mirror of
https://github.com/MLSysBook/TinyTorch.git
synced 2026-06-05 12:32:30 -05:00
b57f485369
- Add tinytorch.utils.profiler following PyTorch's utils pattern - Includes SimpleProfiler class for educational performance measurement - Provides timing, memory usage, and system metrics - Follows PyTorch's torch.utils.* organizational pattern - Module 11: Kernels uses profiler for performance demonstrations Features: - Wall time and CPU time measurement - Memory usage tracking (peak, delta, percentages) - Array information (shape, size, dtype) - CPU and system metrics - Clean educational interface for ML performance learning Import pattern: from tinytorch.utils.profiler import SimpleProfiler
77 lines
2.2 KiB
YAML
77 lines
2.2 KiB
YAML
# TinyTorch Module Metadata
|
|
# Essential system information for CLI tools and build systems
|
|
|
|
name: "11_kernels"
|
|
title: "Kernels - Hardware-Aware Optimization"
|
|
description: "Custom operations, performance optimization, and hardware-aware computing for ML systems"
|
|
version: "1.0.0"
|
|
author: "TinyTorch Team"
|
|
|
|
# Dependencies - Used by CLI for module ordering and prerequisites
|
|
dependencies:
|
|
prerequisites: [
|
|
"00_setup", "01_tensor", "02_activations", "03_layers",
|
|
"04_networks", "05_cnn", "06_dataloader", "07_autograd",
|
|
"08_optimizers", "09_training", "10_compression"
|
|
]
|
|
enables: ["12_benchmarking", "13_mlops"]
|
|
|
|
# Package Export - What gets built into tinytorch package
|
|
exports_to: "tinytorch.core.kernels"
|
|
|
|
# File Structure - What files exist in this module
|
|
files:
|
|
dev_file: "kernels_dev.py"
|
|
test_file: "tests/test_kernels.py"
|
|
readme: "README.md"
|
|
benchmark_dir: "benchmarks/"
|
|
|
|
# Components - What's implemented in this module
|
|
components:
|
|
# Custom Operations
|
|
- "matmul_custom"
|
|
- "relu_custom"
|
|
- "conv2d_custom"
|
|
|
|
# Optimized Implementations
|
|
- "matmul_vectorized"
|
|
- "matmul_cache_optimized"
|
|
- "matmul_parallel"
|
|
|
|
# Compressed Model Kernels
|
|
- "quantized_matmul"
|
|
- "sparse_matmul"
|
|
- "pruned_conv2d"
|
|
|
|
# Performance Tools
|
|
- "KernelProfiler"
|
|
- "PerformanceBenchmark"
|
|
- "HardwareProfiler"
|
|
|
|
# Learning Objectives - What students will achieve
|
|
learning_objectives:
|
|
- "Implement custom ML operations beyond NumPy"
|
|
- "Apply SIMD vectorization and CPU optimization"
|
|
- "Optimize memory layout and cache efficiency"
|
|
- "Understand GPU-style parallel computing"
|
|
- "Build performance profiling tools"
|
|
- "Create hardware-optimized compressed model operations"
|
|
|
|
# Educational Approach
|
|
pedagogy:
|
|
framework: "Build → Use → Optimize"
|
|
difficulty: "Expert"
|
|
time_estimate: "8-10 hours"
|
|
|
|
# Integration Points - How this connects to other modules
|
|
integration:
|
|
builds_on: "10_compression" # Extends compression with hardware optimization
|
|
enables: "12_benchmarking" # Provides optimized kernels for benchmarking
|
|
connects_to: "13_mlops" # Hardware optimization for production deployment
|
|
|
|
# Testing Strategy
|
|
testing:
|
|
inline_tests: true
|
|
performance_tests: true
|
|
integration_tests: true
|
|
benchmark_tests: true |