Update app.py

app.py

@@ -4,7 +4,6 @@ import streamlit as st
 import pandas as pd
 import torch
 import os
-import re
 from huggingface_hub import login
 from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer
 from peft import PeftModel, PeftConfig
@@ -55,18 +54,6 @@ model, tokenizer = load_model()
 if 'history' not in st.session_state:
     st.session_state.history = []
 
-# Placeholder MET prediction function (replace with actual model)
-def predict_met(age, sex, weight_kg, bfp):
-    try:
-        # Simulated MET prediction (replace with your model)
-        # Example: Linear regression based on AntDatpromt.csv features
-        lbm = weight_kg * (1 - (bfp / 100))
-        met = 3.5 + 0.1 * lbm - 0.05 * age + (0.3 if sex == "male" else 0.0)  # Placeholder formula
-        return round(met, 2)
-    except Exception as e:
-        st.warning(f"MET prediction error: {str(e)}")
-        return None
-
 # Prediction function
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
@@ -136,44 +123,27 @@ def generate_response(age, sex, height_cm, weight_kg, wc_cm):
 
         # Generate output
         st.write("🤖 Model response:")
-        output_container = st.empty()
-        text_streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
-        output = model.generate(
-            input_ids=input_ids,
-            attention_mask=attention_mask,
-            max_new_tokens=250,
-            temperature=0.7,
-            top_p=0.95,
-            do_sample=True,
-            pad_token_id=tokenizer.eos_token_id,
-            use_cache=True,
-            streamer=text_streamer
-        )
+        with st.empty():
+            text_streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=False)
+            output = model.generate(
+                input_ids=input_ids,
+                attention_mask=attention_mask,
+                max_new_tokens=250,
+                temperature=0.7,
+                top_p=0.95,
+                do_sample=True,
+                pad_token_id=tokenizer.eos_token_id,
+                use_cache=True,
+                streamer=text_streamer
+            )
 
         # Decode the output
         decoded = tokenizer.decode(output[0], skip_special_tokens=False)
         st.write(f"Decoded output: {decoded}")
 
-        # Extract assistant response (remove system header and tokens)
-        assistant_response = re.search(r"<|start_header_id|>assistant<|end_header_id>\n(.*)<|eot_id>", decoded, re.DOTALL)
-        clean_response = assistant_response.group(1).strip() if assistant_response else decoded
-
-        # MET prediction
-        try:
-            bfp_match = re.search(r"BFP:\s*(\d+\.\d+)%", clean_response)
-            bfp = float(bfp_match.group(1)) if bfp_match else 18.36  # Fallback
-            met_pred = predict_met(age, sex, weight_kg, bfp)
-            if met_pred is not None:
-                clean_response += f", Predicted MET: {met_pred}"
-        except Exception as e:
-            st.warning(f"MET prediction skipped: {str(e)}")
-
         # Update history
-        st.session_state.history.append((prompt, clean_response))
-
-        # Display clean response
-        output_container.write(clean_response)
-        return clean_response
+        st.session_state.history.append((prompt, decoded))
+        return decoded
 
     except Exception as e:
         st.error(f"Error during generation: {str(e)}")
@@ -241,4 +211,5 @@ with tab2:
 # Clear history button
 if st.button("Clear History"):
     st.session_state.history = []
-    st.rerun()
+    st.rerun()
+