What is the teacher-student distillation workflow?

It involves training a complex CSI-heavy teacher model offline and distilling its knowledge into a lightweight RSRP student model suitable for real-time inference.

Is it compatible with Sionna?

Yes, it is specifically designed to help refactor research ideas from Sionna-based simulations into modular, production-safe components.

Does this skill help with RIC xApp development?

Yes, it provides specific guidance on where to place code and how to structure schemas to ensure models are compatible with xApp skeletons and runtime constraints.

How does it handle tensor shapes?

The skill enforces strict shape contracts, such as (B, N_BEAMS), and provides testing patterns to ensure data consistency across the ML pipeline.

Beam Tracking ML & RL Pipeline

Name: Beam Tracking ML & RL Pipeline
Author: thc1006

bythc1006

0•

Data Science & ML

Design and refactor machine learning pipelines for 5G beam tracking using teacher-student distillation architectures.

This skill streamlines the development of reinforcement learning and machine learning models for beam tracking in telecommunications environments. It provides specialized guidance on architecting CSI-based teacher models and RSRP-based student models, enforcing strict tensor shape contracts, and ensuring models are optimized for real-time inference within RAN Intelligent Controller (RIC) xApps. It is particularly useful for developers bridging the gap between Sionna simulations and production-ready modular components, facilitating a structured workflow from offline training to online fine-tuning.

Key Features

01Modular refactoring of Sionna-based beam tracking scripts

02Standardized project structure for models and training scripts

03Integration-ready schemas for RIC xApp interfaces

040 GitHub stars

05Enforcement of strict tensor shape contracts for inference safety

06Teacher-student distillation workflow for RSRP student models

Use Cases

01Distilling CSI-heavy offline models into lightweight real-time RSRP policies

02Translating RL architectural diagrams into modular Python code

03Designing observation and action schemas for beam management systems

Key Features

01Modular refactoring of Sionna-based beam tracking scripts

02Standardized project structure for models and training scripts

03Integration-ready schemas for RIC xApp interfaces

040 GitHub stars

05Enforcement of strict tensor shape contracts for inference safety

06Teacher-student distillation workflow for RSRP student models

Use Cases

01Distilling CSI-heavy offline models into lightweight real-time RSRP policies

02Translating RL architectural diagrams into modular Python code

03Designing observation and action schemas for beam management systems