config_num = int(sys.argv[2]) if len(sys.argv) > 3: log_space = sys.argv[3][0] in "TtYy1" ################################ # Network Parameters midlayer, midlayerargs = configs[config_num] chars = data['chars'] nClasses = len(chars) nDims = len(data['x'][0]) nSamples = len(data['x']) nTrainSamples = nSamples * .75 nEpochs = 100 labels_print, labels_len = prediction_printer(chars) print("\nConfig: {}" "\n Midlayer: {} {}" "\nInput Dim: {}" "\nNum Classes: {}" "\nNum Samples: {}" "\nFloatX: {}" "\nUsing log space: {}" "\n".format(config_num, midlayer, midlayerargs, nDims, nClasses, nSamples, th.config.floatX, log_space)) ################################ print("Preparing the Data") try: conv_sz = midlayerargs["conv_sz"]
data = pickle.load(pkl_file) try: nHidden = int(sys.argv[2]) except IndexError: pass nClasses = data['nChars'] nDims = len(data['x'][0]) nSamples = len(data['x']) nTrainSamples = nSamples * .75 ntwk = RnnCTC(nDims, nHidden, nClasses) train_fn = ntwk.get_train_fn() test_fn = ntwk.get_test_fn() pred_print = prediction_printer(nClasses) data_x, data_y = [], [] for x, y in zip(data['x'], data['y']): # Need to make alternate characters blanks (index as nClasses) y1 = [nClasses] for char in y: y1 += [char, nClasses] data_y.append(np.asarray(y1, dtype=np.int32)) data_x.append(np.asarray(x, dtype=theano.config.floatX)) # Actual training for epoch in range(100): print('Epoch : ', epoch) for samp in range(nSamples): x = data_x[samp]
data = pickle.load(pkl_file) if len(sys.argv) > 2: config_num = int(sys.argv[2]) ################################ # Network Parameters midlayer, midlayerargs = configs[config_num] nClasses = data['nChars'] nDims = len(data['x'][0]) nSamples = len(data['x']) nTrainSamples = nSamples * .75 nEpochs = 100 labels_print, labels_len = prediction_printer(nClasses) print("\nConfig {}" "\n\tMidlayer: {} {}" "\nInput Dim: {}" "\nNum Classes: {}" "\nNum Samples: {}" "\n".format(config_num, midlayer, midlayerargs, nDims, nClasses, nSamples)) ################################ print("Preparing the Data") try: conv_sz = midlayerargs["conv_sz"] except KeyError: conv_sz = 1