Google Cloud Tech: AI/ML frameworks for cloud TPUs

Author: Google Cloud Tech - Bewertung: 0x - Views:7

Explore the diverse landscape of AI/ML frameworks, including, PyTorch, JAX, and Keras. This video breaks down what these frameworks are, how they differ, and when to choose one over another for model training, inference, and fine tuning. We cover additional specific purpose frameworks like Nvidia's Triton Inference Server and Hugging Face's TGI and PEFT, providing insights into their strengths for building and deploying efficient AI models.



Chapters:

0:00 - Introduction to AI/ML frameworks

0:34 - Defining your models: The builders

0:57 - Keras: The user friendly API

1:30 - PyTorch and JAX: Granular control and performance

2:23 - Inference: putting models to work

3:38 - vLLM for LLM serving

4:13 - Fine tuning: The learning process

4:58 - Conclusion and recap





Resources:

Keras Overview → https://goo.gle/3JOiCCX

PyTorch Overview → https://goo.gle/3LyarLE

Get Started with JAX → https://goo.gle/3LUjlTK

Get Started with vLLM → https://goo.gle/47xuYbQ

Hugging Face PEFT Documentation → https://goo.gle/49MZyzz

NVIDIA Triton Inference Server → https://goo.gle/4qTBv8f





Subscribe to Google Cloud Tech → https://goo.gle/GoogleCloudTech



#GoogleCloud #AIFrameworks



Speakers: Duncan Campbell

Products Mentioned: AI Infrastructure, JAX, PyTorch, Keras, Cloud TPUs