+
|
87 |
+
generate_kwargs = dict(
|
88 |
+
temperature=temperature,
|
89 |
+
max_new_tokens=max_new_tokens,
|
90 |
+
top_p=top_p,
|
91 |
+
repetition_penalty=repetition_penalty,
|
92 |
+
do_sample=True,
|
93 |
+
seed=seed,
|
94 |
+
)
|
95 |
+
#history.append((prompt,""))
|
96 |
+
formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
|
97 |
+
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
98 |
+
output = ""
|
99 |
+
|
100 |
+
for response in stream:
|
101 |
+
output += response.token.text
|
102 |
+
#history.append((output,history))
|
103 |
+
|
104 |
+
return output
|
105 |
+
def create_valid_filename(invalid_filename: str) -> str:
|
106 |
+
"""Converts invalid characters in a string to be suitable for a filename."""
|
107 |
+
invalid_filename.replace(" ","-")
|
108 |
+
valid_chars = '-'.join(invalid_filename.split())
|
109 |
+
allowed_chars = ('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm',
|
110 |
+
'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z',
|
111 |
+
'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M',
|
112 |
+
'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z',
|
113 |
+
'0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '_', '-')
|
114 |
+
return ''.join(char for char in valid_chars if char in allowed_chars)
|
115 |
+
|
116 |
+
def generate(prompt, history, agent_name=agents[0], sys_prompt="", temperature=0.9, max_new_tokens=1048, top_p=0.95, repetition_penalty=1.0,):
|
117 |
+
main_point[0]=prompt
|
118 |
+
#print(datetime.datetime.now())
|
119 |
+
uid=uuid.uuid4()
|
120 |
+
current_time = str(datetime.datetime.now())
|
121 |
+
|
122 |
+
current_time=current_time.replace(":","-")
|
123 |
+
current_time=current_time.replace(".","-")
|
124 |
+
print (current_time)
|
125 |
+
agent=prompts.BLOG_POSTER
|
126 |
+
system_prompt=agent
|
127 |
+
temperature = float(temperature)
|
128 |
+
if temperature < 1e-2:
|
129 |
+
temperature = 1e-2
|
130 |
+
top_p = float(top_p)
|
131 |
+
hist_out=[]
|
132 |
+
sum_out=[]
|
133 |
+
json_hist={}
|
134 |
+
json_obj={}
|
135 |
+
filename=create_valid_filename(f'{prompt}---{current_time}')
|
136 |
+
while True:
|
137 |
+
seed = random.randint(1,1111111111111111)
|
138 |
+
|
139 |
+
generate_kwargs = dict(
|
140 |
+
temperature=temperature,
|
141 |
+
max_new_tokens=max_new_tokens2,
|
142 |
+
top_p=top_p,
|
143 |
+
repetition_penalty=repetition_penalty,
|
144 |
+
do_sample=True,
|
145 |
+
seed=seed,
|
146 |
+
)
|
147 |
+
if prompt.startswith(' \"'):
|
148 |
+
prompt=prompt.strip(' \"')
|
149 |
+
formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
|
150 |
+
if len(formatted_prompt) < (50000):
|
151 |
+
print(len(formatted_prompt))
|
152 |
+
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
153 |
+
output = ""
|
154 |
+
#if history:
|
155 |
+
# yield history
|
156 |
+
|
157 |
+
for response in stream:
|
158 |
+
output += response.token.text
|
159 |
+
yield '', [(prompt,output)],summary[0],json_obj, json_hist
|
160 |
+
out_json = {"prompt":prompt,"output":output}
|
161 |
+
|
162 |
+
prompt = question_generate(output, history)
|
163 |
+
#output += prompt
|
164 |
+
history.append((prompt,output))
|
165 |
+
print ( f'Prompt:: {len(prompt)}')
|
166 |
+
#print ( f'output:: {output}')
|
167 |
+
print ( f'history:: {len(formatted_prompt)}')
|
168 |
+
hist_out.append(out_json)
|
169 |
+
#try:
|
170 |
+
# for ea in
|
171 |
+
with open(f'{uid}.json', 'w') as f:
|
172 |
+
json_hist=json.dumps(hist_out, indent=4)
|
173 |
+
f.write(json_hist)
|
174 |
+
f.close()
|
175 |
+
|
176 |
+
upload_file(
|
177 |
+
path_or_fileobj =f"{uid}.json",
|
178 |
+
path_in_repo = f"test/{filename}.json",
|
179 |
+
repo_id =f"{username}/{dataset_name}",
|
180 |
+
repo_type = "dataset",
|
181 |
+
token=token,
|
182 |
+
)
|
183 |
+
else:
|
184 |
+
formatted_prompt = format_prompt(f"{prompts.COMPRESS_HISTORY_PROMPT.format(history=summary[0],focus=main_point[0])}, {summary[0]}", history)
|
185 |
+
|
186 |
+
#current_time = str(datetime.datetime.now().timestamp()).split(".",1)[0]
|
187 |
+
#filename=f'{filename}-{current_time}'
|
188 |
+
history = []
|
189 |
+
output = compress_history(formatted_prompt)
|
190 |
+
summary[0]=output
|
191 |
+
sum_json = {"summary":summary[0]}
|
192 |
+
sum_out.append(sum_json)
|
193 |
+
with open(f'{uid}-sum.json', 'w') as f:
|
194 |
+
json_obj=json.dumps(sum_out, indent=4)
|
195 |
+
f.write(json_obj)
|
196 |
+
f.close()
|
197 |
+
upload_file(
|
198 |
+
path_or_fileobj =f"{uid}-sum.json",
|
199 |
+
path_in_repo = f"summary/{filename}-summary.json",
|
200 |
+
repo_id =f"{username}/{dataset_name}",
|
201 |
+
repo_type = "dataset",
|
202 |
+
token=token,
|
203 |
+
)
|
204 |
+
|
205 |
+
|
206 |
+
prompt = question_generate(output, history)
|
207 |
+
|
208 |
+
return prompt, history, summary[0],json_obj,json_hist
|
209 |
|
210 |
def load_html():
|
211 |
with open('index.html','r') as h:
|