成功借助tgui调用更多LLM

config.py

@@ -1,5 +1,5 @@
 # [step 1]>> 例如： API_KEY = "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" （此key无效）
-API_KEY = "sk-此处填API密钥"
+API_KEY = "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"
 
 # [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
 USE_PROXY = False

predict.py

@@ -244,7 +244,10 @@ def generate_payload(inputs, top_p, temperature, history, system_prompt, stream)
 
 
 if not LLM_MODEL.startswith('gpt'):
-    from request_llm.bridge_tgui import predict_tgui
+    # 函数重载到另一个文件
+    from request_llm.bridge_tgui import predict_tgui, predict_tgui_no_ui
     predict = predict_tgui
+    predict_no_ui = predict_tgui_no_ui
+    predict_no_ui_long_connection = predict_tgui_no_ui
 
     

request_llm/README.md

@@ -0,0 +1,39 @@
+# 如何使用其他大语言模型
+
+## 1. 先运行text-generation
+``` sh
+# 下载模型
+git clone https://github.com/oobabooga/text-generation-webui.git
+
+# 安装text-generation的额外依赖
+pip install accelerate bitsandbytes flexgen gradio llamacpp markdown numpy peft requests rwkv safetensors sentencepiece tqdm datasets git+https://github.com/huggingface/transformers
+
+# 切换路径
+cd text-generation-webui
+
+# 下载模型
+python download-model.py facebook/opt-1.3b
+
+# 其他可选如 facebook/galactica-1.3b
+#           facebook/galactica-6.7b
+#           facebook/galactica-120b
+
+# Pymalion 6B is a proof-of-concept dialogue model based on EleutherAI's GPT-J-6B.
+#           facebook/pygmalion-1.3b
+
+# 启动text-generation，注意把模型的斜杠改成下划线
+python server.py --cpu --listen --listen-port 7860 --model facebook_galactica-1.3b
+```
+
+## 2. 修改config.py
+```
+# LLM_MODEL格式为     [模型]@[ws地址] @[ws端口]
+LLM_MODEL = "pygmalion-1.3b@localhost@7860"
+```
+
+
+## 3. 运行！
+```
+cd chatgpt-academic
+python main.py
+```

request_llm/bridge_tgui.py

@@ -11,6 +11,7 @@ import websockets
 import logging
 import time
 import threading
+import importlib
 from toolbox import get_conf
 LLM_MODEL, = get_conf('LLM_MODEL')
 
@@ -22,7 +23,7 @@ def random_hash():
 
 async def run(context):
     params = {
-        'max_new_tokens': 200,
+        'max_new_tokens': 1024,
         'do_sample': True,
         'temperature': 0.5,
         'top_p': 0.9,
@@ -103,9 +104,10 @@ def predict_tgui(inputs, top_p, temperature, chatbot=[], history=[], system_prom
         if "PreProcess" in functional[additional_fn]: inputs = functional[additional_fn]["PreProcess"](inputs)  # 获取预处理函数（如果有的话）
         inputs = functional[additional_fn]["Prefix"] + inputs + functional[additional_fn]["Suffix"]
 
-    raw_input = inputs
+    raw_input = "What I would like to say is the following: " + inputs
     logging.info(f'[raw_input] {raw_input}')
-    chatbot.append((inputs, ""))
+    history.extend([inputs, ""])
+    chatbot.append([inputs, ""])
     yield chatbot, history, "等待响应"
 
     prompt = inputs
@@ -113,11 +115,11 @@ def predict_tgui(inputs, top_p, temperature, chatbot=[], history=[], system_prom
 
     mutable = [""]
     def run_coorotine(mutable):
-        async def get_result():
+        async def get_result(mutable):
             async for response in run(prompt):
                 # Print intermediate steps
-                mutable += response
-        asyncio.run(get_result())
+                mutable[0] = response
+        asyncio.run(get_result(mutable))
 
     thread_listen = threading.Thread(target=run_coorotine, args=(mutable,))
     thread_listen.start()
@@ -129,9 +131,25 @@ def predict_tgui(inputs, top_p, temperature, chatbot=[], history=[], system_prom
             tgui_say = mutable[0]
             history[-1] = tgui_say
             chatbot[-1] = (history[-2], history[-1])
-            yield chatbot, history, status_text
+            yield chatbot, history, "status_text"
 
     logging.info(f'[response] {tgui_say}')
 
 
-    
+
+def predict_tgui_no_ui(inputs, top_p, temperature, history=[], sys_prompt=""):
+    raw_input = "What I would like to say is the following: " + inputs
+    prompt = inputs
+    tgui_say = ""
+    mutable = [""]
+    def run_coorotine(mutable):
+        async def get_result(mutable):
+            async for response in run(prompt):
+                # Print intermediate steps
+                mutable[0] = response
+        asyncio.run(get_result(mutable))
+    thread_listen = threading.Thread(target=run_coorotine, args=(mutable,))
+    thread_listen.start()
+    thread_listen.join()
+    tgui_say = mutable[0]
+    return tgui_say