aimet_onnx.analysis

Warning

This feature is under heavy development and API changes may occur without notice in future versions.

Visualization Tools

aimet_onnx.analysis.visualize_stats(sim, save_path='./quant_stats_visualization.html')[source]

Produces two interactive htmls to view the activation and weight encoding ranges from a calibrated QuantSim model.

Note

The QuantizationSimModel input is expected to have been calibrated before using this function. Stats will only be plotted for quantizers containing calibration statistics.

Creates interactive visualizations of min and max ranges of all quantized activations and weights in the input QuantSim object. The features include:

  • Adjustable threshold values to flag layers whose min or max values exceed the set thresholds

  • Tables containing names and ranges for layers exceeding threshold values

Two files are saved, derived from save_path by inserting _activations and _weights before the extension. For example, the default ./quant_stats_visualization.html produces ./quant_stats_visualization_activations.html and ./quant_stats_visualization_weights.html.

Example

>>> sim = QuantizationSimModel(model, quant_scheme=QuantScheme.post_training_tf)
>>> with compute_encodings(sim):
...     for input_data in data_loader:
...         sim.session.run(None, input_data)
...
>>> visualize_stats(sim, save_path="./quant_stats_visualization.html")
Parameters:

  • sim (QuantizationSimModel) – Calibrated QuantizationSimModel

  • save_path (str) – Base path for saving the visualizations. Default is “./quant_stats_visualization.html”

Return type:

None