Update app.py

app.py

@@ -1,174 +1,40 @@
-# import gradio as gr
-# from langchain.llms import HuggingFacePipeline
-# from transformers import AutoTokenizer, AutoModel
-# import transformers
-# import torch
-# from langchain.prompts import PromptTemplate
-# from langchain.chains import LLMChain
-# import warnings
-# warnings.filterwarnings('ignore')
-
-
-# model = 'MD1998/FLAN-T5-V1'
-# tokenizer=AutoTokenizer.from_pretrained(model)
-
-# prompt_template=PromptTemplate(input_variables=["conversation"],
-#                                template="""\
-# You are a helpful, respectful, and honest assistant designed to improve English language skills. Always provide accurate and helpful responses to language improvement tasks, while ensuring safety and ethical standards. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased, positive, and focused on enhancing language skills.
-# If a question does not make sense or is not factually coherent, explain why instead of answering something incorrect. If you don't know the answer to a question, please don't share false information.
-# Your role is to guide users through various language exercises and challenges, helping them to practice and improve their English skills in a fun and engaging way. Always encourage users to try different approaches and provide constructive feedback to help them progress.
-
-# {conversation}
-# """)
-
-
-# pipeline=transformers.pipeline(
-#     "text-generation",
-#     model=model,
-#     tokenizer=tokenizer,
-#     torch_dtype=torch.bfloat16,
-#     trust_remote_code=True,
-#     device_map="auto",
-#     max_length=15,
-#     do_sample=True,
-#     top_k=10,
-#     top_p=5,
-#     num_return_sequences=1,
-#     eos_token_id=tokenizer.eos_token_id
-#     )
-
-# llm=HuggingFacePipeline(pipeline=pipeline, model_kwargs={'temperature':0.1})
-
-
-# chain = LLMChain(llm=llm, prompt=prompt_template, verbose=True)
-
-
-##########################
-
-# from transformers import T5Tokenizer
-# from transformers import T5ForConditionalGeneration
-
-# finetuned_model = T5ForConditionalGeneration.from_pretrained("MD1998/chating_beginner_v2")
-# tokenizer = T5Tokenizer.from_pretrained("MD1998/chating_beginner_v2")
-
-# # Initialize conversation history
-# conversation_history = "System_prompt: You establish that the assistant is intelligent and helpful, and that you want to have an engaging conversation.\n"
-# generation_params = {
-#     "max_length": 100,
-#     "repetition_penalty": 1.2,
-#     "temperature": 0.2,
-#     "top_p": 0.99,
-#     "top_k": 1
-# }
-
-# # Function to handle conversation
-# def chat_with_model(input_text):
-#     global conversation_history
-    
-#     # Combine the new input with the conversation history
-#     my_inputs = conversation_history + input_text
-    
-#     # Encode the inputs
-#     inputs = tokenizer(my_inputs, return_tensors="pt")
-    
-#     # Generate outputs using the model
-#     outputs = finetuned_model.generate(**inputs, **generation_params)
-    
-#     # Decode the outputs to get the answer
-#     answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    
-#     # Update conversation history (append the new input and answer)
-#     conversation_history += f"\nUser: {input_text}\nAssistant: {answer}\n"
-    
-#     # Display the answer using text wrapping for readability
-#     print(fill(answer, width=80))
-    
-#     # Return the answer for further use (if needed)
-#     return answer
-
-# # Example usage
-# # user_input = "What is the weather like today?"
-# # chat_with_model(user_input)
-
-
-# def greet(user_input):
-#     response = chat_with_model(user_input)
-    
-#     return response
-
-# iface = gr.Interface(fn=greet, inputs="text", outputs="text")
-# iface.launch()
-##########################
-
-
-
-
-
-
-
-
-
-
-
+from langchain import PromptTemplate
+from langchain import LLMChain
+from langchain.llms import CTransformers
+import gradio as gr
 
-# def greet(prompt):
-#     response = chain.run(prompt)
-    
-#     return response
+B_INST, E_INST = "[INST]", "[/INST]"
+B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"
 
+DEFAULT_SYSTEM_PROMPT="\
+You are a helpful, respectful, and honest assistant designed to improve English language skills. Your name is Nemo\
+Always provide accurate and helpful responses to language improvement tasks, while ensuring safety and ethical standards. \
+Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. \
+Please ensure that your responses are socially unbiased, positive, and focused on enhancing language skills. \
+If a question does not make sense or is not factually coherent, explain why instead of answering something incorrect. \
+If you don't know the answer to a question, please don't share false information. \
+Your role is to guide users through various language exercises and challenges, helping them to practice and improve their English skills in a fun and engaging way. \
+Always encourage users to try different approaches and provide constructive feedback to help them progress."
 
+instruction = "Have a good conversation: \n\n {text}"
 
-# iface = gr.Interface(fn=greet, inputs="text", outputs="text")
-# iface.launch()
 
-#@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
-from transformers import pipeline, T5Tokenizer
+SYSTEM_PROMPT = B_SYS + DEFAULT_SYSTEM_PROMPT + E_SYS
 
-# Load the tokenizer and the pipeline for text generation
-tokenizer = T5Tokenizer.from_pretrained("MD1998/chating_beginner_v2")
-chat_pipeline = pipeline("text2text-generation", model="MD1998/chating_beginner_v2", tokenizer=tokenizer)
+template = B_INST + SYSTEM_PROMPT + instruction + E_INST
 
-# Initialize conversation history
-conversation_history = "System_prompt: You are a helpful, respectful, and honest assistant designed to improve English language skills. Always provide accurate and helpful responses to language improvement tasks, while ensuring safety and ethical standards. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased, positive, and focused on enhancing language skills. If a question does not make sense or is not factually coherent, explain why instead of answering something incorrect. If you don't know the answer to a question, please don't share false information. Your role is to guide users through various language exercises and challenges, helping them to practice and improve their English skills in a fun and engaging way. Always encourage users to try different approaches and provide constructive feedback to help them progress.\n"
-generation_params = {
-    "max_length": 100,
-    "repetition_penalty": 1.2,
-    "temperature": 0.2,
-    "top_p": 0.99,
-    "top_k": 1
-}
+prompt = PromptTemplate(template=template, input_variables=["text"])
 
-# Function to handle conversation
-def chat_with_model(input_text):
-    global conversation_history
-    
-    # Combine the new input with the conversation history
-    prompt = conversation_history + input_text
-    
-    # Generate outputs using the pipeline with the provided prompt and generation parameters
-    response = chat_pipeline(prompt, **generation_params)[0]["generated_text"]
-    
-    # Update conversation history (append the new input and answer)
-    conversation_history += f"\nUser: {input_text}\nAssistant: {response}\n"
-    
-    # Display the answer using text wrapping for readability
-    print(fill(response, width=80))
-    
-    # Return the answer for further use (if needed)
-    return response
+llm = CTransformers(model="TheBloke/Llama-2-7B-Chat-GGUF", model_file="llama-2-7b-chat.Q3_K_S.gguf",
+                    model_type='llama',
+                    config={'max_new_tokens': 128,
+                            'temperature': 0.01}
+                    )
 
-# Example usage
-# user_input = "What is the weather like today?"
-# chat_with_model(user_input)
+LLM_Chain = LLMChain(prompt=prompt, llm=llm)
 
-def greet(user_input):
-    response = chat_with_model(user_input)
-    return response
+def greet(prompt):
+    return LLM_Chain.run(prompt)
 
-# Launch a Gradio interface
-import gradio as gr
 iface = gr.Interface(fn=greet, inputs="text", outputs="text")
-iface.launch()
-
-
-
+iface.launch()