Spaces:
Sleeping
Sleeping
File size: 6,267 Bytes
391cdfe f1218fc 391cdfe f1218fc 391cdfe f1218fc 391cdfe f1218fc 391cdfe f1218fc 391cdfe f1218fc 391cdfe f1218fc 391cdfe f1218fc 391cdfe f1218fc c706328 f1218fc 55b0c51 f1218fc 55b0c51 |
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 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 |
import copy
import json
import re
import requests
import uuid
# from curl_cffi import requests
from tclogger import logger
from transformers import AutoTokenizer
from constants.models import (
MODEL_MAP,
STOP_SEQUENCES_MAP,
TOKEN_LIMIT_MAP,
TOKEN_RESERVED,
)
from constants.envs import PROXIES
from constants.headers import (
REQUESTS_HEADERS,
HUGGINGCHAT_POST_HEADERS,
HUGGINGCHAT_SETTINGS_POST_DATA,
)
from messagers.message_outputer import OpenaiStreamOutputer
class HuggingchatStreamer:
def __init__(self, model: str):
if model in MODEL_MAP.keys():
self.model = model
else:
self.model = "mixtral-8x7b"
self.model_fullname = MODEL_MAP[self.model]
self.message_outputer = OpenaiStreamOutputer(model=self.model)
self.tokenizer = AutoTokenizer.from_pretrained(self.model_fullname)
def count_tokens(self, text):
tokens = self.tokenizer.encode(text)
token_count = len(tokens)
logger.note(f"Prompt Token Count: {token_count}")
return token_count
def get_hf_chat_id(self):
request_url = "https://huggingface.co/chat/settings"
request_body = copy.deepcopy(HUGGINGCHAT_SETTINGS_POST_DATA)
extra_body = {
"activeModel": self.model_fullname,
}
request_body.update(extra_body)
logger.note(f"> hf-chat ID:", end=" ")
res = requests.post(
request_url,
headers=HUGGINGCHAT_POST_HEADERS,
json=request_body,
proxies=PROXIES,
timeout=10,
)
self.hf_chat_id = res.cookies.get("hf-chat")
if self.hf_chat_id:
logger.success(f"[{self.hf_chat_id}]")
else:
logger.warn(f"[{res.status_code}]")
logger.warn(res.text)
raise ValueError("Failed to get hf-chat ID!")
def get_conversation_id(self, preprompt: str = ""):
request_url = "https://huggingface.co/chat/conversation"
request_headers = HUGGINGCHAT_POST_HEADERS
extra_headers = {
"Cookie": f"hf-chat={self.hf_chat_id}",
}
request_headers.update(extra_headers)
request_body = {
"model": self.model_fullname,
"preprompt": preprompt,
}
logger.note(f"> Conversation ID:", end=" ")
res = requests.post(
request_url,
headers=request_headers,
json=request_body,
proxies=PROXIES,
timeout=10,
)
if res.status_code == 200:
conversation_id = res.json()["conversationId"]
logger.success(f"[{conversation_id}]")
else:
logger.warn(f"[{res.status_code}]")
raise ValueError("Failed to get conversation ID!")
self.conversation_id = conversation_id
return conversation_id
def log_request(self, url, method="GET"):
logger.note(f"> {method}:", end=" ")
logger.mesg(f"{url}", end=" ")
def log_response(
self, res: requests.Response, stream=False, iter_lines=False, verbose=False
):
status_code = res.status_code
status_code_str = f"[{status_code}]"
if status_code == 200:
logger_func = logger.success
else:
logger_func = logger.warn
logger_func(status_code_str)
logger.enter_quiet(not verbose)
if status_code != 200:
logger_func(res.text)
if stream:
if not iter_lines:
return
for line in res.iter_lines():
line = line.decode("utf-8")
line = re.sub(r"^data:\s*", "", line)
line = line.strip()
if line:
try:
data = json.loads(line, strict=False)
msg_type = data.get("type")
if msg_type == "status":
msg_status = data.get("status")
elif msg_type == "stream":
content = data.get("token", "")
logger_func(content, end="")
elif msg_type == "finalAnswer":
full_content = data.get("text")
logger.success("\n[Finished]")
break
else:
pass
except Exception as e:
logger.warn(e)
else:
logger_func(res.json())
logger.exit_quiet(not verbose)
def chat_response(
self,
prompt: str = None,
temperature: float = 0.5,
top_p: float = 0.95,
max_new_tokens: int = None,
api_key: str = None,
use_cache: bool = False,
):
self.get_hf_chat_id()
self.get_conversation_id()
message_id = self.get_message_id()
request_url = f"https://huggingface.co/chat/conversation/{self.conversation_id}"
request_headers = copy.deepcopy(HUGGINGCHAT_POST_HEADERS)
extra_headers = {
"Content-Type": "text/event-stream",
"Referer": request_url,
"Cookie": f"hf-chat={self.hf_chat_id}",
}
request_headers.update(extra_headers)
request_body = {
"files": [],
"id": message_id,
"inputs": prompt,
"is_continue": False,
"is_retry": False,
"web_search": False,
}
self.log_request(request_url, method="POST")
res = requests.post(
request_url,
headers=request_headers,
json=request_body,
proxies=PROXIES,
stream=True,
)
self.log_response(res, stream=True, iter_lines=True, verbose=True)
return res
def chat_return_dict(self, stream_response):
pass
def chat_return_generator(self, stream_response):
pass
if __name__ == "__main__":
streamer = HuggingchatStreamer(model="mixtral-8x7b")
prompt = "who are you?"
streamer.chat_response(prompt=prompt)
# HF_ENDPOINT=https://hf-mirror.com python -m networks.huggingchat_streamer
|