Set up and run a first evaluation
Install the framework, evaluate the wav2vec 2.0 model for speech recognition with no compression and with NNCodec, and read the result file.
Sweep quantisation parameters and plot the trade-off
Run the NNCodec evaluation over a range of quantisation parameters, plot word error rate against compressed size, and read the plot against the test cases of TR 26.847.
Add a compression method
Write a new coder, register it in the pipeline and compare it with the anchor and with NNCodec.
Network emulation and traffic characterisation
The framework in this repository does not emulate a network. For that, use the 6G AI Traffic Characterization Testbed tutorials.
Developer Exchange
No AI/ML Evaluation Framework-specific videos are curated here yet. In the meantime, here's what's recent from the wider Developer Exchange:
Connecting MBSTF to the MBS System - Developer Exchange
Introducing the Avatar Real-Time Communication Project - Developer Exchange
5G Media Streaming: Latest Additions - Developer Exchange
MBS User Services for 5G MBS (Release 18) - Developer Exchange
Advanced Media Delivery: CMCD Metrics in Action - Developer Exchange
Initial steps exposing APIs to AI systems through the MCP - Developer Exchange