DataLoader.ixx File Reference

Device-agnostic data loader interface using abstract tensor data types. More...

#include <type_traits>
#include <cstddef>
#include <memory>
#include <stdexcept>
#include <cstdint>
#include <string>
#include <sstream>
import Compute.CpuMemoryResource;
import Compute.MemoryResource;
import Compute.DeviceType;
import Dnn.TensorDataTypeTraits;
import Dnn.TensorDataType;
import Dnn.Tensor;

Classes
class	Mila::Data::DataLoader< TInputDataType, TTargetDataType, TMemoryResource >
	Device-agnostic data loader interface using abstract tensor data types. More...

Namespaces
namespace	Mila
	Mila main API namespace.

Typedefs
template<TensorDataType TInputDataType = TensorDataType::FP32, TensorDataType TTargetDataType = TInputDataType>
using	Mila::Data::CpuDataLoader = DataLoader<TInputDataType, TTargetDataType, CpuMemoryResource>
	CPU data loader with single precision floating point.

Detailed Description

Device-agnostic data loader interface using abstract tensor data types.

This module provides a sophisticated data loading framework for efficiently feeding heterogeneous data into neural network models during training and evaluation processes. Uses abstract TensorDataType enumeration to enable seamless operation across different compute devices (CPU, CUDA, Metal, Rocm ) without exposing device-specific concrete types to host compilation.

Key architectural features:

• Abstract data type system prevents device-specific compilation dependencies
• Support for mixed-precision training with different input/target data types
• Optimized memory resource selection for efficient host-device data transfers
• Type-safe batch operations with compile-time validation
• Extensible design for various data sources and formats

Typedef Documentation

CpuDataLoader

template<TensorDataType TInputDataType = TensorDataType::FP32, TensorDataType TTargetDataType = TInputDataType>

using Mila::Data::CpuDataLoader = DataLoader<TInputDataType, TTargetDataType, CpuMemoryResource>

export

CPU data loader with single precision floating point.

Convenient alias for data loaders using standard CPU memory with FP32 data types for both inputs and targets. Suitable for CPU-only training and development workflows.

Template Parameters

TInputDataType	Input tensor data type (defaults to FP32)
TTargetDataType	Target tensor data type (defaults to input type)