HPC Guru (Twitter)
RT @thoefler: Don't forget to submit to our #TinyML session! Still three weeks left until the Jan 31st deadline. We're open to a wide range of topics from hardware to algorithms and sparsity in #DeepLearning and other #MachineLearning tasks on small devices and the edge.
RT @thoefler: Don't forget to submit to our #TinyML session! Still three weeks left until the Jan 31st deadline. We're open to a wide range of topics from hardware to algorithms and sparsity in #DeepLearning and other #MachineLearning tasks on small devices and the edge.
HPC Guru (Twitter)
What Is #TinyML?
The need for #HPC has confined many ML applications to the #Cloud
We must find ways to facilitate ML inference on smaller, more resource-constrained devices, typically found at the Edge
https://www.allaboutcircuits.com/technical-articles/what-is-tinyml/
What Is #TinyML?
The need for #HPC has confined many ML applications to the #Cloud
We must find ways to facilitate ML inference on smaller, more resource-constrained devices, typically found at the Edge
https://www.allaboutcircuits.com/technical-articles/what-is-tinyml/
HPC Guru (Twitter)
RT @thoefler: 2.5 bits per weight are enough to run inference on extremely large language models! Using a simple blocked scaling scheme and moderate block sizes.
Meet GPTQ: https://arxiv.org/abs/2210.17323
#DeepLearning #TinyML @DAlistarh @elias_frantar @AshkboosSaleh https://twitter.com/thoefler/status/1599653546483408896/photo/1
RT @thoefler: 2.5 bits per weight are enough to run inference on extremely large language models! Using a simple blocked scaling scheme and moderate block sizes.
Meet GPTQ: https://arxiv.org/abs/2210.17323
#DeepLearning #TinyML @DAlistarh @elias_frantar @AshkboosSaleh https://twitter.com/thoefler/status/1599653546483408896/photo/1