Update app.py

app.py

@@ -9,8 +9,8 @@ import faiss
 import numpy as np
 from sentence_transformers import SentenceTransformer
 from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline, BitsAndBytesConfig
-from langchain.document_loaders import DirectoryLoader, TextLoader  # Import these from langchain
-from langchain.text_splitter import RecursiveCharacterTextSplitter  # Import the text splitter
+from langchain.document_loaders import DirectoryLoader, TextLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
 
 class DocumentRetrievalAndGeneration:
     def __init__(self, embedding_model_name, lm_model_id, data_folder):
@@ -48,7 +48,7 @@ class DocumentRetrievalAndGeneration:
             bnb_4bit_compute_dtype=torch.bfloat16
         )
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-        model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=bnb_config)
+        model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=bnb_config).to(device)
         tokenizer = AutoTokenizer.from_pretrained(model_id)
         generate_text = pipeline(
             model=model,
@@ -94,7 +94,8 @@ class DocumentRetrievalAndGeneration:
                 content += "-" * 50 + "\n"
                 content += self.all_splits[idx].page_content + "\n"
 
-            prompt = f"""<s>
+            prompt = f"""
+            <s>
             Here's my question:
             Query: {query}
             Solution:
@@ -103,7 +104,7 @@ class DocumentRetrievalAndGeneration:
             """
             messages = [{"role": "user", "content": prompt}]
             encodeds = self.llm.tokenizer.apply_chat_template(messages, return_tensors="pt")
-            model_inputs = encodeds.to(self.llm.device)
+            model_inputs = encodeds.to(self.llm.model.device)
 
             start_time = datetime.now()
             generated_ids = self.generate_response_with_timeout(model_inputs)
@@ -120,9 +121,63 @@ class DocumentRetrievalAndGeneration:
 
             print("Generated response:", generated_response)
             print("Time elapsed:", elapsed_time)
-            print("Device in use:", self.llm.device)
+            print("Device in use:", self.llm.model.device)
 
             return solution_text, content
 
         except TimeoutError:
-            return "timeout", content
+            return "timeout", content
+
+if __name__ == "__main__":
+    # Example usage
+    embedding_model_name = 'flax-sentence-embeddings/all_datasets_v3_MiniLM-L12'
+    lm_model_id = "mistralai/Mistral-7B-Instruct-v0.2"
+    data_folder = 'sample_embedding_folder2'
+
+    doc_retrieval_gen = DocumentRetrievalAndGeneration(embedding_model_name, lm_model_id, data_folder)
+
+    # Define Gradio interface function
+    def launch_interface():
+        css_code = """
+            .gradio-container {
+                background-color: #daccdb;
+            }
+            /* Button styling for all buttons */
+            button {
+                background-color: #927fc7; /* Default color for all other buttons */
+                color: black;
+                border: 1px solid black;
+                padding: 10px;
+                margin-right: 10px;
+                font-size: 16px; /* Increase font size */
+                font-weight: bold; /* Make text bold */
+            }
+            """
+        EXAMPLES = ["Can the VIP and CSI2 modules operate simultaneously?", 
+                    "I'm using Code Composer Studio 5.4.0.00091 and enabled FPv4SPD16 floating point support for CortexM4 in TDA2. However, after building the project, the .asm file shows --float_support=vfplib instead of FPv4SPD16. Why is this happening?", 
+                    "Could you clarify the maximum number of cameras that can be connected simultaneously to the video input ports on the TDA2x SoC, considering it supports up to 10 multiplexed input ports and includes 3 dedicated video input modules?"]
+        
+        file_path = "ticketNames.txt"
+
+        # Read the file content
+        with open(file_path, "r") as file:
+            content = file.read()
+        ticket_names = json.loads(content)
+        dropdown = gr.Dropdown(label="Sample queries", choices=ticket_names)
+        
+        # Define Gradio interface
+        interface = gr.Interface(
+            fn=doc_retrieval_gen.qa_infer_gradio,
+            inputs=[gr.Textbox(label="QUERY", placeholder="Enter your query here")],
+            allow_flagging='never',
+            examples=EXAMPLES,
+            cache_examples=False,
+            outputs=[gr.Textbox(label="SOLUTION"), gr.Textbox(label="RELATED QUERIES")],
+            css=css_code
+        )
+
+        # Launch Gradio interface
+        interface.launch(debug=True)
+
+    # Launch the interface
+    launch_interface()