Using the Winograd algorithm for convolution execution can provide increased performance compared to common implementation. However, it can be difficult to understand which algorithm would be faster due to dependency on convolution layer parameters and hardware configuration. The Winograd Algorithmic Tuner solves this problem automatically.
For more detailed information about the algorithm, refer to this whitepaper.
The Winograd Algorithmic Tuner workflow is:
Input: Original IR
Output: Modified IR with preassigned algorithm priority for each convolution layer
NOTE: OpenVINO™ supports the Winograd algorithm only for a limited set of convolution parameters. Some convolution layers with prioritization of the Winograd algorithm cannot be executed with this approach.
Once a job is finished, you can configure model optimization to Winograd.
NOTE: Using Winograd optimization, you can tune an original (top-level) model, or a model that has already been tuned.
Select the Optimize section, and then check Winograd.
Click Optimize and a new row of your model appears.
Once the job is complete, you can click on it to view inference results.