dunedn.networks.onnx package

Submodules

dunedn.networks.onnx.onnx_abstract_net module

class dunedn.networks.onnx.onnx_abstract_net.OnnxNetwork(ckpt: Path, metrics: MetricsList, providers: Optional[list[str]] = None)

Bases: InferenceSession

Subclass

dunedn.networks.onnx.onnx_gcnn_net module

class dunedn.networks.onnx.onnx_gcnn_net.OnnxGcnnNetwork(ckpt: Path, metrics: MetricsList, providers: Optional[list[str]] = None)

Bases: OnnxNetwork

Subclass

predict(generator: GcnnDataset, profiler: Optional[BatchProfiler] = None) → Tensor

ONNX GCNN network inference.

Parameters

• generator (GcnnDataset) – The inference generator.

• profiler (BatchProfiler) – The profiler object to record batch inference time.

Returns

Output tensor of shape=(N,C,H,W).

Return type

torch.Tensor

dunedn.networks.onnx.uscg_onnx_net module

This module implements ONNX port of the USCG network.

dunedn.networks.onnx.utils module

dunedn.networks.onnx.utils.gcnn_onnx_inference_pass(test_loader: DataLoader, ort_session: InferenceSession, verbose: int = 1, profiler: Optional[BatchProfiler] = None) → Tensor

Parameters

• generator (torch.utils.data.DataLoader) – The inference dataset generator.

• ort_session (ort.InferenceSession) – The onnxruntime inference session.

• verbose (int) –

  Switch to log information. Defaults to 1. Available options:

  • 0: no logs.

  • 1: display progress bar.

• profiler (BatchProfiler) – The profiler object to record batch inference time.

Returns

Output tensor of shape=(N,C,H,W).

Return type

torch.Tensor

Module contents