Add support to aliyun qwen online models.
Rename model tag "qwen" to "qwen-local" Add model tag "qwen-turbo", "qwen-plus", "qwen-max" Add corresponding model interfaces in request_llms/bridge_all.py Add configuration variable “DASHSCOPE_API_KEY" Rename request_llms/bridge_qwen.py to bridge_qwen_local.py to distinguish it from the online model interface
This commit is contained in:
leike0813
parent
9479dd984c
commit
ac3d4cf073
16
config.py
16
config.py
@ -92,8 +92,9 @@ AVAIL_LLM_MODELS = ["gpt-3.5-turbo-1106","gpt-4-1106-preview","gpt-4-vision-prev
|
||||
"api2d-gpt-3.5-turbo", 'api2d-gpt-3.5-turbo-16k',
|
||||
"gpt-4", "gpt-4-32k", "azure-gpt-4", "api2d-gpt-4",
|
||||
"chatglm3", "moss", "claude-2"]
|
||||
# P.S. 其他可用的模型还包括 ["zhipuai", "qianfan", "deepseekcoder", "llama2", "qwen", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-16k-0613", "gpt-3.5-random"
|
||||
# "spark", "sparkv2", "sparkv3", "chatglm_onnx", "claude-1-100k", "claude-2", "internlm", "jittorllms_pangualpha", "jittorllms_llama"]
|
||||
# P.S. 其他可用的模型还包括 ["zhipuai", "qianfan", "deepseekcoder", "llama2", "qwen-local", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-16k-0613", "gpt-3.5-random"
|
||||
# "spark", "sparkv2", "sparkv3", "chatglm_onnx", "claude-1-100k", "claude-2", "internlm", "jittorllms_pangualpha", "jittorllms_llama"
|
||||
# “qwen-turbo", "qwen-plus", "qwen-max"]
|
||||
|
||||
|
||||
# 定义界面上“询问多个GPT模型”插件应该使用哪些模型，请从AVAIL_LLM_MODELS中选择，并在不同模型之间用`&`间隔，例如"gpt-3.5-turbo&chatglm3&azure-gpt-4"
|
||||
@ -103,7 +104,11 @@ MULTI_QUERY_LLM_MODELS = "gpt-3.5-turbo&chatglm3"
|
||||
# 选择本地模型变体（只有当AVAIL_LLM_MODELS包含了对应本地模型时，才会起作用）
|
||||
# 如果你选择Qwen系列的模型，那么请在下面的QWEN_MODEL_SELECTION中指定具体的模型
|
||||
# 也可以是具体的模型路径
|
||||
QWEN_MODEL_SELECTION = "Qwen/Qwen-1_8B-Chat-Int8"
|
||||
QWEN_LOCAL_MODEL_SELECTION = "Qwen/Qwen-1_8B-Chat-Int8"
|
||||
|
||||
|
||||
# 接入通义千问在线大模型 https://dashscope.console.aliyun.com/
|
||||
DASHSCOPE_API_KEY = "此处填阿里灵积云API秘钥" # 阿里灵积云API_KEY
|
||||
|
||||
|
||||
# 百度千帆（LLM_MODEL="qianfan"）
|
||||
@ -284,6 +289,9 @@ NUM_CUSTOM_BASIC_BTN = 4
|
||||
│ ├── ZHIPUAI_API_KEY
|
||||
│ └── ZHIPUAI_MODEL
|
||||
│
|
||||
├── "qwen-turbo" 等通义千问大模型
|
||||
│ └── DASHSCOPE_API_KEY
|
||||
│
|
||||
└── "newbing" Newbing接口不再稳定，不推荐使用
|
||||
├── NEWBING_STYLE
|
||||
└── NEWBING_COOKIES
|
||||
@ -300,7 +308,7 @@ NUM_CUSTOM_BASIC_BTN = 4
|
||||
├── "jittorllms_pangualpha"
|
||||
├── "jittorllms_llama"
|
||||
├── "deepseekcoder"
|
||||
├── "qwen"
|
||||
├── "qwen-local"
|
||||
├── RWKV的支持见Wiki
|
||||
└── "llama2"
|
||||
|
||||
|
||||
@ -2932,7 +2932,7 @@
|
||||
"3. 输入修改需求": "3. Enter modification requirements",
|
||||
"刷新界面 由于请求gpt需要一段时间": "Refreshing the interface takes some time due to the request for gpt",
|
||||
"随机小游戏": "Random mini game",
|
||||
"那么请在下面的QWEN_MODEL_SELECTION中指定具体的模型": "So please specify the specific model in QWEN_MODEL_SELECTION below",
|
||||
"那么请在下面的QWEN_LOCAL_MODEL_SELECTION中指定具体的模型": "So please specify the specific model in QWEN_LOCAL_MODEL_SELECTION below",
|
||||
"表值": "Table value",
|
||||
"我画你猜": "I draw, you guess",
|
||||
"狗": "Dog",
|
||||
|
||||
@ -431,16 +431,48 @@ if "chatglm_onnx" in AVAIL_LLM_MODELS:
|
||||
})
|
||||
except:
|
||||
print(trimmed_format_exc())
|
||||
if "qwen" in AVAIL_LLM_MODELS:
|
||||
if "qwen-local" in AVAIL_LLM_MODELS:
|
||||
try:
|
||||
from .bridge_qwen_local import predict_no_ui_long_connection as qwen_local_noui
|
||||
from .bridge_qwen_local import predict as qwen_local_ui
|
||||
model_info.update({
|
||||
"qwen-local": {
|
||||
"fn_with_ui": qwen_local_ui,
|
||||
"fn_without_ui": qwen_local_noui,
|
||||
"endpoint": None,
|
||||
"max_token": 4096,
|
||||
"tokenizer": tokenizer_gpt35,
|
||||
"token_cnt": get_token_num_gpt35,
|
||||
}
|
||||
})
|
||||
except:
|
||||
print(trimmed_format_exc())
|
||||
if "qwen-turbo" in AVAIL_LLM_MODELS or "qwen-plus" in AVAIL_LLM_MODELS or "qwen-max" in AVAIL_LLM_MODELS: # zhipuai
|
||||
try:
|
||||
from .bridge_qwen import predict_no_ui_long_connection as qwen_noui
|
||||
from .bridge_qwen import predict as qwen_ui
|
||||
model_info.update({
|
||||
"qwen": {
|
||||
"qwen-turbo": {
|
||||
"fn_with_ui": qwen_ui,
|
||||
"fn_without_ui": qwen_noui,
|
||||
"endpoint": None,
|
||||
"max_token": 4096,
|
||||
"max_token": 6144,
|
||||
"tokenizer": tokenizer_gpt35,
|
||||
"token_cnt": get_token_num_gpt35,
|
||||
},
|
||||
"qwen-plus": {
|
||||
"fn_with_ui": qwen_ui,
|
||||
"fn_without_ui": qwen_noui,
|
||||
"endpoint": None,
|
||||
"max_token": 30720,
|
||||
"tokenizer": tokenizer_gpt35,
|
||||
"token_cnt": get_token_num_gpt35,
|
||||
},
|
||||
"qwen-max": {
|
||||
"fn_with_ui": qwen_ui,
|
||||
"fn_without_ui": qwen_noui,
|
||||
"endpoint": None,
|
||||
"max_token": 28672,
|
||||
"tokenizer": tokenizer_gpt35,
|
||||
"token_cnt": get_token_num_gpt35,
|
||||
}
|
||||
|
||||
@ -1,59 +1,66 @@
|
||||
model_name = "Qwen"
|
||||
cmd_to_install = "`pip install -r request_llms/requirements_qwen.txt`"
|
||||
import time
|
||||
import os
|
||||
from toolbox import update_ui, get_conf, update_ui_lastest_msg
|
||||
from toolbox import check_packages, report_exception
|
||||
|
||||
from toolbox import ProxyNetworkActivate, get_conf
|
||||
from .local_llm_class import LocalLLMHandle, get_local_llm_predict_fns
|
||||
model_name = 'Qwen'
|
||||
|
||||
def validate_key():
|
||||
DASHSCOPE_API_KEY = get_conf("DASHSCOPE_API_KEY")
|
||||
if DASHSCOPE_API_KEY == '': return False
|
||||
return True
|
||||
|
||||
if not validate_key():
|
||||
raise RuntimeError('请配置DASHSCOPE_API_KEY')
|
||||
os.environ['DASHSCOPE_API_KEY'] = get_conf("DASHSCOPE_API_KEY")
|
||||
|
||||
|
||||
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_llms/bridge_all.py
|
||||
"""
|
||||
watch_dog_patience = 5
|
||||
response = ""
|
||||
|
||||
# ------------------------------------------------------------------------------------------------------------------------
|
||||
# 🔌💻 Local Model
|
||||
# ------------------------------------------------------------------------------------------------------------------------
|
||||
class GetQwenLMHandle(LocalLLMHandle):
|
||||
from .com_qwenapi import QwenRequestInstance
|
||||
sri = QwenRequestInstance()
|
||||
for response in sri.generate(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 load_model_info(self):
|
||||
# 🏃♂️🏃♂️🏃♂️ 子进程执行
|
||||
self.model_name = model_name
|
||||
self.cmd_to_install = cmd_to_install
|
||||
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_llms/bridge_all.py
|
||||
"""
|
||||
chatbot.append((inputs, ""))
|
||||
yield from update_ui(chatbot=chatbot, history=history)
|
||||
|
||||
def load_model_and_tokenizer(self):
|
||||
# 🏃♂️🏃♂️🏃♂️ 子进程执行
|
||||
# from modelscope import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
|
||||
from transformers import AutoModelForCausalLM, AutoTokenizer
|
||||
from transformers.generation import GenerationConfig
|
||||
with ProxyNetworkActivate('Download_LLM'):
|
||||
model_id = get_conf('QWEN_MODEL_SELECTION')
|
||||
self._tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True, resume_download=True)
|
||||
# use fp16
|
||||
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", trust_remote_code=True).eval()
|
||||
model.generation_config = GenerationConfig.from_pretrained(model_id, trust_remote_code=True) # 可指定不同的生成长度、top_p等相关超参
|
||||
self._model = model
|
||||
# 尝试导入依赖，如果缺少依赖，则给出安装建议
|
||||
try:
|
||||
check_packages(["dashscope"])
|
||||
except:
|
||||
yield from update_ui_lastest_msg(f"导入软件依赖失败。使用该模型需要额外依赖，安装方法```pip install --upgrade dashscope```。",
|
||||
chatbot=chatbot, history=history, delay=0)
|
||||
return
|
||||
|
||||
return self._model, self._tokenizer
|
||||
if additional_fn is not None:
|
||||
from core_functional import handle_core_functionality
|
||||
inputs, history = handle_core_functionality(additional_fn, inputs, history, chatbot)
|
||||
|
||||
def llm_stream_generator(self, **kwargs):
|
||||
# 🏃♂️🏃♂️🏃♂️ 子进程执行
|
||||
def adaptor(kwargs):
|
||||
query = kwargs['query']
|
||||
max_length = kwargs['max_length']
|
||||
top_p = kwargs['top_p']
|
||||
temperature = kwargs['temperature']
|
||||
history = kwargs['history']
|
||||
return query, max_length, top_p, temperature, history
|
||||
# 开始接收回复
|
||||
from .com_qwenapi import QwenRequestInstance
|
||||
sri = QwenRequestInstance()
|
||||
for response in sri.generate(inputs, llm_kwargs, history, system_prompt):
|
||||
chatbot[-1] = (inputs, response)
|
||||
yield from update_ui(chatbot=chatbot, history=history)
|
||||
|
||||
query, max_length, top_p, temperature, history = adaptor(kwargs)
|
||||
|
||||
for response in self._model.chat_stream(self._tokenizer, query, history=history):
|
||||
yield response
|
||||
|
||||
def try_to_import_special_deps(self, **kwargs):
|
||||
# import something that will raise error if the user does not install requirement_*.txt
|
||||
# 🏃♂️🏃♂️🏃♂️ 主进程执行
|
||||
import importlib
|
||||
importlib.import_module('modelscope')
|
||||
|
||||
|
||||
# ------------------------------------------------------------------------------------------------------------------------
|
||||
# 🔌💻 GPT-Academic Interface
|
||||
# ------------------------------------------------------------------------------------------------------------------------
|
||||
predict_no_ui_long_connection, predict = get_local_llm_predict_fns(GetQwenLMHandle, model_name)
|
||||
# 总结输出
|
||||
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)
|
||||
59
request_llms/bridge_qwen_local.py
Normal file
59
request_llms/bridge_qwen_local.py
Normal file
@ -0,0 +1,59 @@
|
||||
model_name = "Qwen_local"
|
||||
cmd_to_install = "`pip install -r request_llms/requirements_qwen_local.txt`"
|
||||
|
||||
from toolbox import ProxyNetworkActivate, get_conf
|
||||
from .local_llm_class import LocalLLMHandle, get_local_llm_predict_fns
|
||||
|
||||
|
||||
|
||||
# ------------------------------------------------------------------------------------------------------------------------
|
||||
# 🔌💻 Local Model
|
||||
# ------------------------------------------------------------------------------------------------------------------------
|
||||
class GetQwenLMHandle(LocalLLMHandle):
|
||||
|
||||
def load_model_info(self):
|
||||
# 🏃♂️🏃♂️🏃♂️ 子进程执行
|
||||
self.model_name = model_name
|
||||
self.cmd_to_install = cmd_to_install
|
||||
|
||||
def load_model_and_tokenizer(self):
|
||||
# 🏃♂️🏃♂️🏃♂️ 子进程执行
|
||||
# from modelscope import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
|
||||
from transformers import AutoModelForCausalLM, AutoTokenizer
|
||||
from transformers.generation import GenerationConfig
|
||||
with ProxyNetworkActivate('Download_LLM'):
|
||||
model_id = get_conf('QWEN_LOCAL_MODEL_SELECTION')
|
||||
self._tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True, resume_download=True)
|
||||
# use fp16
|
||||
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", trust_remote_code=True).eval()
|
||||
model.generation_config = GenerationConfig.from_pretrained(model_id, trust_remote_code=True) # 可指定不同的生成长度、top_p等相关超参
|
||||
self._model = model
|
||||
|
||||
return self._model, self._tokenizer
|
||||
|
||||
def llm_stream_generator(self, **kwargs):
|
||||
# 🏃♂️🏃♂️🏃♂️ 子进程执行
|
||||
def adaptor(kwargs):
|
||||
query = kwargs['query']
|
||||
max_length = kwargs['max_length']
|
||||
top_p = kwargs['top_p']
|
||||
temperature = kwargs['temperature']
|
||||
history = kwargs['history']
|
||||
return query, max_length, top_p, temperature, history
|
||||
|
||||
query, max_length, top_p, temperature, history = adaptor(kwargs)
|
||||
|
||||
for response in self._model.chat_stream(self._tokenizer, query, history=history):
|
||||
yield response
|
||||
|
||||
def try_to_import_special_deps(self, **kwargs):
|
||||
# import something that will raise error if the user does not install requirement_*.txt
|
||||
# 🏃♂️🏃♂️🏃♂️ 主进程执行
|
||||
import importlib
|
||||
importlib.import_module('modelscope')
|
||||
|
||||
|
||||
# ------------------------------------------------------------------------------------------------------------------------
|
||||
# 🔌💻 GPT-Academic Interface
|
||||
# ------------------------------------------------------------------------------------------------------------------------
|
||||
predict_no_ui_long_connection, predict = get_local_llm_predict_fns(GetQwenLMHandle, model_name)
|
||||
85
request_llms/com_qwenapi.py
Normal file
85
request_llms/com_qwenapi.py
Normal file
@ -0,0 +1,85 @@
|
||||
from http import HTTPStatus
|
||||
from toolbox import get_conf
|
||||
import threading
|
||||
import logging
|
||||
|
||||
timeout_bot_msg = '[Local Message] Request timeout. Network error.'
|
||||
|
||||
class QwenRequestInstance():
|
||||
def __init__(self):
|
||||
|
||||
self.time_to_yield_event = threading.Event()
|
||||
self.time_to_exit_event = threading.Event()
|
||||
|
||||
self.result_buf = ""
|
||||
|
||||
def generate(self, inputs, llm_kwargs, history, system_prompt):
|
||||
# import _thread as thread
|
||||
from dashscope import Generation
|
||||
QWEN_MODEL = {
|
||||
'qwen-turbo': Generation.Models.qwen_turbo,
|
||||
'qwen-plus': Generation.Models.qwen_plus,
|
||||
'qwen-max': Generation.Models.qwen_max,
|
||||
}[llm_kwargs['llm_model']]
|
||||
top_p = llm_kwargs.get('top_p', 0.8)
|
||||
if top_p == 0: top_p += 1e-5
|
||||
if top_p == 1: top_p -= 1e-5
|
||||
|
||||
self.result_buf = ""
|
||||
responses = Generation.call(
|
||||
model=QWEN_MODEL,
|
||||
messages=generate_message_payload(inputs, llm_kwargs, history, system_prompt),
|
||||
top_p=top_p,
|
||||
temperature=llm_kwargs.get('temperature', 1.0),
|
||||
result_format='message',
|
||||
stream=True,
|
||||
incremental_output=True
|
||||
)
|
||||
|
||||
for response in responses:
|
||||
if response.status_code == HTTPStatus.OK:
|
||||
if response.output.choices[0].finish_reason == 'stop':
|
||||
yield self.result_buf
|
||||
break
|
||||
elif response.output.choices[0].finish_reason == 'length':
|
||||
self.result_buf += "[Local Message] 生成长度过长，后续输出被截断"
|
||||
yield self.result_buf
|
||||
break
|
||||
else:
|
||||
self.result_buf += response.output.choices[0].message.content
|
||||
yield self.result_buf
|
||||
else:
|
||||
self.result_buf += f"[Local Message] 请求错误：状态码：{response.status_code}，错误码:{response.code}，消息：{response.message}"
|
||||
yield self.result_buf
|
||||
break
|
||||
logging.info(f'[raw_input] {inputs}')
|
||||
logging.info(f'[response] {self.result_buf}')
|
||||
return self.result_buf
|
||||
|
||||
|
||||
def generate_message_payload(inputs, llm_kwargs, history, system_prompt):
|
||||
conversation_cnt = len(history) // 2
|
||||
if system_prompt == '': system_prompt = 'Hello!'
|
||||
messages = [{"role": "user", "content": system_prompt}, {"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
|
||||
@ -1,4 +1 @@
|
||||
modelscope
|
||||
transformers_stream_generator
|
||||
auto-gptq
|
||||
optimum
|
||||
dashscope
|
||||
4
request_llms/requirements_qwen_local.txt
Normal file
4
request_llms/requirements_qwen_local.txt
Normal file
@ -0,0 +1,4 @@
|
||||
modelscope
|
||||
transformers_stream_generator
|
||||
auto-gptq
|
||||
optimum
|
||||
@ -18,7 +18,7 @@ if __name__ == "__main__":
|
||||
# from request_llms.bridge_internlm import predict_no_ui_long_connection
|
||||
# from request_llms.bridge_deepseekcoder import predict_no_ui_long_connection
|
||||
# from request_llms.bridge_qwen_7B import predict_no_ui_long_connection
|
||||
from request_llms.bridge_qwen import predict_no_ui_long_connection
|
||||
from request_llms.bridge_qwen_local import predict_no_ui_long_connection
|
||||
# from request_llms.bridge_spark import predict_no_ui_long_connection
|
||||
# from request_llms.bridge_zhipu import predict_no_ui_long_connection
|
||||
# from request_llms.bridge_chatglm3 import predict_no_ui_long_connection
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user