Meta introduces LLAMA 2 LONG

#9: MIT Course on TinyML & Efficient DL Computing, Graph Neural Prompting, List of GenAI startups in India

Meta introduces LLAMA 2 LONG

• Supports context length of upto 32,768 tokens

• Instruction tuned model outperforms GPT3.5 Turbo on a suite of long-context tasks

• Read the paper here. To know more about ZeroSCROLLS dataset check here and here

From “Effective Long-Context Scaling of Foundation Models” paper

MIT’s Course on TinyML & Efficient DL Computing

• The course introduces efficient computing techniques to enable powerful DL models on resource-constrained devices

• Topics include model compression, pruning, quantization, neural architecture search, distributed training, data/model parallelism, gradient compression, and on-device fine-tuning

• Course page andYouTube Playlist

Image from YouTube Playlist

Paper: Graph Neural Prompting with LLMs

• The paper proposes Graph Neural Prompting (GNP), a novel plug-and-play method to assist pre-trained LLMs in learning beneficial knowledge from Knowledge Graphs

• GNP encompasses various designs, including a standard graph neural network encoder, a cross modality pooling module, a domain projector, and a self-supervised link prediction objective.

• Read the paper here

Image from the paper

GenAI Startups in India

• Matrix Partners India released the list of GenAI startups in India

• Check it out here

  

  Image from Matrix Partners website