AI product Open source
PyTorch is a Python machine-learning library for tensor computation and deep neural networks, with CPU and GPU execution. Its tensor library provides NumPy-like operations, while its tape-based autograd system uses reverse-mode automatic differentiation for differentiable tensor operations and dynamically defined networks. The project also includes neural-network, compilation, multiprocessing, and data-loading components, and supports extensions through Python packages such as NumPy, SciPy, and Cython.
8 uses taken from transcripts — each links to the moment in the video.
Provides the execution framework discussed in relation to kernel ordering, overlapping execution, and race conditions.
Provides the standard TopK baseline replaced by DeepSelect.
Example of a machine-learning library or AI package that uses Python underneath.
An open-source Python package for tensor computation and deep learning. It provides CPU- and GPU-based tensors, automatic differentiation, and neural-network tools for research and production.
Runs the distributed training script whose job is restarted and resumed from a saved checkpoint after the GPU failure.
Used as the basis for a simple model-inference implementation and as part of model execution.
The deep learning framework whose compiled workloads the video discusses, covering how Python-level code like .item() calls interact with compiler optimization and accelerator execution.
Framework whose models can be sped up with one-line compilation; provides the compiled wrapper around the model
8 in the library.