aimet_onnx.batch_norm_fold¶
Top-level API
- aimet_onnx.batch_norm_fold.fold_all_batch_norms_to_weight(model)[source]¶
Fold all possible batch_norm layers in a model into the weight of the corresponding conv layers
- Parameters:
model (
ModelProto) – onnx Model to perform BN fold on- Return type:
[typing.List]
- Returns:
A list of pairs of layers [(Conv/Linear, BN layer that got folded)]