litert-lm: add page (#23600)

Co-authored-by: Wiktor Perskawiec <git@spageektti.cc>

pages/common/litert-lm.md

@@ -0,0 +1,36 @@
+# litert-lm
+
+> Run and manage LiteRT-LM models.
+> More information: <https://developers.google.com/edge/litert-lm/cli/usage>.
+
+- List all imported models:
+
+`litert-lm list`
+
+- Run a local or imported model interactively:
+
+`litert-lm run {{path/to/model.litertlm|model_id}}`
+
+- Run a model directly from a Hugging Face repository with a single prompt:
+
+`litert-lm run --from-huggingface-repo {{owner/repository}} {{model.litertlm}} --prompt "{{What is the capital of France?}}"`
+
+- Run a model using GPU acceleration:
+
+`litert-lm run {{path/to/model.litertlm|model_id}} --backend gpu`
+
+- Run a multimodal model with an image attachment:
+
+`litert-lm run {{path/to/model.litertlm|model_id}} --vision-backend {{gpu|cpu}} --attachment {{path/to/image.jpg}} --prompt "{{Describe this image.}}"`
+
+- Run a model with Python function calling tools via a preset file:
+
+`litert-lm run {{path/to/model.litertlm|model_id}} --preset {{path/to/preset.py}}`
+
+- Benchmark a model's performance:
+
+`litert-lm benchmark {{path/to/model.litertlm|model_id}}`
+
+- Start an OpenAI-compatible API server on a specific port:
+
+`litert-lm serve --port {{9379}}`