module training.orttraining
¶
Short summary¶
module onnxcustom.training.orttraining
Helper for onnxruntime-training.
Functions¶
function |
truncated documentation |
---|---|
Returns a name different from any name in existing_names. |
|
Modifies an ONNX graph to add operators to score and allow training. |
|
Returns the list of initializer to train. |
Documentation¶
Helper for onnxruntime-training.
- onnxcustom.training.orttraining._unique_name(existing_names, name, add=True)¶
Returns a name different from any name in existing_names.
- Parameters
existing_names – set of names
name – current
add – add the name of the list of existing names
- Returns
unique name
- onnxcustom.training.orttraining.add_loss_output(onx, score_name='squared_error', loss_name='loss', label_name='label')¶
Modifies an ONNX graph to add operators to score and allow training.
- Parameters
onx – onx graph
score_name – name of the score
loss_name – name of the output loss
label_name – name of the label input
- Returns
modified graph
Possible values for score_name:
‘squared_error’:
See example Train a scikit-learn neural network with onnxruntime-training on GPU.
- onnxcustom.training.orttraining.get_train_initializer(onx)¶
Returns the list of initializer to train.
- Returns
dictionary {name: (value, tensor)}