PyTorch 2.0 is here! Quick tutorial + NVIDIA RTX 4080 giveaway

Опубликовано: 15 Октябрь 2024
на канале: Daniel Bourke

15,495

444

The official version PyTorch 2.0 just launched! In this video, I break down some of the updates and how you can use them in your PyTorch code.

Sign up for NVIDIA GTC March 2023 - https://nvda.ws/3GUZygQ
Details for giveaway - https://www.mrdbourke.com/pytorch-2

NVIDIA GTC sessions I'm going to:
Compile and train with 43% speedup using PyTorch 2.0 - https://dbourke.link/nvidia-gtc-2023-...
Fireside Chat with Ilya Sutskever and Jensen Huang: AI Today and Vision of the Future - https://dbourke.link/nvidia-gtc-2023-...
Exploring next-generation methods for optimizing PyTorch models for inference with Torch-TensorRT - https://dbourke.link/nvidia-gtc-2023-...
FP8 Mixed-Precision Training with Hugging Face Accelerate - https://dbourke.link/nvidia-gtc-2023-...
Generative AI Demystified - https://dbourke.link/nvidia-gtc-2023-...

PyTorch 2.0 resources:
PyTorch 2.0 technical overview - https://pytorch.org/get-started/pytor...
PyTorch 2.0 release notes - https://pytorch.org/blog/pytorch-2.0-...
PyTorch 2.0 on TIMM/Hugging Face models - https://pytorch.org/blog/Accelerating...
PyTorch 2.0 speedups on Diffusers - https://pytorch.org/blog/accelerated-...
Which GPU for deep learning by Tim Dettmers - https://timdettmers.com/2023/01/30/wh...
Make deep learning models go brrrr by Horace He - https://horace.io/brrr_intro.html
My PyTorch 2.0 code intro - https://github.com/mrdbourke/pytorch-...
My PyTorch 2.0 benchmarks - https://github.com/mrdbourke/pytorch-...

Other links:
Learn AI/ML (beginner-friendly courses I teach) - https://www.mrdbourke.com/ml-courses/
AI/ML courses/books I recommend - https://www.mrdbourke.com/ml-resources/
Read my novel Charlie Walks - https://www.charliewalks.com

Connect elsewhere:
Web - https://dbourke.link/web
Twitter - / mrdbourke
Twitch - / mrdbourke
ArXiv channel (past streams) - https://dbourke.link/archive-channel
Get email updates on my work - https://dbourke.link/newsletter

Timestamps:
0:00 - Intro
0:10 - torch.compile()
0:53 - PyTorch 2.0 speedups
2:01 - PyTorch 2.0 new technologies
2:22 - Operator fusion
4:00 - Graph capture
5:14 - Other notable PyTorch 2.0 features
6:00 - GPU time
11:30 - NVIDIA GTC information and GPU giveaway details

#pytorch #ai