new

request_llms/bridge_qwen_local.py

@@ -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)

request_llms/com_qwenapi.py

@@ -0,0 +1,94 @@
+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):
+        import dashscope
+        self.time_to_yield_event = threading.Event()
+        self.time_to_exit_event = threading.Event()
+        self.result_buf = ""
+
+        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')
+        dashscope.api_key = get_conf("DASHSCOPE_API_KEY")
+
+
+    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

request_llms/requirements_qwen_local.txt

@@ -0,0 +1,5 @@
+modelscope
+transformers_stream_generator
+auto-gptq
+optimum
+urllib3<2