A small, serialized wrapper over LiteRT-LM: one shared engine, GPU→CPU fallback ladder, retry-aware sampling, streamed and one-shot generation.

Artifacts using Stepgrid LLM Kit :: Engine (1)
Sort by:Popular▼

Everything in stepgrid-llm-kit in one dependency: the LiteRT-LM engine wrapper, device tiering and resumable-download decisions.
Last Release on Nov 20, 2026
  • Prev
  • 1
  • Next