Tutorials

Hello, MXA!

Hello, MXA!

Simple guide to quickly benchmark a model using the MemryX Accelerator.

hello/hello_mxa.html

Basic Inference

Classification using Simulator

Learn simple classification with the Neural Compiler & Simulator API.

basic_inf/classification_sim.html

Classification using Accelerator

Classify images using ResNet50 with the Neural Compiler & MXA.

basic_inf/classification_accl.html

Realtime Inference

Depth Estimation

Simple end-to-end Depth Estimation using MIDASv2-small in Python & C++.

realtime_inf/realtime_depth.html

YOLOv7t Object Detection

Single-stream end-to-end Object Detection using YOLOv7-tiny in Python.

realtime_inf/realtime_od.html

CenterNet Object Detection

Object Detection using CenterNet with cropped pre/post sections in C++.

realtime_inf/autocrop_inf/autocrop_centernet.html

YOLOv8 Pose Estimation

Pose Estimation using YOLOv8m-pose in Python & C++.

realtime_inf/realtime_pose.html

Multi-Stream Applications

MultiStream YOLOv7t Object Detection

MultiStream Object Detection using YOLOv7-tiny in Python & C++.

multistream_realtime_inf/multistream_od.html

MultiStream YOLOv8S Object Detection

MultiStream Object Detection using YOLOv8S in Python & C++.

multistream_realtime_inf/multistream_od_yolov8s.html

Multi-Model Applications

Face Detect & Emotion Classification

Perform multi-model inference by chaining

realtime_inf/realtime_multimodel.html

Fun Projects

Chrome Dino Game with Palm Detection

Play the Chrome Dino game using palm detection powered by the MX3 chip.

fun_projects/dino_game.html

Tiny Stories

Generate stories using the TinyStories model in Python.

text_generation/tinystories.html

Simple Integration

Deepface Integration

Learn how to integrate the MXA into a popular Face Recognition software.

app_integration/deepface.html

Before You Start

To complete the tutorials, you may need to install some third-party libraries. Please refer to the Third Party Libraries Installation guide for detailed instructions.

MemryX eXamples

For a complete list of end-to-end example applications, visit the MemryX eXamples GitHub page, where you’ll find full code implementations.

How-To

Compile and Benchmark Using APIs

Learn how to programmatically compile and benchmark a neural network model using MemryX APIs.

how_to/compile_benchmark_api.html

Model Cropping

Learn how to utilize the neural compiler’s model cropping feature.

how_to/cropping.html

Converting PyTorch Models to ONNX

This tutorial walks you through the steps of converting a PyTorch model to ONNX.

how_to/pytorch_to_onnx.html

Automate DFP Generation with Compiler API

This tutorial showcases a way of automating dfp compilation with compiler API.

how_to/automate_using_compiler_api.html

Accuracy Calculation

Classification Accuracy Calculation

Classification Accuracy Calculation for the ResNet50 (MLPerf) model.

accuracy/mlperf_accuracy/resnet50v1.5_mlperf_accuracy.html

Keras Classifier Models

Compare the accuracy of several Keras models on the MXA and the CPU.

accuracy/keras_classifiers_accuracy/keras_accuracy.html

Validating YOLOv8 Accuracy

Validate YOLOv8 mean Average Precision (mAP) on the COCO dataset for:

• Object detection

• Segmentation

• Pose estimation

accuracy/yolov8_accuracy/yolov8_accuracy.html

Advanced

MX3 Power Mode Tweaking

Tune the MX3 M.2’s performance higher or lower.

advanced/set_powermode.html

Weight Precision

Advanced tutorial for configuring per-layer weight precision settings beyond the default options.

advanced/weight_precision.html

Test with Fewer Chips

Deploy and test your model on fewer than 4 chips included in the M.2 module.

advanced/test_with_fewer_chips.html

High Precision Output Channels with YOLOv7

Learn how and when to use HPOC, demonstrated with YOLOv7.

advanced/hpoc_yolov7.html