[NPU]LoRA: Adding Torch Native backend by vlserov · Pull Request #14132 · sgl-project/sglang

vlserov · 2025-11-29T08:27:22Z

Adding Torch Native backend for support LoRA feature for architecture without Triton

Motivation

Not all platform support Triton. Needed backend for LoRA feature without using Triton

Adding new backend and also fix for Ascend related to base interface changes

Format your code according to the Format code with pre-commit.
Add unit tests according to the Run and add unit tests.
Update documentation according to Write documentations.
Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
Follow the SGLang code style guidance.
Work with maintainers to merge your PR. See the PR Merge Process