Update app.py

app.py

@@ -1,11 +1,9 @@
 import gradio as gr
 from transformers import AutoTokenizer, AutoModel, GPT2LMHeadModel, GPT2Tokenizer
 import torch
-from sklearn.metrics.pairwise import cosine_similarity
-import numpy as np
 
 # Load the bi-encoder model and tokenizer
-bi_encoder_model_name = "nasa-impact/nasa-smd-ibm-st-v2"
+bi_encoder_model_name = "sentence-transformers/all-MiniLM-L6-v2"
 bi_tokenizer = AutoTokenizer.from_pretrained(bi_encoder_model_name)
 bi_model = AutoModel.from_pretrained(bi_encoder_model_name)
 
@@ -20,28 +18,36 @@ def encode_text(text):
     # Ensure the output is 2D by averaging the last hidden state along the sequence dimension
     return outputs.last_hidden_state.mean(dim=1).detach().numpy()
 
-def generate_response(user_input):
-    # Encode the user input
-    user_embedding = encode_text(user_input)
+def generate_response(user_input, context_embedding):
+    # Combine user input with context embedding for GPT-2 input
+    combined_input = user_input + " " + context_embedding
     
-    # Generate a response using GPT-2
-    gpt2_inputs = gpt2_tokenizer.encode(user_input, return_tensors='pt')
-    gpt2_outputs = gpt2_model.generate(gpt2_inputs, max_length=150, num_return_sequences=1)
+    # Generate a response using GPT-2 with adjusted parameters
+    gpt2_inputs = gpt2_tokenizer.encode(combined_input, return_tensors='pt')
+    gpt2_outputs = gpt2_model.generate(
+        gpt2_inputs,
+        max_length=150,
+        num_return_sequences=1,
+        temperature=0.5,
+        top_p=0.9,
+        repetition_penalty=1.2
+    )
     generated_text = gpt2_tokenizer.decode(gpt2_outputs[0], skip_special_tokens=True)
     
     return generated_text
 
-def chatbot(user_input):
-    response = generate_response(user_input)
+def chatbot(user_input, context=""):
+    context_embedding = encode_text(context) if context else ""
+    response = generate_response(user_input, context_embedding)
     return response
 
 # Create the Gradio interface
 iface = gr.Interface(
     fn=chatbot,
-    inputs=gr.Textbox(lines=2, placeholder="Enter your message here..."),
+    inputs=[gr.Textbox(lines=2, placeholder="Enter your message here..."), gr.Textbox(lines=2, placeholder="Enter context here (optional)...")],
     outputs="text",
-    title="Dynamic Response Chatbot",
-    description="A chatbot using a bi-encoder model to understand the input and GPT-2 to generate dynamic responses."
+    title="Context-Aware Dynamic Response Chatbot",
+    description="A chatbot using a bi-encoder model to understand the input context and GPT-2 to generate dynamic responses."
 )
 
 # Launch the interface
@@ -49,3 +55,4 @@ iface.launch()
 
 
 
+