RealFeedApp New format runs 11B VLM on Android with 4 GB memory
Researchers from Graphcore and Arm developed a quantized version of Llama 3.2 Vision that runs on Android phones using only 3.7 GB of storage.
- Field
- applying language models
- What they did
- Researchers from Graphcore and Arm created a 3.7 GB version of the Llama 3.2 Vision Instruct model to fit it into a smartphone's memory, as the original model weighed over 21 GB.
- Why it matters
- This enables running powerful computer vision models directly on mobile devices, which is crucial for robotics and wearables with limited power.
#llm#quantization#arm#graphcore#llama 3.2 vision
