Skip to content

Latest commit

 

History

History
59 lines (42 loc) · 2.53 KB

File metadata and controls

59 lines (42 loc) · 2.53 KB

About The Project

This repository hosts a Dockerfile and its dependencies that aims to build a container with TFLite installed to aid in testing Qualcomm platforms.

Build Daily LAVA

How to build

This isn't using any fancy features, so a regular build command will work:

docker build  --platform linux/arm64 .

How to use the container

Start the container with host networking and forwarding the GPU devices nodes inside /dev/dri:

docker run --network host --device /dev/dri -it --entrypoint /bin/bash <container URI>

Once inside run the helper script:

root@qrb2210-rb1-core-kit:/# ./run-tflite.sh
...
LAVA_RESULT test_case_id=tflite-label-image-cpu result=pass measurement=31.488 units=ms
...
LAVA_RESULT test_case_id=tflite-benchmark-mobilenet-quant-v1-224-cpu result=pass measurement=105.784 units=ms

Additional .tflite models mounted under /root/models are benchmarked recursively. The model directory can remain read-only because benchmark results are emitted on standard output instead of being written next to the models.

CI performance results

LAVA records each TensorFlow Lite latency as a native measurement. The workflow summary renders one Mermaid graph per board, with every measured test shown as an adjacent bar and an exact-value table below it. Missing accelerators remain absent rather than being plotted as zero.

Each run uploads a tflite-performance-<suite>-<boards> artifact containing results.json, results.csv, raw-logs/, and summary.md. Keying the artifact by its exact board set keeps targeted runs from displacing another board's baseline. The report includes the qcom-deb-images input, kernel, TensorFlow Lite revision, QAIRT version, container digest, AIML commit, and LAVA job/device. Compatible measurements from the previous report on the same branch are shown as informational changes.

License

AIML container test is licensed under the BSD-3-clause License. See LICENSE for the full license text.