quark.onnx.graph_transformations.transforms_pipeline

Module Contents

Classes

class quark.onnx.graph_transformations.transforms_pipeline.TransformsPipeline(configs: Dict[str, Any] | None = None)

Wrapper of transforms to the model, apply in sequence. Transforms the original model to perform better during quantization.

get_configs() → Dict[str, Any] | None

Get the configurations.

Args:

None

Returns:

Dict of configurations

abstract apply(model: onnx.ModelProto, candidate_layers: Any, layer_metadata: Any) → Any

Apply list of transforms to onnx model.

Args:

model: onnx model to be quantized.

Returns:

New onnx model based on model which has been transformed.