onnx

Version:: 1.9.0, 1.6.0
Category:: ai
Cluster:: Loki

Author / Distributor

Description

ONNX (Open Neural Network Exchange) is an open ecosystem for representing deep learning models across frameworks. The modules listed here provide ONNX 1.9.0 with PyTorch integration, allowing models trained with PyTorch to be exported to ONNX format for interoperability and inference in other runtimes.

Documentation

import torch
import torch.onnx

# Sample model export
model = torch.nn.Linear(10, 5)
dummy_input = torch.randn(1, 10)
torch.onnx.export(model, dummy_input, "model.onnx")

# Validate model
import onnx
onnx_model = onnx.load("model.onnx")
onnx.checker.check_model(onnx_model)

# View model structure
print(onnx.helper.printable_graph(onnx_model.graph))

Full documentation: - ONNX: https://onnx.ai/ - PyTorch ONNX export: https://pytorch.org/docs/stable/onnx.html

Examples/Usage

List available modules:
```
$ module avail onnx
```

Load the onnx module:

$ module load onnx-pytorch-py39-cuda11.2-gcc9/1.9.0

Check the loaded modules:
```
$ module list
```

Unload the onnx module:

$ module unload onnx-pytorch-py39-cuda11.2-gcc9/1.9.0

Installation

Source code is obtained from onnx