Add tensor-cache.json to repo root for WebLLM

Add MLC/WebGPU compiled model for WebLLM

initial commit