TinyTorch/modules/source/14_benchmarking/benchmarking_dev.ipynb

Benchmarking - Systematic ML Performance Evaluation

Welcome to the Benchmarking module! This is where we learn to systematically evaluate ML systems using industry-standard methodology inspired by MLPerf.

Learning Goals

• Understand the four-component MLPerf benchmarking architecture
• Implement different benchmark scenarios (latency, throughput, offline)
• Apply statistical validation for meaningful results
• Create professional performance reports for ML projects
• Learn to avoid common benchmarking pitfalls

Build → Use → Analyze

1. Build: Benchmarking framework with proper statistical validation
2. Use: Apply systematic evaluation to your TinyTorch models
3. Analyze: Generate professional reports with statistical confidence

In [ ]:

#| default_exp core.benchmarking

#| export
import numpy as np
import matplotlib.pyplot as plt
import time
import statistics
import json
import math
from typing import Dict, List, Tuple, Optional, Any, Callable
from dataclasses import dataclass
from enum import Enum
import os
import sys

# Import our TinyTorch dependencies
try:
    from tinytorch.core.tensor import Tensor
    from tinytorch.core.networks import Sequential
    from tinytorch.core.layers import Dense
    from tinytorch.core.activations import ReLU, Softmax
    from tinytorch.core.dataloader import DataLoader
except ImportError:
    # For development, import from local modules
    parent_dirs = [
        os.path.join(os.path.dirname(__file__), '..', '01_tensor'),
        os.path.join(os.path.dirname(__file__), '..', '03_layers'),
        os.path.join(os.path.dirname(__file__), '..', '02_activations'),
        os.path.join(os.path.dirname(__file__), '..', '04_networks'),
        os.path.join(os.path.dirname(__file__), '..', '06_dataloader')
    ]
    for path in parent_dirs:
        if path not in sys.path:
            sys.path.append(path)
    
    try:
        from tensor_dev import Tensor
        from networks_dev import Sequential
        from layers_dev import Dense
        from activations_dev import ReLU, Softmax
        from dataloader_dev import DataLoader
    except ImportError:
        # Fallback for missing modules
        print("⚠️  Some TinyTorch modules not available - using minimal implementations")

In [ ]:

#| hide
#| export
def _should_show_plots():
    """Check if we should show plots (disable during testing)"""
    is_pytest = (
        'pytest' in sys.modules or
        'test' in sys.argv or
        os.environ.get('PYTEST_CURRENT_TEST') is not None or
        any('test' in arg for arg in sys.argv) or
        any('pytest' in arg for arg in sys.argv)
    )
    
    return not is_pytest

In [ ]:

print("📊 TinyTorch Benchmarking Module")
print(f"NumPy version: {np.__version__}")
print(f"Python version: {sys.version_info.major}.{sys.version_info.minor}")
print("Ready to build professional ML benchmarking tools!")

📦 Where This Code Lives in the Final Package

Learning Side: You work in modules/source/12_benchmarking/benchmarking_dev.py
Building Side: Code exports to tinytorch.core.benchmarking

# Final package structure:
from tinytorch.core.benchmarking import TinyTorchPerf, BenchmarkScenarios
from tinytorch.core.benchmarking import StatisticalValidator, PerformanceReporter

Why this matters:

• Learning: Deep understanding of systematic evaluation
• Production: Professional benchmarking methodology
• Projects: Tools for validating your ML project performance
• Career: Industry-standard skills for ML engineering roles

What is ML Benchmarking?

The Systematic Evaluation Problem

When you build ML systems, you need to answer critical questions:

• Is my model actually better? Statistical significance vs random variation
• How does it perform in production? Latency, throughput, resource usage
• Which approach should I choose? Systematic comparison methodology
• Can I trust my results? Avoiding common benchmarking pitfalls

The MLPerf Architecture

MLPerf (Machine Learning Performance) defines the industry standard for ML benchmarking:

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│  Load Generator │───▶│ System Under    │───▶│    Dataset      │
│   (Controls     │    │ Test (Your ML   │    │ (Standardized   │
│    Queries)     │    │    Model)       │    │  Evaluation)    │
└─────────────────┘    └─────────────────┘    └─────────────────┘

The Four Components

1. System Under Test (SUT): Your ML model/system being evaluated
2. Dataset: Standardized evaluation data (CIFAR-10, ImageNet, etc.)
3. Model: The specific architecture and weights being tested
4. Load Generator: Controls how evaluation queries are sent to the SUT

Why This Matters

• Reproducibility: Others can verify your results
• Comparability: Fair comparison between different approaches
• Statistical validity: Meaningful conclusions from your data
• Industry standards: Skills you'll use in ML engineering careers

Real-World Examples

• Google: Uses similar patterns for production ML system evaluation
• Meta: A/B testing frameworks follow these principles
• OpenAI: GPT model comparisons use systematic benchmarking
• Research: All major ML conferences require proper evaluation methodology

Step 1: Benchmark Scenarios - How to Measure Performance

The Three Standard Scenarios

Different use cases require different performance measurements:

1. Single-Stream Scenario

• Use case: Mobile/edge inference, interactive applications
• Pattern: Send next query only after previous completes
• Metric: 90th percentile latency (tail latency)
• Why: Users care about worst-case response time

2. Server Scenario

• Use case: Production web services, API endpoints
• Pattern: Poisson distribution of concurrent queries
• Metric: Queries per second (QPS) at acceptable latency
• Why: Servers handle multiple simultaneous requests

3. Offline Scenario

• Use case: Batch processing, data center workloads
• Pattern: Send all samples at once for batch processing
• Metric: Throughput (samples per second)
• Why: Batch jobs care about total processing time

Mathematical Foundation

Each scenario tests different aspects:

• Latency: Time for single sample = f(model_complexity, hardware)
• Throughput: Samples per second = f(parallelism, batch_size)
• Efficiency: Resource utilization = f(memory, compute, bandwidth)

Why Multiple Scenarios?

Real ML systems have different requirements:

• Chatbot: Low latency for good user experience
• Image API: High throughput for many concurrent users
• Data pipeline: Maximum batch processing efficiency

Step 2: Statistical Validation - Ensuring Meaningful Results

The Significance Problem

Common benchmarking mistakes:

# BAD: Single run, no statistical validation
result_a = model_a.run_once()  # 94.2% accuracy
result_b = model_b.run_once()  # 94.7% accuracy
print("Model B is better!")  # Maybe, maybe not...

The MLPerf Solution

Proper statistical validation:

# GOOD: Multiple runs with confidence intervals
results_a = [model_a.run() for _ in range(10)]  # [93.8, 94.1, 94.3, ...]
results_b = [model_b.run() for _ in range(10)]  # [94.2, 94.5, 94.9, ...]
significance = statistical_test(results_a, results_b)
print(f"Model B is {significance.p_value < 0.05} better with p={significance.p_value}")

Key Statistical Concepts

• Confidence intervals: Range of likely true values
• P-values: Probability that difference is due to chance
• Effect size: Magnitude of improvement (not just significance)
• Multiple comparisons: Adjusting for testing many approaches

Sample Size Calculation

MLPerf uses this formula for minimum samples:

n = Φ^(-1)((1-C)/2)^2 * p(1-p) / MOE^2

Where:

• C = confidence level (0.99)
• p = percentile (0.90 for 90th percentile)
• MOE = margin of error ((1-p)/20)

For 90th percentile with 99% confidence: n = 24,576 samples

In [ ]:

#| export
class BenchmarkScenario(Enum):
    """Standard benchmark scenarios from MLPerf"""
    SINGLE_STREAM = "single_stream"
    SERVER = "server"
    OFFLINE = "offline"

@dataclass
class BenchmarkResult:
    """Results from a benchmark run"""
    scenario: BenchmarkScenario
    latencies: List[float]  # All latency measurements in seconds
    throughput: float      # Samples per second
    accuracy: float        # Model accuracy (0-1)
    metadata: Optional[Dict[str, Any]] = None

#| export
class BenchmarkScenarios:
    """
    Implements the three standard MLPerf benchmark scenarios.
    
    TODO: Implement the three benchmark scenarios following MLPerf patterns.
    
    UNDERSTANDING THE SCENARIOS:
    1. Single-Stream: Send queries one at a time, measure latency
    2. Server: Send queries following Poisson distribution, measure QPS
    3. Offline: Send all queries at once, measure total throughput
    
    IMPLEMENTATION APPROACH:
    1. Each scenario should run the model multiple times
    2. Collect latency measurements for each run
    3. Calculate appropriate metrics for each scenario
    4. Return BenchmarkResult with all measurements
    
    EXAMPLE USAGE:
    scenarios = BenchmarkScenarios()
    result = scenarios.single_stream(model, dataset, num_queries=1000)
    print(f"90th percentile latency: {result.latencies[int(0.9 * len(result.latencies))]} seconds")
    """
    
    def __init__(self):
        self.results = []
    
    def single_stream(self, model: Callable, dataset: List, num_queries: int = 1000) -> BenchmarkResult:
        """
        Run single-stream benchmark scenario.
        
        TODO: Implement single-stream benchmarking.
        
        STEP-BY-STEP:
        1. Initialize empty list for latencies
        2. For each query (up to num_queries):
           a. Get next sample from dataset (cycle if needed)
           b. Record start time
           c. Run model on sample
           d. Record end time
           e. Calculate latency = end - start
           f. Add latency to list
        3. Calculate throughput = num_queries / total_time
        4. Calculate accuracy if possible
        5. Return BenchmarkResult with SINGLE_STREAM scenario
        
        HINTS:
        - Use time.perf_counter() for precise timing
        - Use dataset[i % len(dataset)] to cycle through samples
        - Sort latencies for percentile calculations
        """
        ### BEGIN SOLUTION
        latencies = []
        correct_predictions = 0
        total_start_time = time.perf_counter()
        
        for i in range(num_queries):
            # Get sample (cycle through dataset)
            sample = dataset[i % len(dataset)]
            
            # Time the inference
            start_time = time.perf_counter()
            result = model(sample)
            end_time = time.perf_counter()
            
            latency = end_time - start_time
            latencies.append(latency)
            
            # Simple accuracy calculation (if possible)
            if hasattr(sample, 'target') and hasattr(result, 'data'):
                predicted = np.argmax(result.data)
                if predicted == sample.target:
                    correct_predictions += 1
        
        total_time = time.perf_counter() - total_start_time
        throughput = num_queries / total_time
        accuracy = correct_predictions / num_queries if num_queries > 0 else 0.0
        
        return BenchmarkResult(
            scenario=BenchmarkScenario.SINGLE_STREAM,
            latencies=sorted(latencies),
            throughput=throughput,
            accuracy=accuracy,
            metadata={"num_queries": num_queries}
        )
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def server(self, model: Callable, dataset: List, target_qps: float = 10.0, 
               duration: float = 60.0) -> BenchmarkResult:
        """
        Run server benchmark scenario with Poisson-distributed queries.
        
        TODO: Implement server benchmarking.
        
        STEP-BY-STEP:
        1. Calculate inter-arrival time = 1.0 / target_qps
        2. Run for specified duration:
           a. Wait for next query arrival (Poisson distribution)
           b. Get sample from dataset
           c. Record start time
           d. Run model
           e. Record end time and latency
        3. Calculate actual QPS = total_queries / duration
        4. Return results
        
        HINTS:
        - Use np.random.exponential(inter_arrival_time) for Poisson
        - Track both query arrival times and completion times
        - Server scenario cares about sustained throughput
        """
        ### BEGIN SOLUTION
        latencies = []
        inter_arrival_time = 1.0 / target_qps
        start_time = time.perf_counter()
        current_time = start_time
        query_count = 0
        
        while (current_time - start_time) < duration:
            # Wait for next query (Poisson distribution)
            wait_time = np.random.exponential(inter_arrival_time)
            time.sleep(min(wait_time, 0.001))  # Small sleep to simulate waiting
            
            # Get sample
            sample = dataset[query_count % len(dataset)]
            
            # Time the inference
            query_start = time.perf_counter()
            result = model(sample)
            query_end = time.perf_counter()
            
            latency = query_end - query_start
            latencies.append(latency)
            
            query_count += 1
            current_time = time.perf_counter()
        
        actual_duration = current_time - start_time
        actual_qps = query_count / actual_duration
        
        return BenchmarkResult(
            scenario=BenchmarkScenario.SERVER,
            latencies=sorted(latencies),
            throughput=actual_qps,
            accuracy=0.0,  # Would need labels for accuracy
            metadata={"target_qps": target_qps, "actual_qps": actual_qps, "duration": actual_duration}
        )
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def offline(self, model: Callable, dataset: List, batch_size: int = 32) -> BenchmarkResult:
        """
        Run offline benchmark scenario with batch processing.
        
        TODO: Implement offline benchmarking.
        
        STEP-BY-STEP:
        1. Group dataset into batches of batch_size
        2. For each batch:
           a. Record start time
           b. Run model on entire batch
           c. Record end time
           d. Calculate batch latency
        3. Calculate total throughput = total_samples / total_time
        4. Return results
        
        HINTS:
        - Process data in batches for efficiency
        - Measure total time for all batches
        - Offline cares about maximum throughput
        """
        ### BEGIN SOLUTION
        latencies = []
        total_samples = len(dataset)
        total_start_time = time.perf_counter()
        
        for batch_start in range(0, total_samples, batch_size):
            batch_end = min(batch_start + batch_size, total_samples)
            batch = dataset[batch_start:batch_end]
            
            # Time the batch inference
            batch_start_time = time.perf_counter()
            for sample in batch:
                result = model(sample)
            batch_end_time = time.perf_counter()
            
            batch_latency = batch_end_time - batch_start_time
            latencies.append(batch_latency)
        
        total_time = time.perf_counter() - total_start_time
        throughput = total_samples / total_time
        
        return BenchmarkResult(
            scenario=BenchmarkScenario.OFFLINE,
            latencies=latencies,
            throughput=throughput,
            accuracy=0.0,  # Would need labels for accuracy
            metadata={"batch_size": batch_size, "total_samples": total_samples}
        )
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")

🧪 Unit Test: Benchmark Scenarios

Let's test our benchmark scenarios with a simple mock model.

In [ ]:

def test_benchmark_scenarios():
    """Test that our benchmark scenarios work correctly."""
    print("🔬 Unit Test: Benchmark Scenarios...")
    
    # Create a simple mock model and dataset
    def mock_model(sample):
        # Simulate some processing time
        time.sleep(0.001)  # 1ms processing
        return {"prediction": np.random.rand(10)}
    
    mock_dataset = [{"data": np.random.rand(10)} for _ in range(100)]
    
    # Test scenarios
    scenarios = BenchmarkScenarios()
    
    # Test single-stream
    single_result = scenarios.single_stream(mock_model, mock_dataset, num_queries=10)
    assert single_result.scenario == BenchmarkScenario.SINGLE_STREAM
    assert len(single_result.latencies) == 10
    assert single_result.throughput > 0
    print(f"✅ Single-stream: {len(single_result.latencies)} measurements")
    
    # Test server (short duration for testing)
    server_result = scenarios.server(mock_model, mock_dataset, target_qps=5.0, duration=2.0)
    assert server_result.scenario == BenchmarkScenario.SERVER
    assert len(server_result.latencies) > 0
    assert server_result.throughput > 0
    print(f"✅ Server: {len(server_result.latencies)} queries processed")
    
    # Test offline
    offline_result = scenarios.offline(mock_model, mock_dataset, batch_size=5)
    assert offline_result.scenario == BenchmarkScenario.OFFLINE
    assert len(offline_result.latencies) > 0
    assert offline_result.throughput > 0
    print(f"✅ Offline: {len(offline_result.latencies)} batches processed")
    
    print("✅ All benchmark scenarios working correctly!")

# Run the test
test_benchmark_scenarios()

Step 3: Statistical Validation - Ensuring Meaningful Results

The Confidence Problem

How do we know if one model is actually better than another?

Statistical Testing for ML

We need to test the null hypothesis: "There is no significant difference between models"

In [ ]:

#| export
@dataclass
class StatisticalValidation:
    """Results from statistical validation"""
    is_significant: bool
    p_value: float
    effect_size: float
    confidence_interval: Tuple[float, float]
    recommendation: str

#| export
class StatisticalValidator:
    """
    Validates benchmark results using proper statistical methods.
    
    TODO: Implement statistical validation for benchmark results.
    
    UNDERSTANDING STATISTICAL TESTING:
    1. Null hypothesis: No difference between models
    2. T-test: Compare means of two groups
    3. P-value: Probability of seeing this difference by chance
    4. Effect size: Magnitude of the difference
    5. Confidence interval: Range of likely true values
    
    IMPLEMENTATION APPROACH:
    1. Calculate basic statistics (mean, std, n)
    2. Perform t-test to get p-value
    3. Calculate effect size (Cohen's d)
    4. Calculate confidence interval
    5. Provide clear recommendation
    """
    
    def __init__(self, confidence_level: float = 0.95):
        self.confidence_level = confidence_level
        self.alpha = 1 - confidence_level
    
    def validate_comparison(self, results_a: List[float], results_b: List[float]) -> StatisticalValidation:
        """
        Compare two sets of benchmark results statistically.
        
        TODO: Implement statistical comparison.
        
        STEP-BY-STEP:
        1. Calculate basic statistics for both groups
        2. Perform two-sample t-test
        3. Calculate effect size (Cohen's d)
        4. Calculate confidence interval for the difference
        5. Generate recommendation based on results
        
        HINTS:
        - Use scipy.stats.ttest_ind for t-test (or implement manually)
        - Cohen's d = (mean_a - mean_b) / pooled_std
        - CI = difference ± (critical_value * standard_error)
        """
        ### BEGIN SOLUTION
        import math
        
        # Basic statistics
        mean_a = statistics.mean(results_a)
        mean_b = statistics.mean(results_b)
        std_a = statistics.stdev(results_a)
        std_b = statistics.stdev(results_b)
        n_a = len(results_a)
        n_b = len(results_b)
        
        # Two-sample t-test (simplified)
        pooled_std = math.sqrt(((n_a - 1) * std_a**2 + (n_b - 1) * std_b**2) / (n_a + n_b - 2))
        standard_error = pooled_std * math.sqrt(1/n_a + 1/n_b)
        
        if standard_error == 0:
            t_stat = 0
            p_value = 1.0
        else:
            t_stat = (mean_a - mean_b) / standard_error
            # Simplified p-value calculation (assuming normal distribution)
            p_value = 2 * (1 - abs(t_stat) / (abs(t_stat) + math.sqrt(n_a + n_b - 2)))
        
        # Effect size (Cohen's d)
        effect_size = (mean_a - mean_b) / pooled_std if pooled_std > 0 else 0
        
        # Confidence interval for difference
        difference = mean_a - mean_b
        critical_value = 1.96  # Approximate for 95% CI
        margin_of_error = critical_value * standard_error
        ci_lower = difference - margin_of_error
        ci_upper = difference + margin_of_error
        
        # Determine significance
        is_significant = p_value < self.alpha
        
        # Generate recommendation
        if is_significant:
            if effect_size > 0.8:
                recommendation = "Large significant difference - strong evidence for improvement"
            elif effect_size > 0.5:
                recommendation = "Medium significant difference - good evidence for improvement"
            else:
                recommendation = "Small significant difference - weak evidence for improvement"
        else:
            recommendation = "No significant difference - insufficient evidence for improvement"
        
        return StatisticalValidation(
            is_significant=is_significant,
            p_value=p_value,
            effect_size=effect_size,
            confidence_interval=(ci_lower, ci_upper),
            recommendation=recommendation
        )
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def validate_benchmark_result(self, result: BenchmarkResult, 
                                 min_samples: int = 100) -> StatisticalValidation:
        """
        Validate that a benchmark result has sufficient statistical power.
        
        TODO: Implement validation for single benchmark result.
        
        STEP-BY-STEP:
        1. Check if we have enough samples
        2. Calculate confidence interval for the metric
        3. Check for common pitfalls (outliers, etc.)
        4. Provide recommendations
        """
        ### BEGIN SOLUTION
        latencies = result.latencies
        n = len(latencies)
        
        if n < min_samples:
            return StatisticalValidation(
                is_significant=False,
                p_value=1.0,
                effect_size=0.0,
                confidence_interval=(0.0, 0.0),
                recommendation=f"Insufficient samples: {n} < {min_samples}. Need more data."
            )
        
        # Calculate confidence interval for mean latency
        mean_latency = statistics.mean(latencies)
        std_latency = statistics.stdev(latencies)
        standard_error = std_latency / math.sqrt(n)
        
        critical_value = 1.96  # 95% CI
        margin_of_error = critical_value * standard_error
        ci_lower = mean_latency - margin_of_error
        ci_upper = mean_latency + margin_of_error
        
        # Check for outliers (simple check)
        q1 = latencies[int(0.25 * n)]
        q3 = latencies[int(0.75 * n)]
        iqr = q3 - q1
        outlier_threshold = q3 + 1.5 * iqr
        outliers = [l for l in latencies if l > outlier_threshold]
        
        if len(outliers) > 0.1 * n:  # More than 10% outliers
            recommendation = f"Warning: {len(outliers)} outliers detected. Results may be unreliable."
        else:
            recommendation = "Benchmark result appears statistically valid."
        
        return StatisticalValidation(
            is_significant=True,
            p_value=0.0,  # Not applicable for single result
            effect_size=std_latency / mean_latency,  # Coefficient of variation
            confidence_interval=(ci_lower, ci_upper),
            recommendation=recommendation
        )
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")

🧪 Unit Test: Statistical Validation

Let's test our statistical validation with simulated data.

In [ ]:

def test_statistical_validation():
    """Test statistical validation functionality."""
    print("🔬 Unit Test: Statistical Validation...")
    
    validator = StatisticalValidator(confidence_level=0.95)
    
    # Test 1: No significant difference
    results_a = [0.1 + 0.01 * np.random.randn() for _ in range(100)]
    results_b = [0.1 + 0.01 * np.random.randn() for _ in range(100)]
    
    validation = validator.validate_comparison(results_a, results_b)
    print(f"✅ No difference test: significant={validation.is_significant}, p={validation.p_value:.4f}")
    
    # Test 2: Clear significant difference
    results_a = [0.1 + 0.01 * np.random.randn() for _ in range(100)]
    results_b = [0.2 + 0.01 * np.random.randn() for _ in range(100)]
    
    validation = validator.validate_comparison(results_a, results_b)
    print(f"✅ Clear difference test: significant={validation.is_significant}, p={validation.p_value:.4f}")
    print(f"    Effect size: {validation.effect_size:.3f}")
    print(f"    Recommendation: {validation.recommendation}")
    
    # Test 3: Single result validation
    mock_result = BenchmarkResult(
        scenario=BenchmarkScenario.SINGLE_STREAM,
        latencies=[0.1 + 0.01 * np.random.randn() for _ in range(200)],
        throughput=1000,
        accuracy=0.95
    )
    
    validation = validator.validate_benchmark_result(mock_result)
    print(f"✅ Single result validation: {validation.recommendation}")
    print(f"    Confidence interval: ({validation.confidence_interval[0]:.4f}, {validation.confidence_interval[1]:.4f})")
    
    print("✅ Statistical validation tests passed!")

# Run the test
test_statistical_validation()

Step 4: The TinyTorchPerf Framework - Putting It All Together

The Complete MLPerf-Inspired Framework

Now we combine all components into a professional benchmarking framework.

In [ ]:

#| export
class TinyTorchPerf:
    """
    Complete MLPerf-inspired benchmarking framework for TinyTorch.
    
    TODO: Implement the complete benchmarking framework.
    
    UNDERSTANDING THE FRAMEWORK:
    1. Combines all benchmark scenarios
    2. Integrates statistical validation
    3. Provides easy-to-use API
    4. Generates professional reports
    
    IMPLEMENTATION APPROACH:
    1. Initialize with model and dataset
    2. Provide methods for each scenario
    3. Include statistical validation
    4. Generate comprehensive reports
    """
    
    def __init__(self):
        self.scenarios = BenchmarkScenarios()
        self.validator = StatisticalValidator()
        self.model = None
        self.dataset = None
        self.results = {}
    
    def set_model(self, model: Callable):
        """Set the model to benchmark."""
        self.model = model
    
    def set_dataset(self, dataset: List):
        """Set the dataset for benchmarking."""
        self.dataset = dataset
    
    def run_single_stream(self, num_queries: int = 1000) -> BenchmarkResult:
        """
        Run single-stream benchmark.
        
        TODO: Implement single-stream benchmark with validation.
        
        STEP-BY-STEP:
        1. Check that model and dataset are set
        2. Run single-stream scenario
        3. Validate results statistically
        4. Store results
        5. Return result
        """
        ### BEGIN SOLUTION
        if self.model is None or self.dataset is None:
            raise ValueError("Model and dataset must be set before running benchmarks")
        
        result = self.scenarios.single_stream(self.model, self.dataset, num_queries)
        validation = self.validator.validate_benchmark_result(result)
        
        self.results['single_stream'] = {
            'result': result,
            'validation': validation
        }
        
        return result
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def run_server(self, target_qps: float = 10.0, duration: float = 60.0) -> BenchmarkResult:
        """
        Run server benchmark.
        
        TODO: Implement server benchmark with validation.
        """
        ### BEGIN SOLUTION
        if self.model is None or self.dataset is None:
            raise ValueError("Model and dataset must be set before running benchmarks")
        
        result = self.scenarios.server(self.model, self.dataset, target_qps, duration)
        validation = self.validator.validate_benchmark_result(result)
        
        self.results['server'] = {
            'result': result,
            'validation': validation
        }
        
        return result
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def run_offline(self, batch_size: int = 32) -> BenchmarkResult:
        """
        Run offline benchmark.
        
        TODO: Implement offline benchmark with validation.
        """
        ### BEGIN SOLUTION
        if self.model is None or self.dataset is None:
            raise ValueError("Model and dataset must be set before running benchmarks")
        
        result = self.scenarios.offline(self.model, self.dataset, batch_size)
        validation = self.validator.validate_benchmark_result(result)
        
        self.results['offline'] = {
            'result': result,
            'validation': validation
        }
        
        return result
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def run_all_scenarios(self, quick_test: bool = False) -> Dict[str, BenchmarkResult]:
        """
        Run all benchmark scenarios.
        
        TODO: Implement comprehensive benchmarking.
        """
        ### BEGIN SOLUTION
        if quick_test:
            # Quick test with smaller parameters
            single_result = self.run_single_stream(num_queries=100)
            server_result = self.run_server(target_qps=5.0, duration=10.0)
            offline_result = self.run_offline(batch_size=16)
        else:
            # Full benchmarking
            single_result = self.run_single_stream(num_queries=1000)
            server_result = self.run_server(target_qps=10.0, duration=60.0)
            offline_result = self.run_offline(batch_size=32)
        
        return {
            'single_stream': single_result,
            'server': server_result,
            'offline': offline_result
        }
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def compare_models(self, model_a: Callable, model_b: Callable, 
                      scenario: str = 'single_stream') -> StatisticalValidation:
        """
        Compare two models statistically.
        
        TODO: Implement model comparison.
        """
        ### BEGIN SOLUTION
        # Run both models on the same scenario
        self.set_model(model_a)
        if scenario == 'single_stream':
            result_a = self.run_single_stream(num_queries=100)
        elif scenario == 'server':
            result_a = self.run_server(target_qps=5.0, duration=10.0)
        else:  # offline
            result_a = self.run_offline(batch_size=16)
        
        self.set_model(model_b)
        if scenario == 'single_stream':
            result_b = self.run_single_stream(num_queries=100)
        elif scenario == 'server':
            result_b = self.run_server(target_qps=5.0, duration=10.0)
        else:  # offline
            result_b = self.run_offline(batch_size=16)
        
        # Compare latencies
        return self.validator.validate_comparison(result_a.latencies, result_b.latencies)
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def generate_report(self) -> str:
        """
        Generate a comprehensive benchmark report.
        
        TODO: Implement professional report generation.
        """
        ### BEGIN SOLUTION
        report = "# TinyTorch Benchmark Report\n\n"
        
        for scenario_name, scenario_data in self.results.items():
            result = scenario_data['result']
            validation = scenario_data['validation']
            
            report += f"## {scenario_name.replace('_', ' ').title()} Scenario\n\n"
            report += f"- **Throughput**: {result.throughput:.2f} samples/second\n"
            report += f"- **Mean Latency**: {statistics.mean(result.latencies)*1000:.2f} ms\n"
            report += f"- **90th Percentile**: {result.latencies[int(0.9*len(result.latencies))]*1000:.2f} ms\n"
            report += f"- **95th Percentile**: {result.latencies[int(0.95*len(result.latencies))]*1000:.2f} ms\n"
            report += f"- **Statistical Validation**: {validation.recommendation}\n\n"
        
        return report
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")

🧪 Unit Test: TinyTorchPerf Framework

Let's test our complete benchmarking framework.

In [ ]:

def test_tinytorch_perf():
    """Test the complete TinyTorchPerf framework."""
    print("🔬 Unit Test: TinyTorchPerf Framework...")
    
    # Create test model and dataset
    def test_model(sample):
        time.sleep(0.001)  # Simulate processing
        return {"prediction": np.random.rand(5)}
    
    test_dataset = [{"data": np.random.rand(10)} for _ in range(50)]
    
    # Test the framework
    benchmark = TinyTorchPerf()
    benchmark.set_model(test_model)
    benchmark.set_dataset(test_dataset)
    
    # Test individual scenarios
    single_result = benchmark.run_single_stream(num_queries=20)
    assert single_result.scenario == BenchmarkScenario.SINGLE_STREAM
    print(f"✅ Single-stream: {single_result.throughput:.2f} samples/sec")
    
    server_result = benchmark.run_server(target_qps=5.0, duration=2.0)
    assert server_result.scenario == BenchmarkScenario.SERVER
    print(f"✅ Server: {server_result.throughput:.2f} QPS")
    
    offline_result = benchmark.run_offline(batch_size=10)
    assert offline_result.scenario == BenchmarkScenario.OFFLINE
    print(f"✅ Offline: {offline_result.throughput:.2f} samples/sec")
    
    # Test comprehensive benchmarking
    all_results = benchmark.run_all_scenarios(quick_test=True)
    assert len(all_results) == 3
    print(f"✅ All scenarios: {list(all_results.keys())}")
    
    # Test model comparison
    def slower_model(sample):
        time.sleep(0.002)  # Twice as slow
        return {"prediction": np.random.rand(5)}
    
    comparison = benchmark.compare_models(test_model, slower_model)
    print(f"✅ Model comparison: {comparison.recommendation}")
    
    # Test report generation
    report = benchmark.generate_report()
    assert "TinyTorch Benchmark Report" in report
    print("✅ Report generation working")
    
    print("✅ Complete TinyTorchPerf framework working!")

# Run the test
test_tinytorch_perf()

Step 5: Professional Reporting - Project-Ready Results

Why Professional Reports Matter

Your ML projects need:

• Clear performance metrics for presentations
• Statistical validation for credibility
• Comparison baselines for context
• Professional formatting for academic/industry standards

In [ ]:

#| export
class PerformanceReporter:
    """
    Generates professional performance reports for ML projects.
    
    TODO: Implement professional report generation.
    
    UNDERSTANDING PROFESSIONAL REPORTS:
    1. Executive summary with key metrics
    2. Detailed methodology section
    3. Statistical validation results
    4. Comparison with baselines
    5. Recommendations for improvement
    """
    
    def __init__(self):
        self.reports = []
    
    def generate_project_report(self, benchmark_results: Dict[str, BenchmarkResult], 
                               model_name: str = "TinyTorch Model") -> str:
        """
        Generate a professional performance report for ML projects.
        
        TODO: Implement project report generation.
        
        STEP-BY-STEP:
        1. Create executive summary
        2. Add methodology section
        3. Present detailed results
        4. Include statistical validation
        5. Add recommendations
        """
        ### BEGIN SOLUTION
        report = f"""# {model_name} Performance Report

## Executive Summary

This report presents comprehensive performance benchmarking results for {model_name} using MLPerf-inspired methodology. The evaluation covers three standard scenarios: single-stream (latency), server (throughput), and offline (batch processing).

### Key Findings
"""
        
        # Add key metrics
        for scenario_name, result in benchmark_results.items():
            mean_latency = statistics.mean(result.latencies) * 1000
            p90_latency = result.latencies[int(0.9 * len(result.latencies))] * 1000
            
            report += f"- **{scenario_name.replace('_', ' ').title()}**: {result.throughput:.2f} samples/sec, "
            report += f"{mean_latency:.2f}ms mean latency, {p90_latency:.2f}ms 90th percentile\n"
        
        report += """
## Methodology

### Benchmark Framework
- **Architecture**: MLPerf-inspired four-component system
- **Scenarios**: Single-stream, server, and offline evaluation
- **Statistical Validation**: Multiple runs with confidence intervals
- **Metrics**: Latency distribution, throughput, accuracy

### Test Environment
- **Hardware**: Standard development machine
- **Software**: TinyTorch framework
- **Dataset**: Standardized evaluation dataset
- **Validation**: Statistical significance testing

## Detailed Results

"""
        
        # Add detailed results for each scenario
        for scenario_name, result in benchmark_results.items():
            report += f"### {scenario_name.replace('_', ' ').title()} Scenario\n\n"
            
            latencies_ms = [l * 1000 for l in result.latencies]
            
            report += f"- **Sample Count**: {len(result.latencies)}\n"
            report += f"- **Mean Latency**: {statistics.mean(latencies_ms):.2f} ms\n"
            report += f"- **Median Latency**: {statistics.median(latencies_ms):.2f} ms\n"
            report += f"- **90th Percentile**: {latencies_ms[int(0.9 * len(latencies_ms))]:.2f} ms\n"
            report += f"- **95th Percentile**: {latencies_ms[int(0.95 * len(latencies_ms))]:.2f} ms\n"
            report += f"- **Standard Deviation**: {statistics.stdev(latencies_ms):.2f} ms\n"
            report += f"- **Throughput**: {result.throughput:.2f} samples/second\n"
            
            if result.accuracy > 0:
                report += f"- **Accuracy**: {result.accuracy:.4f}\n"
            
            report += "\n"
        
        report += """## Statistical Validation

All results include proper statistical validation:
- Multiple independent runs for reliability
- Confidence intervals for key metrics
- Outlier detection and handling
- Significance testing for comparisons

## Recommendations

Based on the benchmark results:
1. **Performance Characteristics**: Model shows consistent performance across scenarios
2. **Optimization Opportunities**: Focus on reducing tail latency for production deployment
3. **Scalability**: Server scenario results indicate good potential for production scaling
4. **Further Testing**: Consider testing with larger datasets and different hardware configurations

## Conclusion

This comprehensive benchmarking demonstrates {model_name}'s performance characteristics using industry-standard methodology. The results provide a solid foundation for production deployment decisions and further optimization efforts.
"""
        
        return report
        ### END SOLUTION
        raise NotImplementedError("Student implementation required")
    
    def save_report(self, report: str, filename: str = "benchmark_report.md"):
        """Save report to file."""
        with open(filename, 'w') as f:
            f.write(report)
        print(f"📄 Report saved to {filename}")

🧪 Unit Test: Performance Reporter

Let's test our professional reporting system.

In [ ]:

def test_performance_reporter():
    """Test the performance reporter."""
    print("🔬 Unit Test: Performance Reporter...")
    
    # Create mock benchmark results
    mock_results = {
        'single_stream': BenchmarkResult(
            scenario=BenchmarkScenario.SINGLE_STREAM,
            latencies=[0.01 + 0.002 * np.random.randn() for _ in range(100)],
            throughput=95.0,
            accuracy=0.942
        ),
        'server': BenchmarkResult(
            scenario=BenchmarkScenario.SERVER,
            latencies=[0.012 + 0.003 * np.random.randn() for _ in range(150)],
            throughput=87.0,
            accuracy=0.938
        ),
        'offline': BenchmarkResult(
            scenario=BenchmarkScenario.OFFLINE,
            latencies=[0.008 + 0.001 * np.random.randn() for _ in range(50)],
            throughput=120.0,
            accuracy=0.945
        )
    }
    
    # Test report generation
    reporter = PerformanceReporter()
    report = reporter.generate_project_report(mock_results, "My Project Model")
    
    # Verify report content
    assert "Performance Report" in report
    assert "Executive Summary" in report
    assert "Methodology" in report
    assert "Detailed Results" in report
    assert "Statistical Validation" in report
    assert "Recommendations" in report
    
    print("✅ Report generated successfully")
    print(f"✅ Report length: {len(report)} characters")
    print(f"✅ Contains all required sections")
    
    # Test saving
    reporter.save_report(report, "test_report.md")
    print("✅ Report saving working")
    
    print("✅ Performance reporter tests passed!")

# Run the test
test_performance_reporter()

Comprehensive Integration Test

Let's test everything together with a realistic TinyTorch model.

In [ ]:

def test_comprehensive_benchmarking():
    """Test the complete benchmarking system with a realistic model."""
    print("🔬 Comprehensive Integration Test...")
    
    # Create a realistic TinyTorch model
    def create_simple_model():
        """Create a simple classification model for testing."""
        def model(sample):
            # Simulate a simple neural network
            x = np.array(sample['data'])
            
            # Layer 1: 10 -> 5
            W1 = np.random.randn(10, 5) * 0.1
            b1 = np.zeros(5)
            h1 = np.maximum(0, x @ W1 + b1)  # ReLU
            
            # Layer 2: 5 -> 3
            W2 = np.random.randn(5, 3) * 0.1
            b2 = np.zeros(3)
            output = h1 @ W2 + b2
            
            # Simulate some processing time
            time.sleep(0.001)
            
            return {"prediction": output}
        
        return model
    
    # Create test dataset
    test_dataset = []
    for i in range(100):
        sample = {
            'data': np.random.randn(10),
            'target': np.random.randint(0, 3)
        }
        test_dataset.append(sample)
    
    # Test complete workflow
    model = create_simple_model()
    
    # 1. Run comprehensive benchmarking
    benchmark = TinyTorchPerf()
    benchmark.set_model(model)
    benchmark.set_dataset(test_dataset)
    
    print("📊 Running comprehensive benchmarking...")
    all_results = benchmark.run_all_scenarios(quick_test=True)
    
    # 2. Generate professional report
    reporter = PerformanceReporter()
    report = reporter.generate_project_report(all_results, "TinyTorch CNN Model")
    
    # 3. Validate results
    for scenario_name, result in all_results.items():
        assert result.throughput > 0, f"{scenario_name} should have positive throughput"
        assert len(result.latencies) > 0, f"{scenario_name} should have latency measurements"
        print(f"✅ {scenario_name}: {result.throughput:.2f} samples/sec")
    
    # 4. Test model comparison
    def create_slower_model():
        """Create a slower model for comparison."""
        def model(sample):
            x = np.array(sample['data'])
            W1 = np.random.randn(10, 5) * 0.1
            b1 = np.zeros(5)
            h1 = np.maximum(0, x @ W1 + b1)
            
            W2 = np.random.randn(5, 3) * 0.1
            b2 = np.zeros(3)
            output = h1 @ W2 + b2
            
            time.sleep(0.002)  # Slower
            return {"prediction": output}
        
        return model
    
    slower_model = create_slower_model()
    comparison = benchmark.compare_models(model, slower_model)
    print(f"✅ Model comparison: {comparison.recommendation}")
    
    # 5. Test report quality
    assert len(report) > 1000, "Report should be comprehensive"
    print(f"✅ Generated {len(report)} character report")
    
    print("✅ Comprehensive integration test passed!")
    print("🎉 Complete benchmarking system working!")

# Run the comprehensive test
test_comprehensive_benchmarking()

🧪 Module Testing

Time to test your implementation! This section uses TinyTorch's standardized testing framework to ensure your implementation works correctly.

This testing section is locked - it provides consistent feedback across all modules and cannot be modified.

In [ ]:

# =============================================================================
# STANDARDIZED MODULE TESTING - DO NOT MODIFY
# This cell is locked to ensure consistent testing across all TinyTorch modules
# =============================================================================

if __name__ == "__main__":
    from tito.tools.testing import run_module_tests_auto
    
    # Automatically discover and run all tests in this module
    success = run_module_tests_auto("Benchmarking")

🎯 Module Summary: Systematic ML Performance Evaluation

What You've Built

You've implemented a comprehensive MLPerf-inspired benchmarking framework:

1. Benchmark Scenarios: Single-stream (latency), server (throughput), and offline (batch processing)
2. Statistical Validation: Confidence intervals, significance testing, and effect size calculation
3. MLPerf Architecture: Four-component system with load generator, model, dataset, and evaluation
4. Professional Reporting: Generate conference-quality performance reports with proper methodology
5. Model Comparison: Systematic comparison framework with statistical validation

Key Insights

• Systematic evaluation beats intuition: Proper benchmarking reveals true performance characteristics
• Statistics matter: Single measurements are meaningless; confidence intervals provide real insights
• Scenarios capture reality: Different use cases (mobile, server, batch) require different metrics
• Reproducibility is crucial: Others must be able to verify your results
• Professional presentation: Clear methodology and statistical validation build credibility

Real-World Connections

• MLPerf: Uses identical four-component architecture and scenario patterns
• Production systems: A/B testing frameworks follow these statistical principles
• Research papers: Proper experimental methodology is required for publication
• ML engineering: Systematic evaluation prevents costly production mistakes
• Open source: Contributing benchmarks to libraries like PyTorch and TensorFlow

Next Steps

In real ML systems, you'd:

1. GPU benchmarking: Extend to CUDA/OpenCL performance measurement
2. Distributed evaluation: Scale benchmarking across multiple machines
3. Continuous monitoring: Integrate with CI/CD pipelines for regression detection
4. Domain-specific metrics: Develop specialized benchmarks for your problem domain
5. Hardware optimization: Evaluate performance across different architectures

🏆 Achievement Unlocked

You've mastered systematic ML evaluation using industry-standard methodology. You understand how to design proper experiments, validate results statistically, and present findings professionally!

You've completed the TinyTorch Benchmarking module! 🎉