Export from PyTorch
Save your trained network as ONNX, as it is.
pip install torch onnx onnxscriptExport
import torch
torch.onnx.export(model, torch.zeros(1, 4), "model.onnx", external_data=False)model is your trained network, and torch.zeros(1, 4) one row of input: change 4 to your number of features. external_data=False keeps everything in one file.
Then
Publish model.onnx. PeekML shows one box per feature. To name them and the classes, see Naming your features.
To show probabilities as bars, the network must return probabilities. If it returns raw scores (most do, because the loss applies softmax), export torch.nn.Sequential(model, torch.nn.Softmax(dim=1)) instead of model.
Image model? See Image models.