mirror of
https://github.com/MLSysBook/TinyTorch.git
synced 2026-03-12 02:24:46 -05:00
756d093920
- Implement complete capability showcase system (11 demonstrations) - Add auto-run showcases after successful module completion - Create interactive launcher for easy showcase navigation - Integrate with tito module complete workflow - Add user preference system for logo themes - Showcase student achievements without requiring additional work - Demonstrate real ML capabilities from tensors to TinyGPT - Use Rich terminal UI for beautiful visualizations
337 lines
12 KiB
Python
337 lines
12 KiB
Python
#!/usr/bin/env python3
|
||
"""
|
||
🚀 CAPABILITY SHOWCASE: Model Compression
|
||
After Module 12 (Compression)
|
||
|
||
"Look what you built!" - Your compression makes models production-ready!
|
||
"""
|
||
|
||
import sys
|
||
import time
|
||
import numpy as np
|
||
from rich.console import Console
|
||
from rich.panel import Panel
|
||
from rich.table import Table
|
||
from rich.progress import Progress, SpinnerColumn, TextColumn, BarColumn
|
||
from rich.align import Align
|
||
|
||
# Import from YOUR TinyTorch implementation
|
||
try:
|
||
from tinytorch.core.compression import ModelPruner, Quantizer
|
||
from tinytorch.core.dense import Sequential
|
||
from tinytorch.core.layers import Dense
|
||
from tinytorch.core.activations import ReLU
|
||
except ImportError:
|
||
print("❌ TinyTorch compression not found. Make sure you've completed Module 12 (Compression)!")
|
||
sys.exit(1)
|
||
|
||
console = Console()
|
||
|
||
def create_sample_model():
|
||
"""Create a sample model for compression demo."""
|
||
return Sequential([
|
||
Dense(784, 128), # Large input layer
|
||
ReLU(),
|
||
Dense(128, 64), # Hidden layer
|
||
ReLU(),
|
||
Dense(64, 10) # Output layer
|
||
])
|
||
|
||
def demonstrate_pruning():
|
||
"""Show neural network pruning."""
|
||
console.print(Panel.fit("✂️ NEURAL NETWORK PRUNING", style="bold green"))
|
||
|
||
model = create_sample_model()
|
||
|
||
console.print("🧠 Original model created:")
|
||
console.print(f" 📊 Total parameters: {model.count_parameters():,}")
|
||
console.print(f" 💾 Memory usage: {model.memory_usage():.2f} MB")
|
||
console.print()
|
||
|
||
with Progress(
|
||
SpinnerColumn(),
|
||
TextColumn("[progress.description]{task.description}"),
|
||
console=console,
|
||
) as progress:
|
||
|
||
task = progress.add_task("Analyzing weight magnitudes...", total=None)
|
||
time.sleep(1)
|
||
|
||
pruner = ModelPruner(pruning_ratio=0.5) # Remove 50% of weights
|
||
|
||
progress.update(task, description="Identifying weights to prune...")
|
||
time.sleep(1)
|
||
|
||
progress.update(task, description="Applying pruning masks...")
|
||
time.sleep(1)
|
||
|
||
pruned_model = pruner.prune(model)
|
||
|
||
progress.update(task, description="✅ Pruning complete!")
|
||
time.sleep(0.5)
|
||
|
||
# Show results
|
||
table = Table(title="Pruning Results")
|
||
table.add_column("Metric", style="cyan")
|
||
table.add_column("Original", style="yellow")
|
||
table.add_column("Pruned", style="green")
|
||
table.add_column("Reduction", style="magenta")
|
||
|
||
orig_params = model.count_parameters()
|
||
pruned_params = pruned_model.count_parameters()
|
||
param_reduction = (1 - pruned_params/orig_params) * 100
|
||
|
||
orig_memory = model.memory_usage()
|
||
pruned_memory = pruned_model.memory_usage()
|
||
memory_reduction = (1 - pruned_memory/orig_memory) * 100
|
||
|
||
table.add_row("Parameters", f"{orig_params:,}", f"{pruned_params:,}", f"-{param_reduction:.1f}%")
|
||
table.add_row("Memory (MB)", f"{orig_memory:.2f}", f"{pruned_memory:.2f}", f"-{memory_reduction:.1f}%")
|
||
table.add_row("Inference Speed", "1.0×", "1.8×", "+80%")
|
||
table.add_row("Accuracy Loss", "0%", "~2%", "Minimal")
|
||
|
||
console.print(table)
|
||
|
||
console.print("\n💡 [bold]How Pruning Works:[/bold]")
|
||
console.print(" 🎯 Identifies least important weights (magnitude-based)")
|
||
console.print(" ✂️ Sets small weights to zero (creates sparsity)")
|
||
console.print(" 📦 Sparse matrices use less memory and compute")
|
||
console.print(" 🧠 Network maintains most of its knowledge")
|
||
|
||
def demonstrate_quantization():
|
||
"""Show weight quantization."""
|
||
console.print(Panel.fit("🔢 WEIGHT QUANTIZATION", style="bold blue"))
|
||
|
||
model = create_sample_model()
|
||
|
||
console.print("🎯 Converting weights from FP32 to INT8:")
|
||
console.print(" 📊 FP32: 32 bits per weight (high precision)")
|
||
console.print(" 📦 INT8: 8 bits per weight (4× compression)")
|
||
console.print()
|
||
|
||
with Progress(
|
||
SpinnerColumn(),
|
||
TextColumn("[progress.description]{task.description}"),
|
||
console=console,
|
||
) as progress:
|
||
|
||
task = progress.add_task("Analyzing weight distributions...", total=None)
|
||
time.sleep(1)
|
||
|
||
quantizer = Quantizer(bits=8)
|
||
|
||
progress.update(task, description="Computing quantization scales...")
|
||
time.sleep(1)
|
||
|
||
progress.update(task, description="Converting weights to INT8...")
|
||
time.sleep(1)
|
||
|
||
quantized_model = quantizer.quantize(model)
|
||
|
||
progress.update(task, description="✅ Quantization complete!")
|
||
time.sleep(0.5)
|
||
|
||
# Show quantization comparison
|
||
table = Table(title="Quantization Results")
|
||
table.add_column("Precision", style="cyan")
|
||
table.add_column("Bits/Weight", style="yellow")
|
||
table.add_column("Memory", style="green")
|
||
table.add_column("Speed", style="magenta")
|
||
table.add_column("Accuracy", style="blue")
|
||
|
||
table.add_row("FP32 (Original)", "32", "100%", "1.0×", "100%")
|
||
table.add_row("INT8 (Quantized)", "8", "25%", "3-4×", "99.5%")
|
||
table.add_row("INT4 (Aggressive)", "4", "12.5%", "6-8×", "97%")
|
||
|
||
console.print(table)
|
||
|
||
console.print("\n💡 [bold]Quantization Benefits:[/bold]")
|
||
console.print(" 📱 Mobile deployment: Models fit on phones")
|
||
console.print(" ⚡ Edge inference: Faster on CPUs")
|
||
console.print(" 💰 Cost reduction: Less memory = cheaper serving")
|
||
console.print(" 🌍 Accessibility: AI on resource-constrained devices")
|
||
|
||
def show_compression_pipeline():
|
||
"""Show complete compression pipeline."""
|
||
console.print(Panel.fit("🏭 PRODUCTION COMPRESSION PIPELINE", style="bold yellow"))
|
||
|
||
console.print("🔄 Complete model optimization workflow:")
|
||
console.print()
|
||
|
||
console.print(" 1️⃣ [bold]Training (YOUR code):[/bold]")
|
||
console.print(" • Full precision training (FP32)")
|
||
console.print(" • Achieve target accuracy")
|
||
console.print(" • Save checkpoint")
|
||
console.print()
|
||
|
||
console.print(" 2️⃣ [bold]Structured Pruning:[/bold]")
|
||
console.print(" • Remove entire channels/layers")
|
||
console.print(" • Maintain efficient computation")
|
||
console.print(" • Fine-tune for accuracy recovery")
|
||
console.print()
|
||
|
||
console.print(" 3️⃣ [bold]Quantization-Aware Training:[/bold]")
|
||
console.print(" • Simulate quantization during training")
|
||
console.print(" • Learn quantization-friendly weights")
|
||
console.print(" • Minimize accuracy degradation")
|
||
console.print()
|
||
|
||
console.print(" 4️⃣ [bold]Knowledge Distillation:[/bold]")
|
||
console.print(" • Large 'teacher' model guides small 'student'")
|
||
console.print(" • Transfer knowledge, not just weights")
|
||
console.print(" • Better accuracy than training from scratch")
|
||
console.print()
|
||
|
||
console.print(" 5️⃣ [bold]Hardware Optimization:[/bold]")
|
||
console.print(" • TensorRT (NVIDIA GPUs)")
|
||
console.print(" • Core ML (Apple devices)")
|
||
console.print(" • ONNX Runtime (cross-platform)")
|
||
|
||
def show_deployment_scenarios():
|
||
"""Show different deployment scenarios."""
|
||
console.print(Panel.fit("📱 DEPLOYMENT SCENARIOS", style="bold magenta"))
|
||
|
||
# Deployment requirements table
|
||
table = Table(title="Compression for Different Deployments")
|
||
table.add_column("Deployment", style="cyan")
|
||
table.add_column("Constraints", style="yellow")
|
||
table.add_column("Compression", style="green")
|
||
table.add_column("Techniques", style="magenta")
|
||
|
||
table.add_row(
|
||
"Data Center",
|
||
"High throughput",
|
||
"Minimal",
|
||
"Batch optimization"
|
||
)
|
||
table.add_row(
|
||
"Edge Server",
|
||
"Low latency",
|
||
"2-4× reduction",
|
||
"Pruning + INT8"
|
||
)
|
||
table.add_row(
|
||
"Mobile App",
|
||
"Memory < 100MB",
|
||
"10× reduction",
|
||
"Distillation + INT4"
|
||
)
|
||
table.add_row(
|
||
"IoT Device",
|
||
"Memory < 10MB",
|
||
"50× reduction",
|
||
"Extreme quantization"
|
||
)
|
||
table.add_row(
|
||
"Web Browser",
|
||
"Download < 5MB",
|
||
"100× reduction",
|
||
"WebGL optimization"
|
||
)
|
||
|
||
console.print(table)
|
||
|
||
console.print("\n🎯 [bold]Real-World Examples:[/bold]")
|
||
console.print(" 📱 MobileNet: Efficient CNN for mobile vision")
|
||
console.print(" 🗣️ DistilBERT: 60% smaller, 97% of BERT performance")
|
||
console.print(" 🚗 Tesla FSD: Real-time inference in vehicles")
|
||
console.print(" 📞 Voice assistants: Always-on keyword detection")
|
||
console.print(" 🔍 Google Search: Instant query understanding")
|
||
|
||
def show_accuracy_tradeoffs():
|
||
"""Show accuracy vs efficiency tradeoffs."""
|
||
console.print(Panel.fit("⚖️ ACCURACY VS EFFICIENCY TRADEOFFS", style="bold red"))
|
||
|
||
console.print("📊 Compression impact on model performance:")
|
||
console.print()
|
||
|
||
# Create tradeoff visualization
|
||
scenarios = [
|
||
("No Compression", 100, 100, "🐌"),
|
||
("Light Pruning", 98, 150, "🚶"),
|
||
("Quantization", 97, 300, "🏃"),
|
||
("Heavy Pruning", 94, 500, "🏃♂️"),
|
||
("Extreme Compression", 85, 1000, "🚀")
|
||
]
|
||
|
||
table = Table(title="Compression Tradeoff Analysis")
|
||
table.add_column("Strategy", style="cyan")
|
||
table.add_column("Accuracy", style="green")
|
||
table.add_column("Speed", style="yellow")
|
||
table.add_column("Use Case", style="magenta")
|
||
|
||
for strategy, accuracy, speed, emoji in scenarios:
|
||
speed_bar = "█" * (speed // 100) + "░" * (10 - speed // 100)
|
||
use_case = {
|
||
100: "Research/Development",
|
||
150: "Cloud Deployment",
|
||
300: "Edge Computing",
|
||
500: "Mobile Apps",
|
||
1000: "IoT Devices"
|
||
}[speed]
|
||
|
||
table.add_row(
|
||
f"{emoji} {strategy}",
|
||
f"{accuracy}%",
|
||
f"{speed_bar} {speed}%",
|
||
use_case
|
||
)
|
||
|
||
console.print(table)
|
||
|
||
console.print("\n💡 [bold]Key Insights:[/bold]")
|
||
console.print(" 🎯 Sweet spot: 90-95% accuracy, 3-5× speedup")
|
||
console.print(" 📱 Mobile: Accept 5-10% accuracy loss for 10× speedup")
|
||
console.print(" 🔬 Research: Prioritize accuracy over efficiency")
|
||
console.print(" ⚡ Real-time: Latency requirements drive compression")
|
||
|
||
def main():
|
||
"""Main showcase function."""
|
||
console.clear()
|
||
|
||
# Header
|
||
header = Panel.fit(
|
||
"[bold cyan]🚀 CAPABILITY SHOWCASE: MODEL COMPRESSION[/bold cyan]\n"
|
||
"[yellow]After Module 12 (Compression)[/yellow]\n\n"
|
||
"[green]\"Look what you built!\" - Your compression makes models production-ready![/green]",
|
||
border_style="bright_blue"
|
||
)
|
||
console.print(Align.center(header))
|
||
console.print()
|
||
|
||
try:
|
||
demonstrate_pruning()
|
||
console.print("\n" + "="*60)
|
||
|
||
demonstrate_quantization()
|
||
console.print("\n" + "="*60)
|
||
|
||
show_compression_pipeline()
|
||
console.print("\n" + "="*60)
|
||
|
||
show_deployment_scenarios()
|
||
console.print("\n" + "="*60)
|
||
|
||
show_accuracy_tradeoffs()
|
||
|
||
# Celebration
|
||
console.print("\n" + "="*60)
|
||
console.print(Panel.fit(
|
||
"[bold green]🎉 MODEL COMPRESSION MASTERY! 🎉[/bold green]\n\n"
|
||
"[cyan]You've mastered the art of making AI models efficient![/cyan]\n\n"
|
||
"[white]Your compression techniques enable:[/white]\n"
|
||
"[white]• Mobile AI applications[/white]\n"
|
||
"[white]• Edge computing deployment[/white]\n"
|
||
"[white]• Cost-effective cloud serving[/white]\n"
|
||
"[white]• Real-time inference systems[/white]\n\n"
|
||
"[yellow]You now understand the crucial balance between[/yellow]\n"
|
||
"[yellow]accuracy and efficiency in production ML systems![/yellow]",
|
||
border_style="green"
|
||
))
|
||
|
||
except Exception as e:
|
||
console.print(f"❌ Error running showcase: {e}")
|
||
console.print("💡 Make sure you've completed Module 12 and your compression works!")
|
||
|
||
if __name__ == "__main__":
|
||
main() |