Update app.py

app.py

@@ -14,13 +14,51 @@ from langchain_community.llms import HuggingFacePipeline
 from langchain.chains import ConversationChain
 from langchain.memory import ConversationBufferMemory
 from langchain_community.llms import HuggingFaceEndpoint
+from huggingface_hub import InferenceClient
 import torch
+
 api_token = os.getenv("HF_TOKEN")
 
+client = InferenceClient(
+    "mistralai/Mistral-7B-Instruct-v0.3"
+)
+
 list_llm = ["meta-llama/Meta-Llama-3-8B-Instruct", "mistralai/Mistral-7B-Instruct-v0.3"]
 list_llm_simple = [os.path.basename(llm) for llm in list_llm]
 
-# Load PDF document and create doc splits
+def format_prompt(message, history):
+    prompt = "<s>"
+    for user_prompt, bot_response in history:
+        prompt += f"[INST] {user_prompt} [/INST]"
+        prompt += f" {bot_response}</s> "
+    prompt += f"[INST] {message} [/INST]"
+    return prompt
+
+def generate(prompt, history, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0):
+    temperature = float(temperature)
+    if temperature < 1e-2:
+        temperature = 1e-2
+    top_p = float(top_p)
+
+    generate_kwargs = dict(
+        temperature=temperature,
+        max_new_tokens=max_new_tokens,
+        top_p=top_p,
+        repetition_penalty=repetition_penalty,
+        do_sample=True,
+        seed=42,
+    )
+
+    formatted_prompt = format_prompt(prompt, history)
+
+    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
+    output = ""
+
+    for response in stream:
+        output += response.token.text
+        yield output
+    return output
+
 def load_doc(list_file_path, chunk_size, chunk_overlap):
     loaders = [PyPDFLoader(x) for x in list_file_path]
     pages = []
@@ -30,7 +68,6 @@ def load_doc(list_file_path, chunk_size, chunk_overlap):
     doc_splits = text_splitter.split_documents(pages)
     return doc_splits
 
-# Create vector database
 def create_db(splits, collection_name, db_type):
     embedding = HuggingFaceEmbeddings()
     
@@ -63,10 +100,8 @@ def create_db(splits, collection_name, db_type):
     
     return vectordb
 
-# Initialize langchain LLM chain
 def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, initial_prompt, progress=gr.Progress()):
     progress(0.1, desc="Initializing HF tokenizer...")
-    
     progress(0.5, desc="Initializing HF Hub...")
 
     llm = HuggingFaceEndpoint(
@@ -229,27 +264,58 @@ def demo():
                 clear_btn = gr.ClearButton([msg, chatbot], value="Clear conversation")
 
         with gr.Tab("Step 6 - Chatbot without document"):
-            with gr.Row():
-                llm_no_doc_btn = gr.Radio(list_llm_simple, 
-                    label="LLM models", value=list_llm_simple[0], type="index", info="Choose your LLM model for chatbot without document")
-            with gr.Accordion("Advanced options - LLM model", open=False):
-                with gr.Row():
-                    slider_temperature_no_doc = gr.Slider(minimum=0.01, maximum=1.0, value=0.7, step=0.1, label="Temperature", info="Model temperature", interactive=True)
-                with gr.Row():
-                    slider_maxtokens_no_doc = gr.Slider(minimum=224, maximum=4096, value=1024, step=32, label="Max Tokens", info="Model max tokens", interactive=True)
-                with gr.Row():
-                    slider_topk_no_doc = gr.Slider(minimum=1, maximum=10, value=3, step=1, label="top-k samples", info="Model top-k samples", interactive=True)
-            with gr.Row():
-                llm_no_doc_progress = gr.Textbox(value="None", label="LLM initialization for chatbot without document")
-            with gr.Row():
-                llm_no_doc_init_btn = gr.Button("Initialize LLM for Chatbot without document")
             chatbot_no_doc = gr.Chatbot(height=300)
+            additional_inputs=[
+                gr.Slider(
+                    label="Temperature",
+                    value=0.9,
+                    minimum=0.0,
+                    maximum=1.0,
+                    step=0.05,
+                    interactive=True,
+                    info="Higher values produce more diverse outputs",
+                ),
+                gr.Slider(
+                    label="Max new tokens",
+                    value=256,
+                    minimum=0,
+                    maximum=1048,
+                    step=64,
+                    interactive=True,
+                    info="The maximum numbers of new tokens",
+                ),
+                gr.Slider(
+                    label="Top-p (nucleus sampling)",
+                    value=0.90,
+                    minimum=0.0,
+                    maximum=1,
+                    step=0.05,
+                    interactive=True,
+                    info="Higher values sample more low-probability tokens",
+                ),
+                gr.Slider(
+                    label="Repetition penalty",
+                    value=1.2,
+                    minimum=1.0,
+                    maximum=2.0,
+                    step=0.05,
+                    interactive=True,
+                    info="Penalize repeated tokens",
+                )
+            ]
             with gr.Row():
                 msg_no_doc = gr.Textbox(placeholder="Type message to chat with lucIAna", container=True)
             with gr.Row():
                 submit_btn_no_doc = gr.Button("Submit message")
                 clear_btn_no_doc = gr.ClearButton([msg_no_doc, chatbot_no_doc], value="Clear conversation")
 
+            gr.ChatInterface(
+                fn=generate,
+                chatbot=chatbot_no_doc,
+                additional_inputs=additional_inputs,
+                title="Mistral 7B v0.3"
+            )
+
         # Preprocessing events
         db_btn.click(initialize_database, 
             inputs=[document, slider_chunk_size, slider_chunk_overlap, db_type_radio], 
@@ -257,7 +323,7 @@ def demo():
         set_prompt_btn.click(lambda prompt: gr.update(value=prompt), 
                              inputs=prompt_input, 
                              outputs=initial_prompt)
-        qachain_btn.click(initialize_LLM, 
+        qachain_btn.click(initialize_llmchain, 
             inputs=[llm_btn, slider_temperature, slider_maxtokens, slider_topk, vector_db, initial_prompt], 
             outputs=[qa_chain, llm_progress]).then(lambda:[None,"",0,"",0,"",0], 
             inputs=None, 
@@ -279,10 +345,6 @@ def demo():
             queue=False)
 
         # Initialize LLM without document for conversation
-        llm_no_doc_init_btn.click(initialize_llm_no_doc, 
-            inputs=[llm_no_doc_btn, slider_temperature_no_doc, slider_maxtokens_no_doc, slider_topk_no_doc, initial_prompt], 
-            outputs=[llm_no_doc, llm_no_doc_progress])
-
         submit_btn_no_doc.click(conversation_no_doc, 
             inputs=[llm_no_doc, msg_no_doc, chatbot_no_doc], 
             outputs=[llm_no_doc, msg_no_doc, chatbot_no_doc],