In this folder, I have three reference documents: pto2_rt.md describes the pto2 runtime design principles. the pypto-frontend-coding-style.md describes the coding method that we intend to use to build up pypto-lib library. the pypto-lib.md describes how the pypto-lib library should be built up.
In ../pypto_v2 we have the project folder of old version of pypto (pypto_v2) project. In this project, the models subfolder contains the top down design using tensor operations in pypto_v2 grammar of several important models such as artic, deepseek_v32_exp, glm_v4_5, qwen3_next. this model uses a set of tensor functions. the pypto_v2 framework contains a pass named "expand_function" that converts basic tensor operations (tensor functions) into tile operations (tile functions). I need to analyze all the models to understand the models and the collection of basic tensor operations, and analyze the expand_function pass code to understand how tensor operations as converted into tile operations. I want you to make a plan to create two folders in pypto-lib project: tensor_functions and models. the tensor_functions contains the full collection of tensor operations implementation using pypto coding style described in pypto-frontend-coding-style.md. when implementing these function, define all tensor functions as opaque functions (without specifying the boundary between orchestration and incore function). in the models folder, implement all models found in pypto_v2 models folder, again using pypto front end new coding style specified in pypto-frontend-coding-style.md, implement the whole models as opaque functions. for more information about pypto-front-end, you can can find the front end new version in the pypto folder, and you need check the implemented tensor functions and model functions with the pypto folder to ensure they fully comply with the latest pypto coding style.