Mila Deep Neural Net library
Mila, version 0.9.61-alpha is currently an early, experimental, preview release.
Achilles Mila Deep Neural Network library provides an API to model, train and evaluate Deep Neural Networks. Mila utilizes the NVIDIA CUDA runtime for high-performance GPU acceleration.
1 Deep Neural Nets
- GPT2, Recurrent Neural Networks
- GPU acceleration using CUDA runtime
2 Datasets
- Batch sequence loader
Mila, Version 0.9.61-alpha.1 changes: Refactor GPT-2 components and update device handling
- Updated CMakeLists.txt to include new source files for Gpt2App and removed CudaSoftmaxOp.
- Moved Gpt2DataLoader class to a new file and removed content from DataLoader.ixx.
- Changed module exports in Gpt2Model and Tokenizer to Gpt2App namespace.
- Created ModelConfig.ixx for model configuration parameters.
- Restructured TrainGpt2.cpp for command-line parsing and model initialization.
- Introduced OperationProperties.ixx for common operation properties.
- Updated README.md for version 0.9.61-alpha and recent changes.
- Replaced Compute::CpuDevice and Compute::CudaDevice with Compute::DeviceType for better device management.
- Updated tests in multiple files to ensure compatibility with the new DeviceType system.
Mila uses CMake build. To build Mila, follow the steps below:
- Clone the Mila repository
- Open the Mila directory using Visual Studio Code or any other IDE
- C++ 20 Module API
- NVIDIA Cuda Runtime, 12.8
- CMake 3.31