Upload folder using huggingface_hub

chatmodel.py

@@ -41,9 +41,9 @@ class SwapChatModel(ChatModel):
         super().__init__(model,sysprompt)
         self.conversation=[]
     def __call__(self, msg:str):
+        self.conversation.append(chatmsg(msg,"assistant"))
         if "End of conversation." in [i["content"] for i in self.conversation]:
             return
-        self.conversation.append(chatmsg(msg,"assistant"))
         prompt="".join([
             self.model.start(),
             self.model.conv([chatmsg(self.sysprompt,"system")]),
@@ -80,7 +80,7 @@ class InquiryChatModel(SwapChatModel):
         ret=self.model(prompt, stop=[".","\n \n","?\n",".\n","tile|>","\n"],max_tokens=10)
         print("system prompt:",ret["choices"][0]["text"])
         if "true" in ret["choices"][0]["text"].lower():
-            self.conversation.append(chatmsg(msg,"user"))
+            self.conversation.append(chatmsg(msg,"assistant"))
             self.conversation.append(chatmsg("End of conversation.","user"))
     def __call__(self, msg:str):
         self.inquire(msg)

interactive_test.py

@@ -7,20 +7,18 @@ from models import Phi35,models
 
 sysprompt=r"""
 {{! This comment will not show up in the output}}
-The User will make an inquiry to the assistant.
-Fullfill the users inquiry.
+You are a user of an artificial assistant. You ask the artificial assistant one inquiry.
+
+Rules:
 {{#if (eq model "SwapChat")}}
-The User will write a message with his closing thoughts and the keyword "<|endtile|>" if his inquiry is fulfilled.
+- The User will write a message with his closing thoughts and the keyword "<|endtile|>" if his inquiry is fulfilled.
 {{/if}}
-The User will never have more than one inquiry in one conversation.
-The User will never complete his own inquiry.
-The User will never be a assistant.
-The User keep his message short in one sentence.
 {{#if (eq model "SwapChat")}}
-All conversations will end with "<|endtile|>".
+- All conversations will end with "<|endtile|>".
 {{/if}}
-After each User message is one assistant response.
-There can never be more than one assistant response in succession.
+- The User will never have more than one inquiry in one conversation.
+- The User will never complete his inquiry.
+
 {{#if (eq model "SwapChat")}}
 Example:
 User: What is the capital?

models.py

@@ -1,7 +1,7 @@
 from typing import Dict, List
 
 from llama_cpp import Llama
-llama_args={"n_gpu_layers":100,"main_gpu":0,"verbose":True}
+llama_args={"n_gpu_layers":100,"main_gpu":0,"verbose":False}
 
 class Model:
     def __init__(self):
@@ -23,10 +23,10 @@ class Phi35RPMax(Model):
             repo_id="ArliAI/Phi-3.5-mini-3.8B-ArliAI-RPMax-v1.1-GGUF",
             filename="ArliAI-RPMax-3.8B-v1.1-fp16.gguf",
             **llama_args,
-
         )
         
     def __call__(self, msg:str, stop:List[str], max_tokens:int):
+        print("Autocomplete: ",msg)
         ret=self.llm(msg, stop=stop, max_tokens=max_tokens)
         return ret
     
@@ -45,6 +45,7 @@ class Phi35(Model):
             **llama_args,
         )
     def __call__(self, msg:str, stop:List[str], max_tokens:int):
+        print("Autocomplete: ",msg)
         return self.llm(msg, stop=stop, max_tokens=max_tokens)
 
     def conv(self,msgs:List[Dict[str, str]]):
@@ -90,6 +91,7 @@ class Llama31uncensored(Model):
             **llama_args,
         )
     def __call__(self, msg:str, stop:List[str], max_tokens:int):
+        print("Autocomplete: ",msg)
         return self.llm(msg, stop=stop, max_tokens=max_tokens)
     
     def start(self):
@@ -111,6 +113,7 @@ class Llama31(Model):
             **llama_args,
         )
     def __call__(self, msg:str, stop:List[str], max_tokens:int):
+        print("Autocomplete: ",msg)
         return self.llm(msg, stop=stop, max_tokens=max_tokens)
     
     def conv(self,msgs:List[Dict[str, str]]):