Microsoft has released a new beta version of Microsoft Cognitive Toolkit (link), and soon the Release Candidate version will arrive.
Note: Thanks to @PabloDoval who point me in the right direction.
We already knew this tools by CNTK (something knew before but not the official mode). This set of tools is not for the typical Microsoft programmer, this Toolkit is designed for data scientists.
The best way to start is the homepage, GitHub or the Wiki. The samples Gallery is quite comprehensive and we can also reuse and modify these examples to adapt them to our needs. The best thing to understand what we can find is this video
And, of course, the copy & paste of the CNTK features.
Highly optimized, built-in components
- Components can handle multi-dimensional dense or sparse data from Python, C++ or BrainScript
- FFN, CNN, RNN/LSTM, Batch normalization, Sequence-to-Sequence with attention and more
- Reinforcement learning, generative adversarial networks, supervised and unsupervised learning
- Ability to add new user-defined core-components on the GPU from Python
- Automatic hyperparameter tuning
- Built-in readers optimized for massive datasets
Efficient resource usage
- Parallelism with accuracy on multiple GPUs/machines via 1-bit SGD and Block Momentum
- Memory sharing and other built-in methods to fit even the largest models in GPU memory
Easily express your own networks
- Full APIs for defining networks, learners, readers, training and evaluation from Python, C++ and BrainScript
- Evaluate models with Python, C++, C# and BrainScript
- Interoperation with NumPy
- Both high-level and low-level APIs available for ease of use and flexibility
- Automatic shape inference based on your data
- Fully optimized symbolic RNN loops (no unrolling needed)
Training and hosting with Azure
- Takes advantage of high-speed resources when used with Azure GPU and Azure networks
- Host trained models easily on Azure and add real-time training if desired
Greetings @ Toronto