Demo
Embedl Studio

Standalone Edge AI model visualizer and profiling app

A holistic Edge AI debugging tool to maximize developer productivity

Frame 1481 (6)

Designed to simplify Edge model debugging

Features for visual profiling and structural graph validation

Frame 1177 (3)

Graph & problem analysis

Trace structural changes through each compilation step. Visualize how vendor toolchains modify the graph to detect silent failures and unsupported operator fusions.

 

Performance analysis

Profile on-device latency, quantization parity, and operation breakdown. Map execution data directly to model layers to isolate hardware-level bottlenecks.

Frame 1180 (3)
Frame 1179 (3)

Compare model versions

Compare how different quantization scales and compiler flags affect the final execution graph. Analyze iterations side-by-side to verify accuracy and hardware constraints.

 

 
Supported frameworks
 

How does the app work in practice?

Understand your models inner workings

Understand and debug your edge models with Embedl Studio

Frequently Asked Questions

What is Embedl Studio

Embedl Studio is a tool for exploring and comparing multiple computation graphs side-by-side (for example PyTorch vs ONNX vs TensorRT), with synchronized navigation and cross-view highlighting, as well as per-layer latency and other profiling information.

How do I install the app?

The Embedl IDE is distributed as a Python package that you can `pip install`

What model and benchmark format are supported?

The Embedl Compiler and Embedl SDK outputs the correct metadata in a special `<model>.embedl` file which can be inspected in the Embedl IDE

What frameworks and hardware is supported?

All frameworks and hardware supported by the Embedl Compiler or Embedl SDK can be used with the Embedl IDE