Upload 2 files

app.py

@@ -1,14 +1,16 @@
 import gradio as gr
 import time
-from llama_cpp import Llama
-from pathlib import Path
-from tqdm.auto import tqdm
+from openai import OpenAI
 import re
+import os
 
-llm = Llama(model_path="./snorkel-mistral-pairrm-dpo.Q4_K_M.gguf", 
-                  chat_format="chatml", 
-                  n_gpu_layers=0, # cpu only
-                  n_ctx=6000)
+client = OpenAI(api_key=os.environ.get("openai"))
+model="gpt-3.5-turbo"
+
+# llm = Llama(model_path="./snorkel-mistral-pairrm-dpo.Q4_K_M.gguf", 
+#                   chat_format="chatml", 
+#                   n_gpu_layers=0, # cpu only
+#                   n_ctx=6000)
 def split_text(text, llm, chunk_size):
     text_newline = text.split('\n')
     text_newline = [t for t in text_newline if len(t)>0]
@@ -83,17 +85,22 @@ def adverse(message, history):
         # response_iter
         # response_str = f'writer {len(writer_messages)}' if turn=='writer' else f'editor {len(editor_messages)}'
         # response_iter = iter(response_str.split(' '))
-        response_iter = llm.create_chat_completion(
-                        list_of_messages, # Prompt
-                        max_tokens=-1,
-                        stop=["###"],
-                        stream=True
-                        )
+        # response_iter = llm.create_chat_completion(
+        #                 list_of_messages, # Prompt
+        #                 max_tokens=-1,
+        #                 stop=["###"],
+        #                 stream=True
+        #                 )
+        response_iter = client.chat.completions.create(
+            model=model,
+            messages=list_of_messages,
+            stream=True,
+        )
         response=''
         for chunk in response_iter:
             try:
-                response+=chunk['choices'][0]['delta']['content']
-                total_response+=chunk['choices'][0]['delta']['content']
+                response+=chunk.choices[0].delta.content
+                total_response+=chunk.choices[0].delta.content
                 time.sleep(1)
                 # print(f'chunk: {chunk}')
                 yield total_response
@@ -130,19 +137,14 @@ max_words = 10_000
 print(f'max_words: {max_words}')
 # llm = Llama(model_path="E:\\yt\\bookSummary/Snorkel-Mistral-PairRM-DPO/snorkel-mistral-pairrm-dpo.Q4_K_M.gguf", chat_format="chatml", n_gpu_layers=-1, n_ctx=6000)
 
-setting = 'Scotland in twenty twenty five'
-book = 'The Great Gatsby by F. Scott Fitzgerald'
-total_summary=''
 writer_system_prompt_unformatted = '''You are a writer.
-The topic is {topic}.
-I am your writing editor. I will give you feedback to guide your response in the right direction. You will obey my feedback adjust your response based on my feedback.'''
+The topic is {topic}.'''
 
-editor_system_prompt_unformatted = '''You are an editor.
+editor_system_prompt_unformatted = '''You are an editor. You give feedback about my writing.
 The topic is {topic}.
 You should reinforce me to adjust my content and improve it. 
-You should push me to make my response as close as possible to the topic'''
+You should push me to make my response as close as possible to the topic.'''
 
-# print(writer_system_prompt.format(book=book, setting=setting))
 
 writer_messages = [{'role':'system','content':writer_system_prompt_unformatted}]
 editor_messages = [{'role':'system','content':editor_system_prompt_unformatted}]
@@ -161,11 +163,6 @@ with gr.Blocks() as demo:
     # Multi Agent LLMs for End-to-End Story Generation
     Start typing below to see the output.
     """)
-    # inp_topic = gr.Textbox(placeholder="What is the topic?", label = 'Topic', lines=4)
-    # # inp_setting = gr.Textbox(placeholder="What is the setting of the book?", label='Setting')
-    # out_system = gr.Textbox(label='System prompt to use', lines=4)
-    # button = gr.Button("Set system prompt")
-    # button.click(set_system_prompts, inputs=inp_topic, outputs=out_system)
     gr.Interface(
         fn=set_system_prompts, 
         inputs=gr.Textbox(placeholder="What is the topic?", label = 'Topic', lines=4), 

requirements.txt

@@ -1 +1 @@
-llama-cpp-python
+openai