class
OnnxRuntime::TrainingSession
- OnnxRuntime::TrainingSession
- Reference
- Object
Overview
TrainingSession class provides high-level API for training ONNX models. This is a placeholder for future implementation when ONNX Runtime training API is supported.
Defined in:
onnxruntime/training_session.crConstructors
-
.new(env : OrtEnvironment, model_path : String, **options)
Creates a new TrainingSession instance.
Instance Method Summary
-
#eval_step(input_feed)
Evaluate the model
-
#learning_rate=(rate : Float64)
Set learning rate
-
#load_checkpoint(checkpoint_path : String)
Load a checkpoint
-
#optimizer_state
Get optimizer state
-
#save_checkpoint(checkpoint_path : String)
Save the trained model
-
#train_step(input_feed)
Train the model for one step
Constructor Detail
def self.new(env : OrtEnvironment, model_path : String, **options)
#
Creates a new TrainingSession instance.