Husnain
commited on
Commit
•
2f7e54b
1
Parent(s):
d860ba7
🔨 [WIP] Enabling get_message_id
Browse files- networks/huggingchat_streamer.py +153 -7
networks/huggingchat_streamer.py
CHANGED
@@ -1,8 +1,10 @@
|
|
|
|
1 |
import json
|
2 |
import re
|
3 |
import requests
|
|
|
4 |
|
5 |
-
|
6 |
from tclogger import logger
|
7 |
from transformers import AutoTokenizer
|
8 |
|
@@ -13,7 +15,11 @@ from constants.models import (
|
|
13 |
TOKEN_RESERVED,
|
14 |
)
|
15 |
from constants.envs import PROXIES
|
16 |
-
from constants.
|
|
|
|
|
|
|
|
|
17 |
from messagers.message_outputer import OpenaiStreamOutputer
|
18 |
|
19 |
|
@@ -25,7 +31,6 @@ class HuggingchatStreamer:
|
|
25 |
self.model = "mixtral-8x7b"
|
26 |
self.model_fullname = MODEL_MAP[self.model]
|
27 |
self.message_outputer = OpenaiStreamOutputer(model=self.model)
|
28 |
-
# export HF_ENDPOINT=https://hf-mirror.com
|
29 |
self.tokenizer = AutoTokenizer.from_pretrained(self.model_fullname)
|
30 |
|
31 |
def count_tokens(self, text):
|
@@ -34,16 +39,46 @@ class HuggingchatStreamer:
|
|
34 |
logger.note(f"Prompt Token Count: {token_count}")
|
35 |
return token_count
|
36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
def get_conversation_id(self, preprompt: str = ""):
|
38 |
request_url = "https://huggingface.co/chat/conversation"
|
|
|
|
|
|
|
|
|
|
|
39 |
request_body = {
|
40 |
"model": self.model_fullname,
|
41 |
"preprompt": preprompt,
|
42 |
}
|
43 |
logger.note(f"> Conversation ID:", end=" ")
|
|
|
44 |
res = requests.post(
|
45 |
request_url,
|
46 |
-
headers=
|
47 |
json=request_body,
|
48 |
proxies=PROXIES,
|
49 |
timeout=10,
|
@@ -55,6 +90,86 @@ class HuggingchatStreamer:
|
|
55 |
logger.warn(f"[{res.status_code}]")
|
56 |
raise ValueError("Failed to get conversation ID!")
|
57 |
self.conversation_id = conversation_id
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
def chat_response(
|
60 |
self,
|
@@ -65,7 +180,37 @@ class HuggingchatStreamer:
|
|
65 |
api_key: str = None,
|
66 |
use_cache: bool = False,
|
67 |
):
|
68 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
def chat_return_dict(self, stream_response):
|
71 |
pass
|
@@ -76,5 +221,6 @@ class HuggingchatStreamer:
|
|
76 |
|
77 |
if __name__ == "__main__":
|
78 |
streamer = HuggingchatStreamer(model="mixtral-8x7b")
|
79 |
-
|
80 |
-
|
|
|
|
1 |
+
import copy
|
2 |
import json
|
3 |
import re
|
4 |
import requests
|
5 |
+
import uuid
|
6 |
|
7 |
+
# from curl_cffi import requests
|
8 |
from tclogger import logger
|
9 |
from transformers import AutoTokenizer
|
10 |
|
|
|
15 |
TOKEN_RESERVED,
|
16 |
)
|
17 |
from constants.envs import PROXIES
|
18 |
+
from constants.headers import (
|
19 |
+
REQUESTS_HEADERS,
|
20 |
+
HUGGINGCHAT_POST_HEADERS,
|
21 |
+
HUGGINGCHAT_SETTINGS_POST_DATA,
|
22 |
+
)
|
23 |
from messagers.message_outputer import OpenaiStreamOutputer
|
24 |
|
25 |
|
|
|
31 |
self.model = "mixtral-8x7b"
|
32 |
self.model_fullname = MODEL_MAP[self.model]
|
33 |
self.message_outputer = OpenaiStreamOutputer(model=self.model)
|
|
|
34 |
self.tokenizer = AutoTokenizer.from_pretrained(self.model_fullname)
|
35 |
|
36 |
def count_tokens(self, text):
|
|
|
39 |
logger.note(f"Prompt Token Count: {token_count}")
|
40 |
return token_count
|
41 |
|
42 |
+
def get_hf_chat_id(self):
|
43 |
+
request_url = "https://huggingface.co/chat/settings"
|
44 |
+
request_body = copy.deepcopy(HUGGINGCHAT_SETTINGS_POST_DATA)
|
45 |
+
extra_body = {
|
46 |
+
"activeModel": self.model_fullname,
|
47 |
+
}
|
48 |
+
request_body.update(extra_body)
|
49 |
+
logger.note(f"> hf-chat ID:", end=" ")
|
50 |
+
|
51 |
+
res = requests.post(
|
52 |
+
request_url,
|
53 |
+
headers=HUGGINGCHAT_POST_HEADERS,
|
54 |
+
json=request_body,
|
55 |
+
proxies=PROXIES,
|
56 |
+
timeout=10,
|
57 |
+
)
|
58 |
+
self.hf_chat_id = res.cookies.get("hf-chat")
|
59 |
+
if self.hf_chat_id:
|
60 |
+
logger.success(f"[{self.hf_chat_id}]")
|
61 |
+
else:
|
62 |
+
logger.warn(f"[{res.status_code}]")
|
63 |
+
logger.warn(res.text)
|
64 |
+
raise ValueError("Failed to get hf-chat ID!")
|
65 |
+
|
66 |
def get_conversation_id(self, preprompt: str = ""):
|
67 |
request_url = "https://huggingface.co/chat/conversation"
|
68 |
+
request_headers = HUGGINGCHAT_POST_HEADERS
|
69 |
+
extra_headers = {
|
70 |
+
"Cookie": f"hf-chat={self.hf_chat_id}",
|
71 |
+
}
|
72 |
+
request_headers.update(extra_headers)
|
73 |
request_body = {
|
74 |
"model": self.model_fullname,
|
75 |
"preprompt": preprompt,
|
76 |
}
|
77 |
logger.note(f"> Conversation ID:", end=" ")
|
78 |
+
|
79 |
res = requests.post(
|
80 |
request_url,
|
81 |
+
headers=request_headers,
|
82 |
json=request_body,
|
83 |
proxies=PROXIES,
|
84 |
timeout=10,
|
|
|
90 |
logger.warn(f"[{res.status_code}]")
|
91 |
raise ValueError("Failed to get conversation ID!")
|
92 |
self.conversation_id = conversation_id
|
93 |
+
return conversation_id
|
94 |
+
|
95 |
+
def get_message_id(self):
|
96 |
+
request_url = f"https://huggingface.co/chat/conversation/{self.conversation_id}/__data.json?x-sveltekit-invalidated=11"
|
97 |
+
request_headers = HUGGINGCHAT_POST_HEADERS
|
98 |
+
extra_headers = {
|
99 |
+
"Cookie": f"hf-chat={self.hf_chat_id}",
|
100 |
+
}
|
101 |
+
request_headers.update(extra_headers)
|
102 |
+
logger.note(f"> Message ID:", end=" ")
|
103 |
+
|
104 |
+
message_id = None
|
105 |
+
res = requests.post(
|
106 |
+
request_url,
|
107 |
+
headers=request_headers,
|
108 |
+
proxies=PROXIES,
|
109 |
+
timeout=10,
|
110 |
+
)
|
111 |
+
if res.status_code == 200:
|
112 |
+
data = res.json()
|
113 |
+
# TODO - extract message_id
|
114 |
+
|
115 |
+
logger.success(f"[{message_id}]")
|
116 |
+
else:
|
117 |
+
logger.warn(f"[{res.status_code}]")
|
118 |
+
raise ValueError("Failed to get conversation ID!")
|
119 |
+
|
120 |
+
return message_id
|
121 |
+
|
122 |
+
def log_request(self, url, method="GET"):
|
123 |
+
logger.note(f"> {method}:", end=" ")
|
124 |
+
logger.mesg(f"{url}", end=" ")
|
125 |
+
|
126 |
+
def log_response(
|
127 |
+
self, res: requests.Response, stream=False, iter_lines=False, verbose=False
|
128 |
+
):
|
129 |
+
status_code = res.status_code
|
130 |
+
status_code_str = f"[{status_code}]"
|
131 |
+
|
132 |
+
if status_code == 200:
|
133 |
+
logger_func = logger.success
|
134 |
+
else:
|
135 |
+
logger_func = logger.warn
|
136 |
+
|
137 |
+
logger_func(status_code_str)
|
138 |
+
|
139 |
+
logger.enter_quiet(not verbose)
|
140 |
+
|
141 |
+
if status_code != 200:
|
142 |
+
logger_func(res.text)
|
143 |
+
|
144 |
+
if stream:
|
145 |
+
if not iter_lines:
|
146 |
+
return
|
147 |
+
|
148 |
+
for line in res.iter_lines():
|
149 |
+
line = line.decode("utf-8")
|
150 |
+
line = re.sub(r"^data:\s*", "", line)
|
151 |
+
line = line.strip()
|
152 |
+
if line:
|
153 |
+
try:
|
154 |
+
data = json.loads(line, strict=False)
|
155 |
+
msg_type = data.get("type")
|
156 |
+
if msg_type == "status":
|
157 |
+
msg_status = data.get("status")
|
158 |
+
elif msg_type == "stream":
|
159 |
+
content = data.get("token", "")
|
160 |
+
logger_func(content, end="")
|
161 |
+
elif msg_type == "finalAnswer":
|
162 |
+
full_content = data.get("text")
|
163 |
+
logger.success("\n[Finished]")
|
164 |
+
break
|
165 |
+
else:
|
166 |
+
pass
|
167 |
+
except Exception as e:
|
168 |
+
logger.warn(e)
|
169 |
+
else:
|
170 |
+
logger_func(res.json())
|
171 |
+
|
172 |
+
logger.exit_quiet(not verbose)
|
173 |
|
174 |
def chat_response(
|
175 |
self,
|
|
|
180 |
api_key: str = None,
|
181 |
use_cache: bool = False,
|
182 |
):
|
183 |
+
self.get_hf_chat_id()
|
184 |
+
self.get_conversation_id()
|
185 |
+
message_id = self.get_message_id()
|
186 |
+
|
187 |
+
request_url = f"https://huggingface.co/chat/conversation/{self.conversation_id}"
|
188 |
+
request_headers = copy.deepcopy(HUGGINGCHAT_POST_HEADERS)
|
189 |
+
extra_headers = {
|
190 |
+
"Content-Type": "text/event-stream",
|
191 |
+
"Referer": request_url,
|
192 |
+
"Cookie": f"hf-chat={self.hf_chat_id}",
|
193 |
+
}
|
194 |
+
request_headers.update(extra_headers)
|
195 |
+
request_body = {
|
196 |
+
"files": [],
|
197 |
+
"id": message_id,
|
198 |
+
"inputs": prompt,
|
199 |
+
"is_continue": False,
|
200 |
+
"is_retry": False,
|
201 |
+
"web_search": False,
|
202 |
+
}
|
203 |
+
self.log_request(request_url, method="POST")
|
204 |
+
|
205 |
+
res = requests.post(
|
206 |
+
request_url,
|
207 |
+
headers=request_headers,
|
208 |
+
json=request_body,
|
209 |
+
proxies=PROXIES,
|
210 |
+
stream=True,
|
211 |
+
)
|
212 |
+
self.log_response(res, stream=True, iter_lines=True, verbose=True)
|
213 |
+
return res
|
214 |
|
215 |
def chat_return_dict(self, stream_response):
|
216 |
pass
|
|
|
221 |
|
222 |
if __name__ == "__main__":
|
223 |
streamer = HuggingchatStreamer(model="mixtral-8x7b")
|
224 |
+
prompt = "who are you?"
|
225 |
+
streamer.chat_response(prompt=prompt)
|
226 |
+
# HF_ENDPOINT=https://hf-mirror.com python -m networks.huggingchat_streamer
|