Get to Know PyTorch - A Cutting Edge Open Source AI Framework for Deep Learning
summary
Overview of PyTorch fundamentals, current Foundation status and practical tooling: PyTorch moves under the Linux Foundation (PyTorch Foundation, Sept 2022) with 3,000+ contributors, provides autograd for automatic gradients, data loaders and torch.nn layers for building models, and supports CPU, NVIDIA CUDA, AMD ROCm and Apple MPS. Shows torch.compile (Dynamo+Inductor) yielding ~30%+ speedups and distributed training options from simple DataParallel to DDP and model-sharding with FSDP — demos report successful runs on 64 nodes / 500+ GPUs and standard checkpointing for resuming long trainings.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:02Deep learning basics
- 8:26PyTorch framework
- 13:22Data prep and model building
- 18:29Training loop and validation
- 24:30Distributed training
- 28:06Compiler and performance
- 31:59GPU support and compatibility
- 32:42Contributing to PyTorch
- 36:13PyTorch roadmap
- 37:37PyTorch in Watson X
- 38:36Q&A
related talks
19mGet to know PyTorch - a cutting edge open source AI framework for Machine Learning
12mKeynote: How PyTorch Became the Foundation of the AI Revolution
19mIncreasing the Impact of AI Through Better Software (Pytorch Updates)
43mPyTorch Foundation: 1 Year of Open Governance
43mPyTorch Technical Deep Dive
09mOpening Remarks
48mFireside Chat
12mLightning Talk: State of PyTorch