You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This issue tracks the further development of autoquant tool
Goal
The goal for autoquant is to be able to get performance speedup over a broad set of models that we care about with minimal accuracy degradations (configurable by user), by reliably selecting the most performant quantization method and kernel implementation for the given input shape for each quantizable layer in the model. It could also be used for selecting hand written kernels that's optimized to get best performance on a specific model, runtime and device.
The content you are editing has changed. Please copy your edits and refresh the page.
This issue tracks the further development of autoquant tool
Goal
The goal for autoquant is to be able to get performance speedup over a broad set of models that we care about with minimal accuracy degradations (configurable by user), by reliably selecting the most performant quantization method and kernel implementation for the given input shape for each quantizable layer in the model. It could also be used for selecting hand written kernels that's optimized to get best performance on a specific model, runtime and device.
Performance
Accuracy
The text was updated successfully, but these errors were encountered: