Create Training Session From Model

 max_norm : float, maximum global gradient norm (enables clipping if > 0).
 norm_type : enum, type of norm used to compute grad_norm (commonly 1 = L1, 2 = L2).
 display_norm : boolean, adds grad_norm as a model output if set to 1.
 keep_outputs : boolean, keeps the original model outputs in addition to the new ones (loss values and, if enabled, gradients or gradient norms).
 Sessions Parameters : cluster

 intra_op_num_threads : integer, number of threads used within each operator to parallelize computations. If the value is 0, ONNX Runtime automatically uses the number of physical CPU cores.
 inter_op_num_threads : integer, number of threads used between operators, to execute multiple graph nodes in parallel. If set to 0, this parameter is ignored when execution_mode is ORT_SEQUENTIAL. In ORT_PARALLEL mode, 0 means ORT automatically selects a suitable number of threads (usually equal to the number of cores).
 execution_mode : enum, controls whether the graph executes nodes one after another or allows parallel execution when possible. ORT_SEQUENTIAL runs nodes in order, ORT_PARALLEL runs them concurrently.
 deterministic_compute : boolean, forces deterministic execution, meaning results will always be identical for the same inputs.
 graph_optimization_level : enum, defines how much ONNX Runtime optimizes the computation graph before running the model.
 optimized_model_file_path : path, file path to save the optimized model after graph analysis.

checkpoint generation : enum, defines what is saved in the checkpoint: "Weight" (weights only) or "Weight + Momentum" (weights and optimizer momentums). Momentums must already be set in the "Model in".