turboloader 1.7.8

Production-ready ML data loading library with distributed training support, SIMD-accelerated transforms, and custom binary format. Built with C++20 for maximum performance, with seamless Python integration for PyTorch, TensorFlow, and JAX workflows.

TurboLoader

Production-Ready ML Data Loading Library

---

Overview

TurboLoader is a high-performance data loading library for machine learning workflows. Built with C++20 and featuring Python bindings, it provides efficient data loading with SIMD-accelerated transforms, custom binary formats, and distributed training support.

Core Features

• Distributed Training Support - Multi-node data loading with deterministic sharding (v1.7.1)
• SIMD-Accelerated Transforms - 19 vectorized transforms using AVX2/AVX-512/NEON
• TBL v2 Binary Format - Custom format with LZ4 compression for reduced storage
• Framework Integration - Seamless support for PyTorch, TensorFlow, and JAX
• Memory-Mapped I/O - Zero-copy file access for improved throughput
• Lock-Free Queues - Concurrent data structures for efficient multi-threading
• GPU JPEG Decoding - Optional NVIDIA nvJPEG support for accelerated decoding

---

Installation

From PyPI (Recommended)

pip install turboloader

From Source

git clone https://github.com/ALJainProjects/TurboLoader.git
cd TurboLoader
pip install -e .

System Requirements

• Python: 3.8 or higher
• Compiler: C++20 capable (GCC 10+, Clang 12+, MSVC 19.29+)
• OS: macOS, Linux, Windows

Optional Dependencies

Install for enhanced performance:

# macOS
brew install jpeg-turbo libpng libwebp lz4

# Ubuntu/Debian
sudo apt-get install libjpeg-turbo8-dev libpng-dev libwebp-dev liblz4-dev

---

Quick Start

Basic Usage

import turboloader

# Create DataLoader
loader = turboloader.DataLoader(
    'imagenet.tar',
    batch_size=128,
    num_workers=8
)

# Iterate over batches
for batch in loader:
    for sample in batch:
        image = sample['image']  # NumPy array (H, W, C)
        label = sample['label']
        # Train your model...

With Transforms

import turboloader

# Create transforms
resize = turboloader.Resize(224, 224)
normalize = turboloader.ImageNetNormalize()
flip = turboloader.RandomHorizontalFlip(p=0.5)

# Apply transforms
loader = turboloader.DataLoader('data.tar', batch_size=64, num_workers=8)

for batch in loader:
    for sample in batch:
        img = sample['image']
        img = resize.apply(img)
        img = flip.apply(img)
        img = normalize.apply(img)
        # Ready for training

PyTorch Integration

import turboloader
import torch

loader = turboloader.DataLoader('imagenet.tar', batch_size=64, num_workers=8)

# Convert to PyTorch tensors
to_tensor = turboloader.ToTensor(
    format=turboloader.TensorFormat.PYTORCH_CHW
)

for batch in loader:
    images = []
    for sample in batch:
        img = to_tensor.apply(sample['image'])
        images.append(torch.from_numpy(img))

    batch_tensor = torch.stack(images)
    # Train model...

Distributed Training

import turboloader
import torch.distributed as dist

# Initialize distributed training
dist.init_process_group(backend='nccl')

# Create loader with distributed support
loader = turboloader.DataLoader(
    data_path="/data/imagenet.tar",
    batch_size=64,
    num_workers=4,
    shuffle=True,
    enable_distributed=True,
    world_rank=dist.get_rank(),
    world_size=dist.get_world_size(),
    drop_last=True
)

# Each rank automatically gets its shard
for batch in loader:
    # Your training code
    pass

---

Transform Library

TurboLoader includes 19 SIMD-accelerated transforms:

Core Transforms

• Resize - Bilinear/Bicubic/Lanczos interpolation
• Normalize - Mean/std normalization with SIMD
• CenterCrop - Center region extraction
• RandomCrop - Random crop with padding

Augmentation Transforms

• RandomHorizontalFlip - SIMD horizontal flip
• RandomVerticalFlip - SIMD vertical flip
• ColorJitter - Brightness/contrast/saturation/hue
• RandomRotation - Arbitrary angle rotation
• GaussianBlur - Separable convolution
• RandomErasing - Cutout augmentation
• Pad - Border padding (CONSTANT/EDGE/REFLECT)

Advanced Transforms

• RandomPosterize - Bit-depth reduction
• RandomSolarize - Threshold inversion
• RandomPerspective - Perspective warp
• AutoAugment - Learned policies (ImageNet/CIFAR10/SVHN)

Tensor Conversion

• ToTensor - PyTorch CHW or TensorFlow HWC format

---

TBL v2 Binary Format

TurboLoader includes a custom binary format optimized for ML workloads:

Features

• LZ4 compression for reduced storage
• Memory-mapped access for fast loading
• O(1) random access via indexed structure
• Data integrity validation with CRC checksums
• Cached image dimensions for filtered loading

Convert TAR to TBL

import turboloader

writer = turboloader.TblWriterV2(
    output_path="/data/imagenet.tbl",
    compression=True
)

reader = turboloader.TarReader("/data/imagenet.tar")
for sample in reader:
    writer.add_sample(
        data=sample.data,
        format=sample.format,
        metadata={"label": sample.label}
    )

writer.finalize()

---

Documentation

Getting Started

• Quick Start Notebook - Interactive tutorial for beginners
• Installation Guide - Detailed setup instructions
• Quick Start - Getting started examples
• Troubleshooting Guide - Common issues and solutions

API Documentation

• API Reference - Complete API documentation
• Transforms API - All 19 transforms with examples

Framework Integration

• PyTorch Integration Guide - Complete PyTorch guide
• TensorFlow Integration Guide - Complete TensorFlow/Keras guide
• PyTorch Lightning Example - Production-ready Lightning integration
• Distributed Training (DDP) - Multi-GPU PyTorch DDP example

Examples

• ImageNet ResNet50 Training - Complete training pipeline with AMP, checkpointing, TensorBoard
• Distributed Training - Multi-node setup guide

---

Architecture

TurboLoader uses a multi-threaded pipeline architecture:

┌─────────────────────────────────────────────┐
│           Memory-Mapped Reader              │
│     (TAR/TBL v2 with zero-copy access)      │
└──────────────┬──────────────────────────────┘
               │
        ┌──────▼──────┐
        │Worker Pool  │
        │  (N threads)│
        ├─────────────┤
        │ Decode      │
        │ Transform   │
        │ Convert     │
        └──────┬──────┘
               │
        ┌──────▼──────────────┐
        │ Lock-Free Queue     │
        └──────┬──────────────┘
               │
        ┌──────▼──────┐
        │Python API   │
        └─────────────┘

Key Components

• Memory-Mapped I/O - Zero-copy file access
• Worker Thread Pool - Parallel processing with per-thread decoders
• SIMD Transforms - Vectorized operations (AVX2/AVX-512/NEON)
• Lock-Free Queues - High-performance concurrent data structures

---

License

TurboLoader is released under the MIT License.

---

Citation

If you use TurboLoader in your research:

@software{turboloader2025,
  author = {Jain, Arnav},
  title = {TurboLoader: Production-Ready ML Data Loading},
  year = {2025},
  version = {1.7.8},
  url = {https://github.com/ALJainProjects/TurboLoader}
}

---

Support

• Documentation: docs/
• Troubleshooting: docs/TROUBLESHOOTING.md
• Verification Script: Run python scripts/verify_installation.py to check your setup
• Issues: GitHub Issues
• Discussions: GitHub Discussions
• PyPI: https://pypi.org/project/turboloader/

---

TurboLoader v1.7.8 - Production-ready ML data loading. Fast. Efficient. Reliable.