Delete src/App3.py

src/App3.py

@@ -1,330 +0,0 @@
-import streamlit as st
-import pandas as pd
-import re
-import plotly.express as px
-import plotly.graph_objects as go
-from typing import Optional, Tuple
-
-# ✅ MUST be the first Streamlit command
-st.set_page_config(
-    page_title="LLM Compatibility Advisor", 
-    layout="wide",
-    page_icon="🧠",
-    initial_sidebar_state="expanded"
-)
-
-# Enhanced data loading with error handling
-@st.cache_data
-def load_data():
-    try:
-        df = pd.read_excel("ICFAI.xlsx", sheet_name="Form Responses 1")
-        df.columns = df.columns.str.strip()
-        return df, None
-    except FileNotFoundError:
-        return None, "Excel file 'ICFAI.xlsx' not found. Please upload the file."
-    except Exception as e:
-        return None, f"Error loading data: {str(e)}"
-
-# Enhanced RAM extraction with better parsing
-def extract_numeric_ram(ram) -> Optional[int]:
-    if pd.isna(ram):
-        return None
-    
-    ram_str = str(ram).lower().replace(" ", "")
-    
-    # Handle various formats: "8GB", "8 GB", "8gb", "8192MB", etc.
-    gb_match = re.search(r"(\d+(?:\.\d+)?)(?:gb|g)", ram_str)
-    if gb_match:
-        return int(float(gb_match.group(1)))
-    
-    # Handle MB format
-    mb_match = re.search(r"(\d+)(?:mb|m)", ram_str)
-    if mb_match:
-        return max(1, int(int(mb_match.group(1)) / 1024))  # Convert MB to GB
-    
-    # Handle plain numbers (assume GB)
-    plain_match = re.search(r"(\d+)", ram_str)
-    if plain_match:
-        return int(plain_match.group(1))
-    
-    return None
-
-# Enhanced LLM recommendation with performance tiers
-def recommend_llm(ram_str) -> Tuple[str, str, str]:
-    """Returns (recommendation, performance_tier, additional_info)"""
-    ram = extract_numeric_ram(ram_str)
-    
-    if ram is None:
-        return "⚪ Check exact specs or test with quantized models.", "Unknown", "Verify RAM specifications"
-    
-    if ram <= 2:
-        return ("🔸 DistilBERT, MobileBERT, TinyBERT", 
-                "Basic", 
-                "Suitable for simple NLP tasks, limited context")
-    elif ram <= 4:
-        return ("🔸 MiniLM, TinyLLaMA, DistilRoBERTa", 
-                "Basic", 
-                "Good for text classification, basic chat")
-    elif ram <= 6:
-        return ("🟠 Phi-1.5, Alpaca.cpp (3B), Mistral Tiny", 
-                "Moderate", 
-                "Decent reasoning, short conversations")
-    elif ram <= 8:
-        return ("🟠 Phi-2, Gemma 2B, LLaMA 2 7B (4-bit)", 
-                "Moderate", 
-                "Good general purpose, coding assistance")
-    elif ram <= 12:
-        return ("🟢 LLaMA 2 7B (GGUF), Mistral 7B (int4)", 
-                "Good", 
-                "Strong performance, longer contexts")
-    elif ram <= 16:
-        return ("🟢 Mixtral 8x7B (4-bit), Command R+", 
-                "Good", 
-                "Excellent reasoning, complex tasks")
-    elif ram <= 24:
-        return ("🔵 LLaMA 2 13B, Mistral 7B FP16, Gemma 7B", 
-                "High", 
-                "Professional grade, high accuracy")
-    else:
-        return ("🔵 Mixtral 8x7B Full, LLaMA 2 70B (quantized)", 
-                "High", 
-                "Top-tier performance, enterprise ready")
-
-# Enhanced OS detection with better icons
-def get_os_info(os_name) -> Tuple[str, str]:
-    """Returns (icon, clean_name)"""
-    if pd.isna(os_name):
-        return "💻", "Not specified"
-    
-    os = str(os_name).lower()
-    if "windows" in os:
-        return "🪟", os_name
-    elif "mac" in os or "darwin" in os:
-        return "🍎", os_name
-    elif "linux" in os or "ubuntu" in os:
-        return "🐧", os_name
-    elif "android" in os:
-        return "🤖", os_name
-    elif "ios" in os:
-        return "📱", os_name
-    else:
-        return "💻", os_name
-
-# Performance visualization
-def create_performance_chart(df):
-    """Create a performance distribution chart"""
-    laptop_rams = df["Laptop RAM"].apply(extract_numeric_ram).dropna()
-    mobile_rams = df["Mobile RAM"].apply(extract_numeric_ram).dropna()
-    
-    fig = go.Figure()
-    
-    fig.add_trace(go.Histogram(
-        x=laptop_rams,
-        name="Laptop RAM",
-        opacity=0.7,
-        nbinsx=10
-    ))
-    
-    fig.add_trace(go.Histogram(
-        x=mobile_rams,
-        name="Mobile RAM",
-        opacity=0.7,
-        nbinsx=10
-    ))
-    
-    fig.update_layout(
-        title="RAM Distribution Across Devices",
-        xaxis_title="RAM (GB)",
-        yaxis_title="Number of Students",
-        barmode='overlay',
-        height=400
-    )
-    
-    return fig
-
-# Main App
-st.title("🧠 LLM Compatibility Advisor")
-st.markdown("Get personalized, device-based suggestions for running LLMs efficiently!")
-
-# Load data
-df, error = load_data()
-
-if error:
-    st.error(error)
-    st.info("Please ensure the Excel file 'BITS_INTERNS.xlsx' is in the same directory as this script.")
-    st.stop()
-
-if df is None or df.empty:
-    st.error("No data found in the Excel file.")
-    st.stop()
-
-# Sidebar filters and info
-with st.sidebar:
-    st.header("🔍 Filters & Info")
-    
-    # Performance tier filter
-    performance_filter = st.multiselect(
-        "Filter by Performance Tier:",
-        ["Basic", "Moderate", "Good", "High", "Unknown"],
-        default=["Basic", "Moderate", "Good", "High", "Unknown"]
-    )
-    
-    # RAM range filter
-    st.subheader("RAM Range Filter")
-    min_ram = st.slider("Minimum RAM (GB)", 0, 32, 0)
-    max_ram = st.slider("Maximum RAM (GB)", 0, 64, 64)
-    
-    st.markdown("---")
-    st.markdown("### 📊 Quick Stats")
-    st.metric("Total Students", len(df))
-    
-    # Calculate average RAM
-    avg_laptop_ram = df["Laptop RAM"].apply(extract_numeric_ram).mean()
-    avg_mobile_ram = df["Mobile RAM"].apply(extract_numeric_ram).mean()
-    
-    if not pd.isna(avg_laptop_ram):
-        st.metric("Avg Laptop RAM", f"{avg_laptop_ram:.1f} GB")
-    if not pd.isna(avg_mobile_ram):
-        st.metric("Avg Mobile RAM", f"{avg_mobile_ram:.1f} GB")
-
-# User selection with search
-st.subheader("👤 Individual Student Analysis")
-selected_user = st.selectbox(
-    "Choose a student:",
-    options=[""] + list(df["Full Name"].unique()),
-    format_func=lambda x: "Select a student..." if x == "" else x
-)
-
-if selected_user:
-    user_data = df[df["Full Name"] == selected_user].iloc[0]
-    
-    # Enhanced user display
-    col1, col2 = st.columns(2)
-    
-    with col1:
-        st.markdown("### 💻 Laptop Configuration")
-        laptop_os_icon, laptop_os_name = get_os_info(user_data.get('Laptop Operating System'))
-        laptop_ram = user_data.get('Laptop RAM', 'Not specified')
-        laptop_rec, laptop_tier, laptop_info = recommend_llm(laptop_ram)
-        
-        st.markdown(f"**OS:** {laptop_os_icon} {laptop_os_name}")
-        st.markdown(f"**RAM:** {laptop_ram}")
-        st.markdown(f"**Performance Tier:** {laptop_tier}")
-        
-        st.success(f"**💡 Recommendation:** {laptop_rec}")
-        st.info(f"**ℹ️ Notes:** {laptop_info}")
-    
-    with col2:
-        st.markdown("### 📱 Mobile Configuration")
-        mobile_os_icon, mobile_os_name = get_os_info(user_data.get('Mobile Operating System'))
-        mobile_ram = user_data.get('Mobile RAM', 'Not specified')
-        mobile_rec, mobile_tier, mobile_info = recommend_llm(mobile_ram)
-        
-        st.markdown(f"**OS:** {mobile_os_icon} {mobile_os_name}")
-        st.markdown(f"**RAM:** {mobile_ram}")
-        st.markdown(f"**Performance Tier:** {mobile_tier}")
-        
-        st.success(f"**💡 Recommendation:** {mobile_rec}")
-        st.info(f"**ℹ️ Notes:** {mobile_info}")
-
-# Batch Analysis Section
-st.markdown("---")
-st.header("📊 Batch Analysis & Insights")
-
-# Create enhanced batch table
-df_display = df[["Full Name", "Laptop RAM", "Mobile RAM"]].copy()
-
-# Add recommendations and performance tiers
-laptop_recommendations = df["Laptop RAM"].apply(lambda x: recommend_llm(x)[0])
-mobile_recommendations = df["Mobile RAM"].apply(lambda x: recommend_llm(x)[0])
-laptop_tiers = df["Laptop RAM"].apply(lambda x: recommend_llm(x)[1])
-mobile_tiers = df["Mobile RAM"].apply(lambda x: recommend_llm(x)[1])
-
-df_display["Laptop LLM"] = laptop_recommendations
-df_display["Mobile LLM"] = mobile_recommendations
-df_display["Laptop Tier"] = laptop_tiers
-df_display["Mobile Tier"] = mobile_tiers
-
-# Filter based on sidebar selections
-laptop_ram_numeric = df["Laptop RAM"].apply(extract_numeric_ram)
-mobile_ram_numeric = df["Mobile RAM"].apply(extract_numeric_ram)
-
-# Apply filters
-mask = (
-    (laptop_tiers.isin(performance_filter) | mobile_tiers.isin(performance_filter)) &
-    ((laptop_ram_numeric.between(min_ram, max_ram)) | (mobile_ram_numeric.between(min_ram, max_ram)))
-)
-
-df_filtered = df_display[mask]
-
-# Display filtered table
-st.subheader(f"📋 Student Recommendations ({len(df_filtered)} students)")
-st.dataframe(
-    df_filtered, 
-    use_container_width=True,
-    column_config={
-        "Full Name": st.column_config.TextColumn("Student Name", width="medium"),
-        "Laptop RAM": st.column_config.TextColumn("Laptop RAM", width="small"),
-        "Mobile RAM": st.column_config.TextColumn("Mobile RAM", width="small"),
-        "Laptop LLM": st.column_config.TextColumn("Laptop Recommendation", width="large"),
-        "Mobile LLM": st.column_config.TextColumn("Mobile Recommendation", width="large"),
-        "Laptop Tier": st.column_config.TextColumn("L-Tier", width="small"),
-        "Mobile Tier": st.column_config.TextColumn("M-Tier", width="small"),
-    }
-)
-
-# Performance distribution chart
-if len(df) > 1:
-    st.subheader("📈 RAM Distribution Analysis")
-    fig = create_performance_chart(df)
-    st.plotly_chart(fig, use_container_width=True)
-
-# Performance tier summary
-st.subheader("��� Performance Tier Summary")
-tier_col1, tier_col2 = st.columns(2)
-
-with tier_col1:
-    st.markdown("**Laptop Performance Tiers:**")
-    laptop_tier_counts = laptop_tiers.value_counts()
-    for tier, count in laptop_tier_counts.items():
-        percentage = (count / len(laptop_tiers)) * 100
-        st.write(f"• {tier}: {count} students ({percentage:.1f}%)")
-
-with tier_col2:
-    st.markdown("**Mobile Performance Tiers:**")
-    mobile_tier_counts = mobile_tiers.value_counts()
-    for tier, count in mobile_tier_counts.items():
-        percentage = (count / len(mobile_tiers)) * 100
-        st.write(f"• {tier}: {count} students ({percentage:.1f}%)")
-
-# Enhanced reference table
-with st.expander("📘 Comprehensive LLM Reference Guide"):
-    st.markdown("""
-    ## RAM-to-LLM Compatibility Matrix
-    
-    | RAM Size | Performance Tier | Recommended Models | Use Cases | Additional Notes |
-    |----------|------------------|-------------------|-----------|------------------|
-    | ≤2GB | 🔸 Basic | DistilBERT, MobileBERT, TinyBERT | Simple NLP, text classification | Limited context, fast inference |
-    | 4GB | 🔸 Basic | MiniLM, TinyLLaMA, DistilRoBERTa | Basic chat, QA systems | Good for mobile deployment |
-    | 6GB | 🟠 Moderate | Phi-1.5, Alpaca.cpp (3B), Mistral Tiny | Reasoning tasks, short conversations | Balanced performance/memory |
-    | 8GB | 🟠 Moderate | Phi-2, Gemma 2B, LLaMA 2 7B (4-bit) | General purpose, coding help | Popular sweet spot |
-    | 12GB | 🟢 Good | LLaMA 2 7B (GGUF), Mistral 7B (int4) | Professional tasks, longer context | Production ready |
-    | 16GB | 🟢 Good | Mixtral 8x7B (4-bit), Command R+ | Complex reasoning, analysis | Excellent capabilities |
-    | 24GB | 🔵 High | LLaMA 2 13B, Mistral 7B FP16, Gemma 7B | High-accuracy tasks, research | Professional grade |
-    | >24GB | 🔵 High | Mixtral 8x7B Full, LLaMA 2 70B (quantized) | Enterprise applications | Top-tier performance |
-    
-    ### 🔧 Optimization Tips:
-    - **Quantization**: Use 4-bit or 8-bit quantization to reduce memory usage
-    - **GGUF Format**: Optimized format for CPU inference with lower memory overhead
-    - **Context Length**: Longer contexts require significantly more memory
-    - **Batch Size**: Reduce batch size for inference to save memory
-    """)
-
-# Footer with additional resources
-st.markdown("---")
-st.markdown("""
-### 🔗 Additional Resources
-- **Quantization Tools**: GPTQ, GGML, bitsandbytes
-- **Inference Engines**: llama.cpp, vLLM, TensorRT-LLM
-- **Model Repositories**: Hugging Face, Ollama, LM Studio
-""")