Deferred GPU-resident sampling and pre-allocated decode tensors#779

mitiskuma wants to merge 1 commit intomlc-ai:mainmlc-ai/web-llm:mainfrom

mitiskuma:Deferred-GPU-resident-sampling-and-pre-allocated-decode-tensorsmitiskuma/web-llm:Deferred-GPU-resident-sampling-and-pre-allocated-decode-tensorsCopy head branch name to clipboard