odel = modelsargsmodel if torchcudais_available model = modelcuda optimizer = torchoptimAdammodelparameters lr=1e-4 weight_decay=1e-5 # optimizerload_state_dicttorchloadsave_optimpt

The first line initializes the model using the specified model type from the models dictionary.
The second line checks if a GPU is available and moves the model to the GPU if it is.
The third line initializes the optimizer with the Adam optimizer, a learning rate of 1e-4, and weight decay of 1e-5.
The fourth line initializes the criterion for calculating the loss, which is the cross-entropy loss.
The fifth line calculates the number of batches per epoch by dividing the length of the training dataset by the batch size.
The next few lines initialize lists to store the training and testing loss and accuracy values.
The next line creates a directory to save the model checkpoints.
The following for loop iterates over the specified number of epochs.
Inside the loop, the training data is iterated over in batches.
The input data and labels are prepared and moved to the GPU if available.
The train function is called to perform a forward pass, calculate the loss, and update the model parameters using the optimizer.
The training labels and predictions are stored in lists.
After each epoch, the training time, training accuracy, and training loss are calculated and stored in the respective lists.
The test function is called to evaluate the model on the test data and calculate the test accuracy, test loss, and other metrics.
The test accuracy and loss values are printed.
The model is saved every 10 epochs and the final model is saved at the end.
The test function is called again with the final model to calculate the test accuracy, test loss, NAR (Normalized Accuracy Rate), and feature list.
The test results are printed.
The feature list is saved as a CSV file.
The draw_result function is called to visualize the confusion matrix.
The draw function is called to plot the training and testing accuracy and loss curves.
Finally, the total training time is printed.