Commit 6b460aa7 authored by Christian Chapman-Bird's avatar Christian Chapman-Bird
Browse files

latest models (needs cleaned up still). Added optional output directory... 

latest models (needs cleaned up still). Added optional output directory location relative to the current working directory
parent 27595f51

EMRI_DET/nn/model_creation.py

+5 −5
Original line number Diff line number Diff line
@@ -37,7 +37,7 @@ class LinearModel(nn.Module): 
            self.initial(m.weight)





def create_mlp(input_features, output_features, neurons, layers, activation, model_name, out_activation=None, init=xavier_uniform_, device=None, norm_type='z-score', use_dropout=False,drop_p=0.25, use_bn=False):

def create_mlp(input_features, output_features, neurons, layers, activation, model_name, out_activation=None, init=xavier_uniform_, device=None, norm_type='z-score', use_dropout=False,drop_p=0.25, use_bn=False, outdir='../models'):

    if isinstance(neurons, list):

        if len(neurons) != layers:

            raise RuntimeError('Length of neuron vector does not equal number of hidden layers.')
@@ -45,8 +45,8 @@ def create_mlp(input_features, output_features, neurons, layers, activation, mod 
        neurons = [neurons, ]

    model = LinearModel(input_features, output_features, neurons, layers, activation, model_name, initialisation=init, use_dropout=use_dropout,drop_p=drop_p,use_bn=use_bn, out_activation=out_activation)

    model.norm_type=norm_type

    Path(get_script_path()+f'/../models/{model_name}/').mkdir(parents=True, exist_ok=True)

    pickle.dump(model, open(get_script_path()+f'/../models/{model_name}/function.pickle', "wb"), pickle.HIGHEST_PROTOCOL)  # save blank model

    Path(get_script_path()+f'/{outdir}/{model_name}/').mkdir(parents=True, exist_ok=True)

    pickle.dump(model, open(get_script_path()+f'/{outdir}/{model_name}/function.pickle', "wb"), pickle.HIGHEST_PROTOCOL)  # save blank model



    if device is not None:

        model = model.to(device)
@@ -54,7 +54,7 @@ def create_mlp(input_features, output_features, neurons, layers, activation, mod 




def load_mlp(model_name, device, get_state_dict=False):

    model = pickle.load(open(get_script_path()+f'/../models/{model_name}/function.pickle', "rb"))  # load blank model

    model = pickle.load(open(get_script_path()+f'/{outdir}/{model_name}/function.pickle', "rb"))  # load blank model

    if get_state_dict:

        model.load_state_dict(torch.load(open(get_script_path()+f'/../models/{model_name}/model.pth', "rb"), map_location=device))

        model.load_state_dict(torch.load(open(get_script_path()+f'/{outdir}/{model_name}/model.pth', "rb"), map_location=device))

    return model

EMRI_DET/nn/model_train_test.py

+11 −11
Original line number Diff line number Diff line
@@ -6,7 +6,7 @@ from sys import stdout 
from pathlib import Path





def model_train_test(data, model, device, n_epochs, n_batches, loss_function, optimizer, verbose=False, return_losses=False, update_every=None, n_test_batches=None, save_best=False, scheduler=None):

def model_train_test(data, model, device, n_epochs, n_batches, loss_function, optimizer, verbose=False, return_losses=False, update_every=None, n_test_batches=None, save_best=False, scheduler=None, outdir='../models'):

    

    if n_test_batches is None:

        n_test_batches = n_batches
@@ -17,13 +17,13 @@ def model_train_test(data, model, device, n_epochs, n_batches, loss_function, op 
    name = model.name

    path = get_script_path()

    norm_type = model.norm_type

    Path(get_script_path()+f'/../models/{name}/').mkdir(parents=True, exist_ok=True)

    Path(get_script_path()+f'/{outdir}/{name}/').mkdir(parents=True, exist_ok=True)

    if norm_type == 'z-score':

        np.save(path+'/../models/'+name+'/xdata_inputs.npy',np.array([xtrain.mean(axis=0), xtrain.std(axis=0)]))

        np.save(path+'/../models/'+name+'/ydata_inputs.npy',np.array([ytrain.mean(), ytrain.std()]))

        np.save(path+f'/{outdir}/{name}/xdata_inputs.npy',np.array([xtrain.mean(axis=0), xtrain.std(axis=0)]))

        np.save(path+f'/{outdir}/{name}/ydata_inputs.npy',np.array([ytrain.mean(), ytrain.std()]))

    elif norm_type == 'uniform':

        np.save(path+'/../models/'+name+'/xdata_inputs.npy',np.array([np.min(xtrain,axis=0), np.max(xtrain,axis=0)]))

        np.save(path+'/../models/'+name+'/ydata_inputs.npy',np.array([np.min(ytrain), np.max(ytrain)]))

        np.save(path+f'/{outdir}/{name}/xdata_inputs.npy',np.array([np.min(xtrain,axis=0), np.max(xtrain,axis=0)]))

        np.save(path+f'/{outdir}/{name}/ydata_inputs.npy',np.array([np.min(ytrain), np.max(ytrain)]))

    xtest = torch.from_numpy(norm_inputs(xtest, ref_dataframe=xtrain, norm_type=norm_type)).to(device).float()

    ytest = torch.from_numpy(norm(ytest, ref_dataframe=ytrain, norm_type=norm_type)).to(device).float()

    xtrain = torch.from_numpy(norm_inputs(xtrain, ref_dataframe=xtrain, norm_type=norm_type)).to(device).float()
@@ -92,7 +92,7 @@ def model_train_test(data, model, device, n_epochs, n_batches, loss_function, op 
        if test_losses[-1] < lowest_loss:

            lowest_loss = test_losses[-1]

            if save_best:

                torch.save(model.state_dict(),path+'/../models/'+name+'/model.pth')

                torch.save(model.state_dict(),path+f'/{outdir}/{name}/model.pth')

                

#         if epoch >= cutoff_LR:

#             scheduler.step()
@@ -110,19 +110,19 @@ def model_train_test(data, model, device, n_epochs, n_batches, loss_function, op 
                plt.xlabel('Epochs')

                plt.ylabel('Loss')

                plt.title('Train and Test Loss Across Train Epochs')

                plt.savefig(path+'/../models/'+name+'/losses.png')

                plt.savefig(path+f'/{outdir}/{name}/losses.png')

                #plt.show()

                plt.close()

                

                if not save_best:

                    torch.save(model.state_dict(),path+'/../models/'+name+'/model.pth')

                    torch.save(model.state_dict(),path+f'/{outdir}/{name}/model.pth')



        

    if verbose:

        print('\nTraining complete - saving.')

    

    if not save_best:

        torch.save(model.state_dict(),path+'/../models/'+name+'/model.pth')

        torch.save(model.state_dict(),path+f'/{outdir}/{name}/model.pth')



    epochs = np.arange(n_epochs)

    plt.semilogy(epochs, train_losses, label='Train')
@@ -131,7 +131,7 @@ def model_train_test(data, model, device, n_epochs, n_batches, loss_function, op 
    plt.xlabel('Epochs')

    plt.ylabel('Loss')

    plt.title('Train and Test Loss Across Train Epochs')

    plt.savefig(path+'/../models/'+name+'/losses.png')

    plt.savefig(path+f'/{outdir}/{name}/losses.png')

    #plt.show()

    plt.close()

EMRI_DET/validate.py

+7 −7
Original line number Diff line number Diff line
@@ -6,7 +6,7 @@ from EMRI_DET.utilities import norm_inputs, unnorm, get_script_path 
import seaborn as sns





def run_on_dataset(model, test_data, distances=None, n_batches=1, device=None, y_transform_fn=None, runtime=False):

def run_on_dataset(model, test_data, distances=None, n_batches=1, device=None, y_transform_fn=None, runtime=False, outdir='../models'):

    """

    Get the re-processed output of the supplied model on a set of supplied test data.
@@ -36,8 +36,8 @@ def run_on_dataset(model, test_data, distances=None, n_batches=1, device=None, y 


    xdata, ydata = test_data



    xscalevals = np.load(get_script_path() + f'/../models/{model.name}/xdata_inputs.npy')

    yscalevals = np.load(get_script_path() + f'/../models/{model.name}/ydata_inputs.npy')

    xscalevals = np.load(get_script_path() + f'/{outdir}/{model.name}/xdata_inputs.npy')

    yscalevals = np.load(get_script_path() + f'/{outdir}/{model.name}/ydata_inputs.npy')



    test_input = torch.Tensor(xdata)

    normed_input = norm_inputs(test_input, ref_inputs=xscalevals,norm_type=model.norm_type).float().to(device)
@@ -108,7 +108,7 @@ def test_threshold_accuracy(comparison_sets, threshold, confusion_matrix=False): 
        return (1-np.mean(np.abs(out_classified-truth_classified)),confmat)



def plot_histograms(comparison_sets, model_name, xlabel, title=None, title_kwargs={}, xlabel_kwargs={}, log=True,

                    fig_kwargs={}, plot_kwargs={}, save_kwargs={}, legend_kwargs={}):

                    fig_kwargs={}, plot_kwargs={}, save_kwargs={}, legend_kwargs={}, outdir='../models'):

    truth, pred = comparison_sets

    if log:

        truth = np.log10(truth)
@@ -130,12 +130,12 @@ def plot_histograms(comparison_sets, model_name, xlabel, title=None, title_kwarg 
        ax.set_title(title, **title_kwargs)



    ax.legend(**legend_kwargs)

    plt.savefig(get_script_path() + f'/../models/{model_name}/hists.png', **save_kwargs)

    plt.savefig(get_script_path() + f'/{outdir}/{model_name}/hists.png', **save_kwargs)

    plt.close()





def plot_difference_histogram(comparison_sets, model_name, xlabel, title=None, title_kwargs={}, xlabel_kwargs={},

                              log=True, one_sided=False, ratio=False, fig_kwargs={}, plot_kwargs={}, save_kwargs={}):

                              log=True, one_sided=False, ratio=False, fig_kwargs={}, plot_kwargs={}, save_kwargs={}, outdir='../models'):

    truth, pred = comparison_sets

    if log:

        truth = np.log10(truth)
@@ -175,7 +175,7 @@ def plot_difference_histogram(comparison_sets, model_name, xlabel, title=None, t 
    else:

        logname = ''



    plt.savefig(get_script_path() + f'/../models/{model_name}/hist_{logname}{diff}diff.png', **save_kwargs)

    plt.savefig(get_script_path() + f'/{outdir}/{model_name}/hist_{logname}{diff}diff.png', **save_kwargs)

    plt.close()

emri_data/models/prograde_SiLU_linearOut/function.pickle

0 → 100644
+465 KiB

File added.

No diff preview for this file type.

View file

emri_data/models/prograde_SiLU_linearOut/losses.png

0 → 100644
+24.4 KiB
Loading image diff...