File size: 44,536 Bytes
1ea2ba0 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 |
from __future__ import annotations
import base64
import json
import time
import logging
import os
import shutil
import time
import traceback
from collections import deque
from enum import Enum
from io import BytesIO
from itertools import islice
from threading import Condition, Thread
from typing import Any, Dict, List, Optional
import colorama
import PIL
import urllib3
from duckduckgo_search import DDGS
from huggingface_hub import hf_hub_download
from langchain.callbacks.base import BaseCallbackHandler
from langchain.chat_models.base import BaseChatModel
from langchain.schema import (AgentAction, AgentFinish, AIMessage, BaseMessage,
HumanMessage, SystemMessage)
from .. import shared
from ..config import retrieve_proxy
from ..index_func import *
from ..presets import *
from ..utils import *
class CallbackToIterator:
def __init__(self):
self.queue = deque()
self.cond = Condition()
self.finished = False
def callback(self, result):
with self.cond:
self.queue.append(result)
self.cond.notify() # Wake up the generator.
def __iter__(self):
return self
def __next__(self):
with self.cond:
# Wait for a value to be added to the queue.
while not self.queue and not self.finished:
self.cond.wait()
if not self.queue:
raise StopIteration()
return self.queue.popleft()
def finish(self):
with self.cond:
self.finished = True
self.cond.notify() # Wake up the generator if it's waiting.
def get_action_description(action):
action_name = action.tool
action_name = " ".join(action_name.split("_")).title()
action_input = action.tool_input
if isinstance(action_input, dict):
action_input = " ".join(action_input.values())
if action_name != "Final Answer":
return f'<!-- S O PREFIX --><p class="agent-prefix">{action_name}: {action_input}\n</p><!-- E O PREFIX -->'
else:
return ""
class ChuanhuCallbackHandler(BaseCallbackHandler):
def __init__(self, callback) -> None:
"""Initialize callback handler."""
self.callback = callback
def on_agent_action(
self, action: AgentAction, color: Optional[str] = None, **kwargs: Any
) -> Any:
self.callback(get_action_description(action))
def on_tool_end(
self,
output: str,
color: Optional[str] = None,
observation_prefix: Optional[str] = None,
llm_prefix: Optional[str] = None,
**kwargs: Any,
) -> None:
"""If not the final action, print out observation."""
# if observation_prefix is not None:
# self.callback(f"\n\n{observation_prefix}")
# self.callback(output)
# if llm_prefix is not None:
# self.callback(f"\n\n{llm_prefix}")
if observation_prefix is not None:
logging.info(observation_prefix)
self.callback(output)
if llm_prefix is not None:
logging.info(llm_prefix)
def on_agent_finish(
self, finish: AgentFinish, color: Optional[str] = None, **kwargs: Any
) -> None:
# self.callback(f"{finish.log}\n\n")
logging.info(finish.log)
def on_llm_new_token(self, token: str, **kwargs: Any) -> None:
"""Run on new LLM token. Only available when streaming is enabled."""
self.callback(token)
def on_chat_model_start(
self,
serialized: Dict[str, Any],
messages: List[List[BaseMessage]],
**kwargs: Any,
) -> Any:
"""Run when a chat model starts running."""
pass
class ModelType(Enum):
Unknown = -1
OpenAI = 0
ChatGLM = 1
LLaMA = 2
XMChat = 3
StableLM = 4
MOSS = 5
YuanAI = 6
Minimax = 7
ChuanhuAgent = 8
GooglePaLM = 9
LangchainChat = 10
Midjourney = 11
Spark = 12
OpenAIInstruct = 13
Claude = 14
Qwen = 15
OpenAIVision = 16
ERNIE = 17
DALLE3 = 18
GoogleGemini = 19
GoogleGemma = 20
Ollama = 21
@classmethod
def get_type(cls, model_name: str):
model_type = None
model_name_lower = model_name.lower()
if "gpt" in model_name_lower:
if "instruct" in model_name_lower:
model_type = ModelType.OpenAIInstruct
elif "vision" in model_name_lower:
model_type = ModelType.OpenAIVision
else:
model_type = ModelType.OpenAI
elif "chatglm" in model_name_lower:
model_type = ModelType.ChatGLM
elif "ollama" in model_name_lower:
model_type = ModelType.Ollama
elif "llama" in model_name_lower or "alpaca" in model_name_lower:
model_type = ModelType.LLaMA
elif "xmchat" in model_name_lower:
model_type = ModelType.XMChat
elif "stablelm" in model_name_lower:
model_type = ModelType.StableLM
elif "moss" in model_name_lower:
model_type = ModelType.MOSS
elif "yuanai" in model_name_lower:
model_type = ModelType.YuanAI
elif "minimax" in model_name_lower:
model_type = ModelType.Minimax
elif "川虎助理" in model_name_lower:
model_type = ModelType.ChuanhuAgent
elif "palm" in model_name_lower:
model_type = ModelType.GooglePaLM
elif "gemini" in model_name_lower:
model_type = ModelType.GoogleGemini
elif "midjourney" in model_name_lower:
model_type = ModelType.Midjourney
elif "azure" in model_name_lower or "api" in model_name_lower:
model_type = ModelType.LangchainChat
elif "星火大模型" in model_name_lower:
model_type = ModelType.Spark
elif "claude" in model_name_lower:
model_type = ModelType.Claude
elif "qwen" in model_name_lower:
model_type = ModelType.Qwen
elif "ernie" in model_name_lower:
model_type = ModelType.ERNIE
elif "dall" in model_name_lower:
model_type = ModelType.DALLE3
elif "gemma" in model_name_lower:
model_type = ModelType.GoogleGemma
else:
model_type = ModelType.LLaMA
return model_type
def download(repo_id, filename, retry=10):
if os.path.exists("./models/downloaded_models.json"):
with open("./models/downloaded_models.json", "r") as f:
downloaded_models = json.load(f)
if repo_id in downloaded_models:
return downloaded_models[repo_id]["path"]
else:
downloaded_models = {}
while retry > 0:
try:
model_path = hf_hub_download(
repo_id=repo_id,
filename=filename,
cache_dir="models",
resume_download=True,
)
downloaded_models[repo_id] = {"path": model_path}
with open("./models/downloaded_models.json", "w") as f:
json.dump(downloaded_models, f)
break
except:
print("Error downloading model, retrying...")
retry -= 1
if retry == 0:
raise Exception("Error downloading model, please try again later.")
return model_path
class BaseLLMModel:
def __init__(
self,
model_name,
system_prompt=INITIAL_SYSTEM_PROMPT,
temperature=1.0,
top_p=1.0,
n_choices=1,
stop=[],
max_generation_token=None,
presence_penalty=0,
frequency_penalty=0,
logit_bias=None,
user="",
single_turn=False,
) -> None:
self.history = []
self.all_token_counts = []
self.model_type = ModelType.get_type(model_name)
try:
self.model_name = MODEL_METADATA[model_name]["model_name"]
except:
self.model_name = model_name
try:
self.multimodal = MODEL_METADATA[model_name]["multimodal"]
except:
self.multimodal = False
if max_generation_token is None:
try:
max_generation_token = MODEL_METADATA[model_name]["max_generation"]
except:
pass
try:
self.token_upper_limit = MODEL_METADATA[model_name]["token_limit"]
except KeyError:
self.token_upper_limit = DEFAULT_TOKEN_LIMIT
self.interrupted = False
self.system_prompt = system_prompt
self.api_key = None
self.need_api_key = False
self.history_file_path = get_first_history_name(user)
self.user_name = user
self.chatbot = []
self.default_single_turn = single_turn
self.default_temperature = temperature
self.default_top_p = top_p
self.default_n_choices = n_choices
self.default_stop_sequence = stop
self.default_max_generation_token = max_generation_token
self.default_presence_penalty = presence_penalty
self.default_frequency_penalty = frequency_penalty
self.default_logit_bias = logit_bias
self.default_user_identifier = user
self.single_turn = single_turn
self.temperature = temperature
self.top_p = top_p
self.n_choices = n_choices
self.stop_sequence = stop
self.max_generation_token = max_generation_token
self.presence_penalty = presence_penalty
self.frequency_penalty = frequency_penalty
self.logit_bias = logit_bias
self.user_identifier = user
self.metadata = {}
def get_answer_stream_iter(self):
"""Implement stream prediction.
Conversations are stored in self.history, with the most recent question in OpenAI format.
Should return a generator that yields the next word (str) in the answer.
"""
logging.warning(
"Stream prediction is not implemented. Using at once prediction instead."
)
response, _ = self.get_answer_at_once()
yield response
def get_answer_at_once(self):
"""predict at once, need to be implemented
conversations are stored in self.history, with the most recent question, in OpenAI format
Should return:
the answer (str)
total token count (int)
"""
logging.warning("at once predict not implemented, using stream predict instead")
response_iter = self.get_answer_stream_iter()
count = 0
for response in response_iter:
count += 1
return response, sum(self.all_token_counts) + count
def billing_info(self):
"""get billing infomation, inplement if needed"""
# logging.warning("billing info not implemented, using default")
return BILLING_NOT_APPLICABLE_MSG
def count_token(self, user_input):
"""get token count from input, implement if needed"""
# logging.warning("token count not implemented, using default")
return len(user_input)
def stream_next_chatbot(self, inputs, chatbot, fake_input=None, display_append=""):
def get_return_value():
return chatbot, status_text
status_text = i18n("开始实时传输回答……")
if fake_input:
chatbot.append((fake_input, ""))
else:
chatbot.append((inputs, ""))
user_token_count = self.count_token(inputs)
self.all_token_counts.append(user_token_count)
logging.debug(f"输入token计数: {user_token_count}")
stream_iter = self.get_answer_stream_iter()
if display_append:
display_append = (
'\n\n<hr class="append-display no-in-raw" />' + display_append
)
partial_text = ""
token_increment = 1
for partial_text in stream_iter:
if type(partial_text) == tuple:
partial_text, token_increment = partial_text
chatbot[-1] = (chatbot[-1][0], partial_text + display_append)
self.all_token_counts[-1] += token_increment
status_text = self.token_message()
yield get_return_value()
if self.interrupted:
self.recover()
break
self.history.append(construct_assistant(partial_text))
def next_chatbot_at_once(self, inputs, chatbot, fake_input=None, display_append=""):
if fake_input:
chatbot.append((fake_input, ""))
else:
chatbot.append((inputs, ""))
if fake_input is not None:
user_token_count = self.count_token(fake_input)
else:
user_token_count = self.count_token(inputs)
self.all_token_counts.append(user_token_count)
ai_reply, total_token_count = self.get_answer_at_once()
self.history.append(construct_assistant(ai_reply))
if fake_input is not None:
self.history[-2] = construct_user(fake_input)
chatbot[-1] = (chatbot[-1][0], ai_reply + display_append)
if fake_input is not None:
self.all_token_counts[-1] += count_token(construct_assistant(ai_reply))
else:
self.all_token_counts[-1] = total_token_count - sum(self.all_token_counts)
status_text = self.token_message()
return chatbot, status_text
def handle_file_upload(self, files, chatbot, language):
"""if the model accepts multi modal input, implement this function"""
status = gr.Markdown()
image_files = []
other_files = []
if files:
for f in files:
if f.name.endswith(IMAGE_FORMATS):
image_files.append(f)
else:
other_files.append(f)
if image_files:
if self.multimodal:
chatbot.extend([(((image.name, None)), None) for image in image_files])
self.history.extend([construct_image(image.name) for image in image_files])
else:
gr.Warning(i18n("该模型不支持多模态输入"))
if other_files:
try:
construct_index(self.api_key, file_src=files)
status = i18n("索引构建完成")
except Exception as e:
import traceback
traceback.print_exc()
status = i18n("索引构建失败!") + str(e)
if other_files:
other_files = [f.name for f in other_files]
else:
other_files = None
return gr.File(value=other_files), chatbot, status
def summarize_index(self, files, chatbot, language):
status = gr.Markdown()
if files:
index = construct_index(self.api_key, file_src=files)
status = i18n("总结完成")
logging.info(i18n("生成内容总结中……"))
os.environ["OPENAI_API_KEY"] = self.api_key
from langchain.callbacks import StdOutCallbackHandler
from langchain.chains.summarize import load_summarize_chain
from langchain.chat_models import ChatOpenAI
from langchain.prompts import PromptTemplate
prompt_template = (
"Write a concise summary of the following:\n\n{text}\n\nCONCISE SUMMARY IN "
+ language
+ ":"
)
PROMPT = PromptTemplate(template=prompt_template, input_variables=["text"])
llm = ChatOpenAI()
chain = load_summarize_chain(
llm,
chain_type="map_reduce",
return_intermediate_steps=True,
map_prompt=PROMPT,
combine_prompt=PROMPT,
)
summary = chain(
{"input_documents": list(index.docstore.__dict__["_dict"].values())},
return_only_outputs=True,
)["output_text"]
print(i18n("总结") + f": {summary}")
chatbot.append([i18n("上传了") + str(len(files)) + "个文件", summary])
return chatbot, status
def prepare_inputs(
self,
real_inputs,
use_websearch,
files,
reply_language,
chatbot,
load_from_cache_if_possible=True,
):
display_append = []
limited_context = False
if type(real_inputs) == list:
fake_inputs = real_inputs[0]["text"]
else:
fake_inputs = real_inputs
if files:
from langchain.embeddings.huggingface import HuggingFaceEmbeddings
from langchain.vectorstores.base import VectorStoreRetriever
limited_context = True
msg = "加载索引中……"
logging.info(msg)
index = construct_index(
self.api_key,
file_src=files,
load_from_cache_if_possible=load_from_cache_if_possible,
)
assert index is not None, "获取索引失败"
msg = "索引获取成功,生成回答中……"
logging.info(msg)
with retrieve_proxy():
retriever = VectorStoreRetriever(
vectorstore=index, search_type="similarity", search_kwargs={"k": 6}
)
# retriever = VectorStoreRetriever(vectorstore=index, search_type="similarity_score_threshold", search_kwargs={
# "k": 6, "score_threshold": 0.2})
try:
relevant_documents = retriever.get_relevant_documents(fake_inputs)
except AssertionError:
return self.prepare_inputs(
fake_inputs,
use_websearch,
files,
reply_language,
chatbot,
load_from_cache_if_possible=False,
)
reference_results = [
[d.page_content.strip("�"), os.path.basename(d.metadata["source"])]
for d in relevant_documents
]
reference_results = add_source_numbers(reference_results)
display_append = add_details(reference_results)
display_append = "\n\n" + "".join(display_append)
if type(real_inputs) == list:
real_inputs[0]["text"] = (
replace_today(PROMPT_TEMPLATE)
.replace("{query_str}", fake_inputs)
.replace("{context_str}", "\n\n".join(reference_results))
.replace("{reply_language}", reply_language)
)
else:
real_inputs = (
replace_today(PROMPT_TEMPLATE)
.replace("{query_str}", real_inputs)
.replace("{context_str}", "\n\n".join(reference_results))
.replace("{reply_language}", reply_language)
)
elif use_websearch:
search_results = []
with retrieve_proxy() as proxy:
if proxy[0] or proxy[1]:
proxies = {}
if proxy[0]:
proxies["http"] = proxy[0]
if proxy[1]:
proxies["https"] = proxy[1]
else:
proxies = None
with DDGS(proxies=proxies) as ddgs:
ddgs_gen = ddgs.text(fake_inputs, backend="lite")
for r in islice(ddgs_gen, 10):
search_results.append(r)
reference_results = []
for idx, result in enumerate(search_results):
logging.debug(f"搜索结果{idx + 1}:{result}")
domain_name = urllib3.util.parse_url(result["href"]).host
reference_results.append([result["body"], result["href"]])
display_append.append(
# f"{idx+1}. [{domain_name}]({result['href']})\n"
f"<a href=\"{result['href']}\" target=\"_blank\">{idx+1}. {result['title']}</a>"
)
reference_results = add_source_numbers(reference_results)
# display_append = "<ol>\n\n" + "".join(display_append) + "</ol>"
display_append = (
'<div class = "source-a">' + "".join(display_append) + "</div>"
)
if type(real_inputs) == list:
real_inputs[0]["text"] = (
replace_today(WEBSEARCH_PTOMPT_TEMPLATE)
.replace("{query}", fake_inputs)
.replace("{web_results}", "\n\n".join(reference_results))
.replace("{reply_language}", reply_language)
)
else:
real_inputs = (
replace_today(WEBSEARCH_PTOMPT_TEMPLATE)
.replace("{query}", fake_inputs)
.replace("{web_results}", "\n\n".join(reference_results))
.replace("{reply_language}", reply_language)
)
else:
display_append = ""
return limited_context, fake_inputs, display_append, real_inputs, chatbot
def predict(
self,
inputs,
chatbot,
stream=False,
use_websearch=False,
files=None,
reply_language="中文",
should_check_token_count=True,
): # repetition_penalty, top_k
status_text = "开始生成回答……"
if type(inputs) == list:
logging.info(
"用户"
+ f"{self.user_name}"
+ "的输入为:"
+ colorama.Fore.BLUE
+ "("
+ str(len(inputs) - 1)
+ " images) "
+ f"{inputs[0]['text']}"
+ colorama.Style.RESET_ALL
)
else:
logging.info(
"用户"
+ f"{self.user_name}"
+ "的输入为:"
+ colorama.Fore.BLUE
+ f"{inputs}"
+ colorama.Style.RESET_ALL
)
if should_check_token_count:
if type(inputs) == list:
yield chatbot + [(inputs[0]["text"], "")], status_text
else:
yield chatbot + [(inputs, "")], status_text
if reply_language == "跟随问题语言(不稳定)":
reply_language = "the same language as the question, such as English, 中文, 日本語, Español, Français, or Deutsch."
(
limited_context,
fake_inputs,
display_append,
inputs,
chatbot,
) = self.prepare_inputs(
real_inputs=inputs,
use_websearch=use_websearch,
files=files,
reply_language=reply_language,
chatbot=chatbot,
)
yield chatbot + [(fake_inputs, "")], status_text
if (
self.need_api_key
and self.api_key is None
and not shared.state.multi_api_key
):
status_text = STANDARD_ERROR_MSG + NO_APIKEY_MSG
logging.info(status_text)
chatbot.append((fake_inputs, ""))
if len(self.history) == 0:
self.history.append(construct_user(fake_inputs))
self.history.append("")
self.all_token_counts.append(0)
else:
self.history[-2] = construct_user(fake_inputs)
yield chatbot + [(fake_inputs, "")], status_text
return
elif len(fake_inputs.strip()) == 0:
status_text = STANDARD_ERROR_MSG + NO_INPUT_MSG
logging.info(status_text)
yield chatbot + [(fake_inputs, "")], status_text
return
if self.single_turn:
self.history = []
self.all_token_counts = []
if type(inputs) == list:
self.history.append(inputs)
else:
self.history.append(construct_user(inputs))
start_time = time.time()
try:
if stream:
logging.debug("使用流式传输")
iter = self.stream_next_chatbot(
inputs,
chatbot,
fake_input=fake_inputs,
display_append=display_append,
)
for chatbot, status_text in iter:
yield chatbot, status_text
else:
logging.debug("不使用流式传输")
chatbot, status_text = self.next_chatbot_at_once(
inputs,
chatbot,
fake_input=fake_inputs,
display_append=display_append,
)
yield chatbot, status_text
except Exception as e:
traceback.print_exc()
status_text = STANDARD_ERROR_MSG + beautify_err_msg(str(e))
yield chatbot, status_text
end_time = time.time()
if len(self.history) > 1 and self.history[-1]["content"] != fake_inputs:
logging.info(
"回答为:"
+ colorama.Fore.BLUE
+ f"{self.history[-1]['content']}"
+ colorama.Style.RESET_ALL
)
logging.info(i18n("Tokens per second:{token_generation_speed}").format(token_generation_speed=str(self.all_token_counts[-1] / (end_time - start_time))))
if limited_context:
# self.history = self.history[-4:]
# self.all_token_counts = self.all_token_counts[-2:]
self.history = []
self.all_token_counts = []
max_token = self.token_upper_limit - TOKEN_OFFSET
if sum(self.all_token_counts) > max_token and should_check_token_count:
count = 0
while (
sum(self.all_token_counts)
> self.token_upper_limit * REDUCE_TOKEN_FACTOR
and sum(self.all_token_counts) > 0
):
count += 1
del self.all_token_counts[0]
del self.history[:2]
logging.info(status_text)
status_text = f"为了防止token超限,模型忘记了早期的 {count} 轮对话"
yield chatbot, status_text
self.chatbot = chatbot
self.auto_save(chatbot)
def retry(
self,
chatbot,
stream=False,
use_websearch=False,
files=None,
reply_language="中文",
):
logging.debug("重试中……")
if len(self.history) > 1:
inputs = self.history[-2]["content"]
del self.history[-2:]
if len(self.all_token_counts) > 0:
self.all_token_counts.pop()
elif len(chatbot) > 0:
inputs = chatbot[-1][0]
if '<div class="user-message">' in inputs:
inputs = inputs.split('<div class="user-message">')[1]
inputs = inputs.split("</div>")[0]
elif len(self.history) == 1:
inputs = self.history[-1]["content"]
del self.history[-1]
else:
yield chatbot, f"{STANDARD_ERROR_MSG}上下文是空的"
return
iter = self.predict(
inputs,
chatbot,
stream=stream,
use_websearch=use_websearch,
files=files,
reply_language=reply_language,
)
for x in iter:
yield x
logging.debug("重试完毕")
# def reduce_token_size(self, chatbot):
# logging.info("开始减少token数量……")
# chatbot, status_text = self.next_chatbot_at_once(
# summarize_prompt,
# chatbot
# )
# max_token_count = self.token_upper_limit * REDUCE_TOKEN_FACTOR
# num_chat = find_n(self.all_token_counts, max_token_count)
# logging.info(f"previous_token_count: {self.all_token_counts}, keeping {num_chat} chats")
# chatbot = chatbot[:-1]
# self.history = self.history[-2*num_chat:] if num_chat > 0 else []
# self.all_token_counts = self.all_token_counts[-num_chat:] if num_chat > 0 else []
# msg = f"保留了最近{num_chat}轮对话"
# logging.info(msg)
# logging.info("减少token数量完毕")
# return chatbot, msg + "," + self.token_message(self.all_token_counts if len(self.all_token_counts) > 0 else [0])
def interrupt(self):
self.interrupted = True
def recover(self):
self.interrupted = False
def set_token_upper_limit(self, new_upper_limit):
self.token_upper_limit = new_upper_limit
self.auto_save()
def set_temperature(self, new_temperature):
self.temperature = new_temperature
self.auto_save()
def set_top_p(self, new_top_p):
self.top_p = new_top_p
self.auto_save()
def set_n_choices(self, new_n_choices):
self.n_choices = new_n_choices
self.auto_save()
def set_stop_sequence(self, new_stop_sequence: str):
new_stop_sequence = new_stop_sequence.split(",")
self.stop_sequence = new_stop_sequence
self.auto_save()
def set_max_tokens(self, new_max_tokens):
self.max_generation_token = new_max_tokens
self.auto_save()
def set_presence_penalty(self, new_presence_penalty):
self.presence_penalty = new_presence_penalty
self.auto_save()
def set_frequency_penalty(self, new_frequency_penalty):
self.frequency_penalty = new_frequency_penalty
self.auto_save()
def set_logit_bias(self, logit_bias):
self.logit_bias = logit_bias
self.auto_save()
def encoded_logit_bias(self):
if self.logit_bias is None:
return {}
logit_bias = self.logit_bias.split()
bias_map = {}
encoding = tiktoken.get_encoding("cl100k_base")
for line in logit_bias:
word, bias_amount = line.split(":")
if word:
for token in encoding.encode(word):
bias_map[token] = float(bias_amount)
return bias_map
def set_user_identifier(self, new_user_identifier):
self.user_identifier = new_user_identifier
self.auto_save()
def set_system_prompt(self, new_system_prompt):
self.system_prompt = new_system_prompt
self.auto_save()
def set_key(self, new_access_key):
if "*" not in new_access_key:
self.api_key = new_access_key.strip()
msg = i18n("API密钥更改为了") + hide_middle_chars(self.api_key)
logging.info(msg)
return self.api_key, msg
else:
return gr.update(), gr.update()
def set_single_turn(self, new_single_turn):
self.single_turn = new_single_turn
self.auto_save()
def reset(self, remain_system_prompt=False):
self.history = []
self.all_token_counts = []
self.interrupted = False
self.history_file_path = new_auto_history_filename(self.user_name)
history_name = self.history_file_path[:-5]
choices = get_history_names(self.user_name)
if history_name not in choices:
choices.insert(0, history_name)
system_prompt = self.system_prompt if remain_system_prompt else ""
self.single_turn = self.default_single_turn
self.temperature = self.default_temperature
self.top_p = self.default_top_p
self.n_choices = self.default_n_choices
self.stop_sequence = self.default_stop_sequence
self.max_generation_token = self.default_max_generation_token
self.presence_penalty = self.default_presence_penalty
self.frequency_penalty = self.default_frequency_penalty
self.logit_bias = self.default_logit_bias
self.user_identifier = self.default_user_identifier
return (
[],
self.token_message([0]),
gr.Radio(choices=choices, value=history_name),
system_prompt,
self.single_turn,
self.temperature,
self.top_p,
self.n_choices,
self.stop_sequence,
self.token_upper_limit,
self.max_generation_token,
self.presence_penalty,
self.frequency_penalty,
self.logit_bias,
self.user_identifier,
)
def delete_first_conversation(self):
if self.history:
del self.history[:2]
del self.all_token_counts[0]
return self.token_message()
def delete_last_conversation(self, chatbot):
if len(chatbot) > 0 and STANDARD_ERROR_MSG in chatbot[-1][1]:
msg = "由于包含报错信息,只删除chatbot记录"
chatbot = chatbot[:-1]
return chatbot, self.history
if len(self.history) > 0:
self.history = self.history[:-2]
if len(chatbot) > 0:
msg = "删除了一组chatbot对话"
chatbot = chatbot[:-1]
if len(self.all_token_counts) > 0:
msg = "删除了一组对话的token计数记录"
self.all_token_counts.pop()
msg = "删除了一组对话"
self.chatbot = chatbot
self.auto_save(chatbot)
return chatbot, msg
def token_message(self, token_lst=None):
if token_lst is None:
token_lst = self.all_token_counts
token_sum = 0
for i in range(len(token_lst)):
token_sum += sum(token_lst[: i + 1])
return (
i18n("Token 计数: ")
+ f"{sum(token_lst)}"
+ i18n(",本次对话累计消耗了 ")
+ f"{token_sum} tokens"
)
def rename_chat_history(self, filename, chatbot):
if filename == "":
return gr.update()
if not filename.endswith(".json"):
filename += ".json"
self.delete_chat_history(self.history_file_path)
# 命名重复检测
repeat_file_index = 2
full_path = os.path.join(HISTORY_DIR, self.user_name, filename)
while os.path.exists(full_path):
full_path = os.path.join(
HISTORY_DIR, self.user_name, f"{repeat_file_index}_{filename}"
)
repeat_file_index += 1
filename = os.path.basename(full_path)
self.history_file_path = filename
save_file(filename, self, chatbot)
return init_history_list(self.user_name)
def auto_name_chat_history(
self, name_chat_method, user_question, chatbot, single_turn_checkbox
):
if len(self.history) == 2 and not single_turn_checkbox:
user_question = self.history[0]["content"]
if type(user_question) == list:
user_question = user_question[0]["text"]
filename = replace_special_symbols(user_question)[:16] + ".json"
return self.rename_chat_history(filename, chatbot)
else:
return gr.update()
def auto_save(self, chatbot=None):
if chatbot is not None:
save_file(self.history_file_path, self, chatbot)
def export_markdown(self, filename, chatbot):
if filename == "":
return
if not filename.endswith(".md"):
filename += ".md"
save_file(filename, self, chatbot)
def load_chat_history(self, new_history_file_path=None):
logging.debug(f"{self.user_name} 加载对话历史中……")
if new_history_file_path is not None:
if type(new_history_file_path) != str:
# copy file from new_history_file_path.name to os.path.join(HISTORY_DIR, self.user_name)
new_history_file_path = new_history_file_path.name
shutil.copyfile(
new_history_file_path,
os.path.join(
HISTORY_DIR,
self.user_name,
os.path.basename(new_history_file_path),
),
)
self.history_file_path = os.path.basename(new_history_file_path)
else:
self.history_file_path = new_history_file_path
try:
if self.history_file_path == os.path.basename(self.history_file_path):
history_file_path = os.path.join(
HISTORY_DIR, self.user_name, self.history_file_path
)
else:
history_file_path = self.history_file_path
if not self.history_file_path.endswith(".json"):
history_file_path += ".json"
with open(history_file_path, "r", encoding="utf-8") as f:
saved_json = json.load(f)
try:
if type(saved_json["history"][0]) == str:
logging.info("历史记录格式为旧版,正在转换……")
new_history = []
for index, item in enumerate(saved_json["history"]):
if index % 2 == 0:
new_history.append(construct_user(item))
else:
new_history.append(construct_assistant(item))
saved_json["history"] = new_history
logging.info(new_history)
except:
pass
if len(saved_json["chatbot"]) < len(saved_json["history"]) // 2:
logging.info("Trimming corrupted history...")
saved_json["history"] = saved_json["history"][
-len(saved_json["chatbot"]) :
]
logging.info(f"Trimmed history: {saved_json['history']}")
logging.debug(f"{self.user_name} 加载对话历史完毕")
self.history = saved_json["history"]
self.single_turn = saved_json.get("single_turn", self.single_turn)
self.temperature = saved_json.get("temperature", self.temperature)
self.top_p = saved_json.get("top_p", self.top_p)
self.n_choices = saved_json.get("n_choices", self.n_choices)
self.stop_sequence = list(saved_json.get("stop_sequence", self.stop_sequence))
self.token_upper_limit = saved_json.get(
"token_upper_limit", self.token_upper_limit
)
self.max_generation_token = saved_json.get(
"max_generation_token", self.max_generation_token
)
self.presence_penalty = saved_json.get(
"presence_penalty", self.presence_penalty
)
self.frequency_penalty = saved_json.get(
"frequency_penalty", self.frequency_penalty
)
self.logit_bias = saved_json.get("logit_bias", self.logit_bias)
self.user_identifier = saved_json.get("user_identifier", self.user_name)
self.metadata = saved_json.get("metadata", self.metadata)
self.chatbot = saved_json["chatbot"]
return (
os.path.basename(self.history_file_path)[:-5],
saved_json["system"],
saved_json["chatbot"],
self.single_turn,
self.temperature,
self.top_p,
self.n_choices,
",".join(self.stop_sequence),
self.token_upper_limit,
self.max_generation_token,
self.presence_penalty,
self.frequency_penalty,
self.logit_bias,
self.user_identifier,
)
except:
# 没有对话历史或者对话历史解析失败
logging.info(f"没有找到对话历史记录 {self.history_file_path}")
self.reset()
return (
os.path.basename(self.history_file_path),
"",
[],
self.single_turn,
self.temperature,
self.top_p,
self.n_choices,
",".join(self.stop_sequence),
self.token_upper_limit,
self.max_generation_token,
self.presence_penalty,
self.frequency_penalty,
self.logit_bias,
self.user_identifier,
)
def delete_chat_history(self, filename):
if filename == "CANCELED":
return gr.update(), gr.update(), gr.update()
if filename == "":
return i18n("你没有选择任何对话历史"), gr.update(), gr.update()
if not filename.endswith(".json"):
filename += ".json"
if filename == os.path.basename(filename):
history_file_path = os.path.join(HISTORY_DIR, self.user_name, filename)
else:
history_file_path = filename
md_history_file_path = history_file_path[:-5] + ".md"
try:
os.remove(history_file_path)
os.remove(md_history_file_path)
return i18n("删除对话历史成功"), get_history_list(self.user_name), []
except:
logging.info(f"删除对话历史失败 {history_file_path}")
return (
i18n("对话历史") + filename + i18n("已经被删除啦"),
get_history_list(self.user_name),
[],
)
def auto_load(self):
self.history_file_path = new_auto_history_filename(self.user_name)
return self.load_chat_history()
def like(self):
"""like the last response, implement if needed"""
return gr.update()
def dislike(self):
"""dislike the last response, implement if needed"""
return gr.update()
def deinitialize(self):
"""deinitialize the model, implement if needed"""
pass
def clear_cuda_cache(self):
import gc
import torch
gc.collect()
torch.cuda.empty_cache()
def get_base64_image(self, image_path):
if image_path.endswith(DIRECTLY_SUPPORTED_IMAGE_FORMATS):
with open(image_path, "rb") as f:
return base64.b64encode(f.read()).decode("utf-8")
else:
# convert to jpeg
image = PIL.Image.open(image_path)
image = image.convert("RGB")
buffer = BytesIO()
image.save(buffer, format="JPEG")
return base64.b64encode(buffer.getvalue()).decode("utf-8")
def get_image_type(self, image_path):
if image_path.lower().endswith(DIRECTLY_SUPPORTED_IMAGE_FORMATS):
return os.path.splitext(image_path)[1][1:].lower()
else:
return "jpeg"
class Base_Chat_Langchain_Client(BaseLLMModel):
def __init__(self, model_name, user_name=""):
super().__init__(model_name, user=user_name)
self.need_api_key = False
self.model = self.setup_model()
def setup_model(self):
# inplement this to setup the model then return it
pass
def _get_langchain_style_history(self):
history = [SystemMessage(content=self.system_prompt)]
for i in self.history:
if i["role"] == "user":
history.append(HumanMessage(content=i["content"]))
elif i["role"] == "assistant":
history.append(AIMessage(content=i["content"]))
return history
def get_answer_at_once(self):
assert isinstance(
self.model, BaseChatModel
), "model is not instance of LangChain BaseChatModel"
history = self._get_langchain_style_history()
response = self.model.generate(history)
return response.content, sum(response.content)
def get_answer_stream_iter(self):
it = CallbackToIterator()
assert isinstance(
self.model, BaseChatModel
), "model is not instance of LangChain BaseChatModel"
history = self._get_langchain_style_history()
def thread_func():
self.model(
messages=history, callbacks=[ChuanhuCallbackHandler(it.callback)]
)
it.finish()
t = Thread(target=thread_func)
t.start()
partial_text = ""
for value in it:
partial_text += value
yield partial_text
|