nonred_out=nOut,
d="cuda:0" if enableCuda else "cpu")
model.double()
if enableCuda:
model.cuda()
# declare optimizer and gradient and loss function
optimizer = optim.Adadelta(model.parameters(), lr=lr_rate)
loss = torch.nn.MSELoss(reduction='mean')
print("Loading model")
model.load_state_dict(torch.load(modelFile))
print("Starting Testing")
out_nrgs, test_val = model.test(images, energies, loss)
print("Testing Finished")
out_nrgs = convertToNumpy(out_nrgs, enableCuda)
# scaling energies to mHa
energies = 1000 * energies
out_nrgs = 1000 * out_nrgs
# calculating median absolute error for energies in range (100-400 mHa)
abs_err = abs(np.sort(energies) - np.sort(out_nrgs))
median_err = np.median(abs_err)
with open(outdir + "median_abs_error.txt", "w+") as f:
f.write("Median Absolute Error: " + str(median_err) + " mHa \n")
# plotting true energies vs. predicted energies
if not ((epoch + 1) % disp_epochs):
print("file name:{}".format(f) +
"\tEpoch [{}/{}]".format(epoch+1, num_epochs) +
"\tTraining Loss: {:.5f}".format(train_val))
print("Training Finished")
pt = files[0][files[0].index("_"):files[0].index("_")+4]
tsc = files[0][(files[0].index(" ")+1)+files[0][files[0].index(" ")+1:].index(" ")+1:files[0].index("]")]
print("Saving Model")
torch.save(model.state_dict(), moddir +
"model_{}{}".format(pt,tsc))
print("Starting Testing")
out_nrgs, test_val = model.test(t_img, t_nrg, loss)
print("Testing Finished")
out_nrgs = convertToNumpy(out_nrgs, enableCuda)
plt.plot(np.linspace(100, 400),
np.linspace(100, 400), color="black")
plt.scatter(np.sort(t_nrg)*1000, np.sort(out_nrgs)*1000, s=0.5, c='#FF0000')
plt.xlim((100,400))
plt.ylim((100,400))
plt.savefig(outdir + "InitialE.png")
plt.close()
plt.plot(obj_vals)
plt.savefig(outdir + "Training error.png")
plt.close()