WebLLM via dianome/webllm

← demo · artifact · WebLLM model id Qwen2.5-0.5B-Instruct-q4f16_1-MLC. The adapter pre-warms WebLLM's webllm/model and webllm/config caches; WebLLM then loads with no shard fetches from the model URL (the wasm model library still comes from WebLLM's CDN). Needs WebGPU.

1. Pre-warm from Dianome2. Load engine and chatClear caches