支持百度云千帆和文心一言
This commit is contained in:
binary-husky
parent
30f28b37c3
commit
8fce49fa02
67
config.py
67
config.py
@ -11,7 +11,7 @@
|
||||
API_KEY = "此处填API密钥" # 可同时填写多个API-KEY，用英文逗号分割，例如API_KEY = "sk-openaikey1,sk-openaikey2,fkxxxx-api2dkey3,azure-apikey4"
|
||||
|
||||
|
||||
# [step 2]>> 改为True应用代理，如果直接在海外服务器部署，此处不修改
|
||||
# [step 2]>> 改为True应用代理，如果直接在海外服务器部署，此处不修改；如果使用本地或无地域限制的大模型时，此处也不需要修改
|
||||
USE_PROXY = False
|
||||
if USE_PROXY:
|
||||
"""
|
||||
@ -71,10 +71,16 @@ MAX_RETRY = 2
|
||||
# 模型选择是 (注意: LLM_MODEL是默认选中的模型, 它*必须*被包含在AVAIL_LLM_MODELS列表中 )
|
||||
LLM_MODEL = "gpt-3.5-turbo" # 可选 ↓↓↓
|
||||
AVAIL_LLM_MODELS = ["gpt-3.5-turbo-16k", "gpt-3.5-turbo", "azure-gpt-3.5", "api2d-gpt-3.5-turbo", "gpt-4", "api2d-gpt-4", "chatglm", "moss", "newbing", "stack-claude"]
|
||||
# P.S. 其他可用的模型还包括 ["llama2", "qwen", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-16k-0613", "spark", "chatglm_onnx", "claude-1-100k", "claude-2", "internlm", "jittorllms_rwkv", "jittorllms_pangualpha", "jittorllms_llama"]
|
||||
# P.S. 其他可用的模型还包括 ["qianfan", "llama2", "qwen", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-16k-0613", "spark", "chatglm_onnx", "claude-1-100k", "claude-2", "internlm", "jittorllms_rwkv", "jittorllms_pangualpha", "jittorllms_llama"]
|
||||
|
||||
|
||||
# ChatGLM(2) Finetune Model Path （如果使用ChatGLM2微调模型，需要把"chatglmft"加入AVAIL_LLM_MODELS中）
|
||||
# 百度千帆（LLM_MODEL="qianfan"）
|
||||
BAIDU_CLOUD_API_KEY = ''
|
||||
BAIDU_CLOUD_SECRET_KEY = ''
|
||||
BAIDU_CLOUD_QIANFAN_MODEL = 'ERNIE-Bot' # 可选 "ERNIE-Bot"(文心一言), "ERNIE-Bot-turbo", "BLOOMZ-7B", "Llama-2-70B-Chat", "Llama-2-13B-Chat", "Llama-2-7B-Chat"
|
||||
|
||||
|
||||
# 如果使用ChatGLM2微调模型，请把 LLM_MODEL="chatglmft"，并在此处指定模型路径
|
||||
ChatGLM_PTUNING_CHECKPOINT = "" # 例如"/home/hmp/ChatGLM2-6B/ptuning/output/6b-pt-128-1e-2/checkpoint-100"
|
||||
|
||||
|
||||
@ -152,5 +158,56 @@ ANTHROPIC_API_KEY = ""
|
||||
CUSTOM_API_KEY_PATTERN = ""
|
||||
|
||||
|
||||
# HUGGINGFACE的TOKEN 下载LLAMA时起作用 https://huggingface.co/docs/hub/security-tokens
|
||||
HUGGINGFACE_ACCESS_TOKEN = "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"
|
||||
# HUGGINGFACE的TOKEN，下载LLAMA时起作用 https://huggingface.co/docs/hub/security-tokens
|
||||
HUGGINGFACE_ACCESS_TOKEN = "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"
|
||||
|
||||
|
||||
|
||||
"""
|
||||
在线大模型配置关联关系示意图
|
||||
│
|
||||
├── "gpt-3.5-turbo" 等openai模型
|
||||
│ ├── API_KEY
|
||||
│ ├── CUSTOM_API_KEY_PATTERN（不常用）
|
||||
│ ├── API_ORG（不常用）
|
||||
│ └── API_URL_REDIRECT（不常用）
|
||||
│
|
||||
├── "azure-gpt-3.5" 等azure模型
|
||||
│ ├── API_KEY
|
||||
│ ├── AZURE_ENDPOINT
|
||||
│ ├── AZURE_API_KEY
|
||||
│ ├── AZURE_ENGINE
|
||||
│ └── API_URL_REDIRECT
|
||||
│
|
||||
├── "spark" 星火认知大模型
|
||||
│ ├── XFYUN_APPID
|
||||
│ ├── XFYUN_API_SECRET
|
||||
│ └── XFYUN_API_KEY
|
||||
│
|
||||
├── "claude-1-100k" 等claude模型
|
||||
│ └── ANTHROPIC_API_KEY
|
||||
│
|
||||
├── "stack-claude"
|
||||
│ ├── SLACK_CLAUDE_BOT_ID
|
||||
│ └── SLACK_CLAUDE_USER_TOKEN
|
||||
│
|
||||
├── "qianfan" 百度千帆大模型库
|
||||
│ ├── BAIDU_CLOUD_QIANFAN_MODEL
|
||||
│ ├── BAIDU_CLOUD_API_KEY
|
||||
│ └── BAIDU_CLOUD_SECRET_KEY
|
||||
│
|
||||
├── "newbing" Newbing接口不再稳定，不推荐使用
|
||||
├── NEWBING_STYLE
|
||||
└── NEWBING_COOKIES
|
||||
|
||||
|
||||
|
||||
插件在线服务配置依赖关系示意图
|
||||
│
|
||||
├── 语音功能
|
||||
├── ENABLE_AUDIO
|
||||
├── ALIYUN_TOKEN
|
||||
├── ALIYUN_APPKEY
|
||||
├── ALIYUN_ACCESSKEY
|
||||
└── ALIYUN_SECRET
|
||||
"""
|
||||
@ -19,6 +19,12 @@ from .bridge_chatgpt import predict as chatgpt_ui
|
||||
from .bridge_chatglm import predict_no_ui_long_connection as chatglm_noui
|
||||
from .bridge_chatglm import predict as chatglm_ui
|
||||
|
||||
from .bridge_chatglm import predict_no_ui_long_connection as chatglm_noui
|
||||
from .bridge_chatglm import predict as chatglm_ui
|
||||
|
||||
from .bridge_qianfan import predict_no_ui_long_connection as qianfan_noui
|
||||
from .bridge_qianfan import predict as qianfan_ui
|
||||
|
||||
colors = ['#FF00FF', '#00FFFF', '#FF0000', '#990099', '#009999', '#990044']
|
||||
|
||||
class LazyloadTiktoken(object):
|
||||
@ -165,7 +171,14 @@ model_info = {
|
||||
"tokenizer": tokenizer_gpt35,
|
||||
"token_cnt": get_token_num_gpt35,
|
||||
},
|
||||
|
||||
"qianfan": {
|
||||
"fn_with_ui": qianfan_ui,
|
||||
"fn_without_ui": qianfan_noui,
|
||||
"endpoint": None,
|
||||
"max_token": 4096,
|
||||
"tokenizer": tokenizer_gpt35,
|
||||
"token_cnt": get_token_num_gpt35,
|
||||
},
|
||||
}
|
||||
|
||||
# -=-=-=-=-=-=- 以下部分是新加入的模型，可能附带额外依赖 -=-=-=-=-=-=-
|
||||
@ -385,7 +398,7 @@ if "spark" in AVAIL_LLM_MODELS: # 讯飞星火认知大模型
|
||||
})
|
||||
except:
|
||||
print(trimmed_format_exc())
|
||||
if "llama2" in AVAIL_LLM_MODELS: # 讯飞星火认知大模型
|
||||
if "llama2" in AVAIL_LLM_MODELS: # llama2
|
||||
try:
|
||||
from .bridge_llama2 import predict_no_ui_long_connection as llama2_noui
|
||||
from .bridge_llama2 import predict as llama2_ui
|
||||
|
||||
150
request_llm/bridge_qianfan.py
Normal file
150
request_llm/bridge_qianfan.py
Normal file
@ -0,0 +1,150 @@
|
||||
|
||||
import time, requests, json
|
||||
from toolbox import update_ui, get_conf
|
||||
from multiprocessing import Process, Pipe
|
||||
from functools import wraps
|
||||
from datetime import datetime, timedelta
|
||||
|
||||
model_name = '千帆大模型平台'
|
||||
timeout_bot_msg = '[Local Message] Request timeout. Network error.'
|
||||
|
||||
def cache_decorator(timeout):
|
||||
cache = {}
|
||||
def decorator(func):
|
||||
@wraps(func)
|
||||
def wrapper(*args, **kwargs):
|
||||
key = (func.__name__, args, frozenset(kwargs.items()))
|
||||
# Check if result is already cached and not expired
|
||||
if key in cache:
|
||||
result, timestamp = cache[key]
|
||||
if datetime.now() - timestamp < timedelta(seconds=timeout):
|
||||
return result
|
||||
|
||||
# Call the function and cache the result
|
||||
result = func(*args, **kwargs)
|
||||
cache[key] = (result, datetime.now())
|
||||
return result
|
||||
return wrapper
|
||||
return decorator
|
||||
|
||||
@cache_decorator(timeout=3600)
|
||||
def get_access_token():
|
||||
"""
|
||||
使用 AK，SK 生成鉴权签名（Access Token）
|
||||
:return: access_token，或是None(如果错误)
|
||||
"""
|
||||
# if (access_token_cache is None) or (time.time() - last_access_token_obtain_time > 3600):
|
||||
BAIDU_CLOUD_API_KEY, BAIDU_CLOUD_SECRET_KEY = get_conf('BAIDU_CLOUD_API_KEY', 'BAIDU_CLOUD_SECRET_KEY')
|
||||
|
||||
if len(BAIDU_CLOUD_SECRET_KEY) == 0: raise RuntimeError("没有配置BAIDU_CLOUD_SECRET_KEY")
|
||||
if len(BAIDU_CLOUD_API_KEY) == 0: raise RuntimeError("没有配置BAIDU_CLOUD_API_KEY")
|
||||
|
||||
url = "https://aip.baidubce.com/oauth/2.0/token"
|
||||
params = {"grant_type": "client_credentials", "client_id": BAIDU_CLOUD_API_KEY, "client_secret": BAIDU_CLOUD_SECRET_KEY}
|
||||
access_token_cache = str(requests.post(url, params=params).json().get("access_token"))
|
||||
return access_token_cache
|
||||
# else:
|
||||
# return access_token_cache
|
||||
|
||||
|
||||
def generate_message_payload(inputs, llm_kwargs, history, system_prompt):
|
||||
conversation_cnt = len(history) // 2
|
||||
messages = [{"role": "user", "content": system_prompt}]
|
||||
messages.append({"role": "assistant", "content": 'Certainly!'})
|
||||
if conversation_cnt:
|
||||
for index in range(0, 2*conversation_cnt, 2):
|
||||
what_i_have_asked = {}
|
||||
what_i_have_asked["role"] = "user"
|
||||
what_i_have_asked["content"] = history[index]
|
||||
what_gpt_answer = {}
|
||||
what_gpt_answer["role"] = "assistant"
|
||||
what_gpt_answer["content"] = history[index+1]
|
||||
if what_i_have_asked["content"] != "":
|
||||
if what_gpt_answer["content"] == "": continue
|
||||
if what_gpt_answer["content"] == timeout_bot_msg: continue
|
||||
messages.append(what_i_have_asked)
|
||||
messages.append(what_gpt_answer)
|
||||
else:
|
||||
messages[-1]['content'] = what_gpt_answer['content']
|
||||
what_i_ask_now = {}
|
||||
what_i_ask_now["role"] = "user"
|
||||
what_i_ask_now["content"] = inputs
|
||||
messages.append(what_i_ask_now)
|
||||
return messages
|
||||
|
||||
|
||||
def generate_from_baidu_qianfan(inputs, llm_kwargs, history, system_prompt):
|
||||
BAIDU_CLOUD_QIANFAN_MODEL, = get_conf('BAIDU_CLOUD_QIANFAN_MODEL')
|
||||
|
||||
url_lib = {
|
||||
"ERNIE-Bot": "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions" ,
|
||||
"ERNIE-Bot-turbo": "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/eb-instant" ,
|
||||
"BLOOMZ-7B": "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/bloomz_7b1",
|
||||
|
||||
"Llama-2-70B-Chat": "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/llama_2_70b",
|
||||
"Llama-2-13B-Chat": "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/llama_2_13b",
|
||||
"Llama-2-7B-Chat": "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/llama_2_7b",
|
||||
}
|
||||
|
||||
url = url_lib[BAIDU_CLOUD_QIANFAN_MODEL]
|
||||
|
||||
url += "?access_token=" + get_access_token()
|
||||
|
||||
|
||||
payload = json.dumps({
|
||||
"messages": generate_message_payload(inputs, llm_kwargs, history, system_prompt),
|
||||
"stream": True
|
||||
})
|
||||
headers = {
|
||||
'Content-Type': 'application/json'
|
||||
}
|
||||
response = requests.request("POST", url, headers=headers, data=payload, stream=True)
|
||||
buffer = ""
|
||||
for line in response.iter_lines():
|
||||
try:
|
||||
dec = line.decode().lstrip('data:')
|
||||
dec = json.loads(dec)
|
||||
incoming = dec['result']
|
||||
buffer += incoming
|
||||
yield buffer
|
||||
except:
|
||||
if 'error_code' in dec: raise RuntimeError(dec['error_msg'])
|
||||
|
||||
|
||||
def predict_no_ui_long_connection(inputs, llm_kwargs, history=[], sys_prompt="", observe_window=[], console_slience=False):
|
||||
"""
|
||||
⭐多线程方法
|
||||
函数的说明请见 request_llm/bridge_all.py
|
||||
"""
|
||||
watch_dog_patience = 5
|
||||
response = ""
|
||||
|
||||
for response in generate_from_baidu_qianfan(inputs, llm_kwargs, history, sys_prompt):
|
||||
if len(observe_window) >= 1:
|
||||
observe_window[0] = response
|
||||
if len(observe_window) >= 2:
|
||||
if (time.time()-observe_window[1]) > watch_dog_patience: raise RuntimeError("程序终止。")
|
||||
return response
|
||||
|
||||
def predict(inputs, llm_kwargs, plugin_kwargs, chatbot, history=[], system_prompt='', stream = True, additional_fn=None):
|
||||
"""
|
||||
⭐单线程方法
|
||||
函数的说明请见 request_llm/bridge_all.py
|
||||
"""
|
||||
chatbot.append((inputs, ""))
|
||||
|
||||
if additional_fn is not None:
|
||||
from core_functional import handle_core_functionality
|
||||
inputs, history = handle_core_functionality(additional_fn, inputs, history, chatbot)
|
||||
|
||||
# 开始接收回复
|
||||
for response in generate_from_baidu_qianfan(inputs, llm_kwargs, history, system_prompt):
|
||||
chatbot[-1] = (inputs, response)
|
||||
yield from update_ui(chatbot=chatbot, history=history)
|
||||
|
||||
# 总结输出
|
||||
response = f"[Local Message]: {model_name}响应异常 ..."
|
||||
if response == f"[Local Message]: 等待{model_name}响应中 ...":
|
||||
response = f"[Local Message]: {model_name}响应异常 ..."
|
||||
history.extend([inputs, response])
|
||||
yield from update_ui(chatbot=chatbot, history=history)
|
||||
@ -128,7 +128,7 @@ def get_local_llm_predict_fns(LLMSingletonClass, model_name):
|
||||
|
||||
# chatglm 没有 sys_prompt 接口，因此把prompt加入 history
|
||||
history_feedin = []
|
||||
history_feedin.append(["What can I do?", sys_prompt])
|
||||
history_feedin.append([sys_prompt, "Certainly!"])
|
||||
for i in range(len(history)//2):
|
||||
history_feedin.append([history[2*i], history[2*i+1]] )
|
||||
|
||||
@ -161,7 +161,7 @@ def get_local_llm_predict_fns(LLMSingletonClass, model_name):
|
||||
|
||||
# 处理历史信息
|
||||
history_feedin = []
|
||||
history_feedin.append(["What can I do?", system_prompt] )
|
||||
history_feedin.append([system_prompt, "Certainly!"])
|
||||
for i in range(len(history)//2):
|
||||
history_feedin.append([history[2*i], history[2*i+1]] )
|
||||
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user