TLDRocket
Sign in

Quantization

7 summarised stories about Quantization, each linking back to the original source. Browse all topics →

Wednesday, 20 March 2024

A Chatbot on your Laptop: Phi-2 on Intel Meteor Lake

Hugging Face Blog 2 years ago

Microsoft's Phi-2, a 2.7-billion parameter language model, can now run on standard laptops using Intel's Meteor Lake processor with 4-bit quantization applied through the Optimum Intel library. The quantized model achieved adequate generation speed on a mid-range Core Ultra 7 155H laptop while maintaining high output quality for tasks like physics explanations and code generation. Local inference eliminates the need for cloud API calls, reducing latency and costs while enabling offline work and data privacy.