Kernel-free, no Cuda, Compiler-only solution.
The current implementation, running on our LPU Inference Engine uses our own processor. The model is Llama 2, 70B 4k sequence at FP16…We haven't even hit the next gear through those other methods. Plenty in the tank left!
RESEARCH
-
Groq LPU Engine: Kernel-free Llama 2 70B Inference Performance
By
–
-
AI Director Shares Insights on Primary Healthcare Applications
By
–
Soma Dhavala, Director, ML at Wadhwani AI sharing his insights on the Application of AI in Comprehensive Primary Healthcare, leading a speaker session at the conference.
-
Outstanding Master’s Thesis in Artificial Intelligence Research
By
–
best masters thesis ever (so far)
-
Image Classification with Global Context Vision Transformer Tutorial
By
–
New tutorial on http://
keras.io: Image classification using Global Context Vision Transformer. Works with JAX/TF/PyTorch. -
Out of the Box Performance Without Optimization Efforts
By
–
It's representative of "out of the box performance", the speed folks who aren't performance experts get when they develop their own models. We have made zero performance optimizations to our model. Also we have found torch compile() rarely ever works, and when it does it may not
-
Experimental Rigor in AI Research Methods Needs Improvement
By
–
very true! I don't think the experiments are particularly thorough here honestly; I would have also liked to see a comparison to random sampling, at the very least; a downstream evaluation of the synthetic data would have been nice too
-
Ecosystem Graphs Framework Enhances Foundation Model Transparency
By
–
A new framework called Ecosystem Graphs summarizes what we know about foundation models and helps to increase transparency:
-
Keras 3 with JAX Significantly Faster Than PyTorch
By
–
It is not uncommon for Keras 3 + JAX to be 3-5x faster than PyTorch models pulled from research repos on GitHub or from Hugging Face. Though the typical figure is more like 30% faster.
-
Revolutionary Thermal Control Circuits to Cool Electronics
By
–
Scientists Finally Invent Heat-Controlling Circuitry That Keeps Electronics Cool | Scientific American https://scientificamerican.com/article/scientists-finally-invent-heat-controlling-circuitry-that-keeps-electronics-cool/
-
Top 10 Deep Learning Algorithms: CNNs to DBNs Explained
By
–
Unlock the potential of AI with this breakdown of the top 10 deep learning algorithms by @techprofreeteam! From CNNs to DBNs, the future is here. Elevate your tech knowledge by following @ingliguori for the latest in #DeepLearning and #AI! #TechTrends #MachineLearning