import matplotlib.cm as cm
from mpl_toolkits.mplot3d import Axes3D
from sklearn.decomposition import PCA
from sklearn.linear_model import LogisticRegression
def warn(*args, **kwargs):
pass
import warnings
warnings.warn = warn
dataset_training = 'datasets/arithmetic/1e2/training.txt'
dataset_testing = 'datasets/arithmetic/1e2/testing.txt'
dataset = Dataset(dataset_training,dataset_testing)
num_layers = 1
hidden_size = 100
num_epochs = 5
input_size = dataset.vector_size
PATH = 'models/arithmetic_l_'+str(num_layers)+'_h_'+str(hidden_size)+'_ep_'+str(num_epochs)
model = GatedGRU(dataset.vector_size,hidden_size,output_size=1)
model.load_state_dict(torch.load(PATH))
model.eval()
#0 - sign, 1 - sign/hundreds, 2 - sign/tens, 3 - ones, 4-plus,
#5 - sign, 6 - sign/hundreds, 7 - sign/tens, 8 - ones, 9-equals
temporal_decoding = {0:[],1:[],2:[],3:[],4:[],
5:[],6:[],7:[],8:[],9:[]}
temporal_labels_first_num = {0:[],1:[],2:[],3:[],4:[],
import random
from tqdm import tqdm
import matplotlib.pyplot as plt
from matplotlib.colors import Normalize
from sklearn.decomposition import PCA
from textwrap import wrap
from tqdm import tqdm
dataset_training = 'datasets/arithmetic/fixed_L4_1e3/training.txt'
dataset_testing = 'datasets/arithmetic/fixed_L4_1e3/testing.txt'
dataset = Dataset(dataset_training, dataset_testing)
hidden_sizes = [100]
losses = []
for hidden in hidden_sizes:
print('Testing:', hidden)
total_loss = 0
num_layers = 1
hidden_size = hidden
num_epochs = 249
input_size = dataset.vector_size
PATH = 'models/arithmetic_L4_1e3_fixed_l_' + str(num_layers) + '_h_' + str(
hidden_size) + '_ep_' + str(num_epochs)
parser.add_argument('--training_set',
type=str,
default='datasets/arithmetic/fixed_1e2/training.txt')
parser.add_argument('--testing_set',
type=str,
default='datasets/arithmetic/fixed_1e2/testing.txt')
parser.add_argument('--model_prefix', type=str, default='arithmetic_1e2_fixed')
parser.add_argument('--dataset_type',
type=str,
default='normal',
help='{normal | polish}')
args = parser.parse_args()
if args.dataset_type == 'normal':
dataset = NormalDataset(args.training_set, args.testing_set)
elif args.dataset_type == 'polish':
dataset = PolishDataset(args.training_set, args.testing_set)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
if torch.cuda.is_available():
print('CUDA AVAILABLE')
input_size = dataset.vector_size
PATH = args.model_prefix + '_l_' + str(args.num_layers) + '_h_' + str(
args.hidden_size) + '_ep_' + str(args.num_epochs)
model = GatedGRU(input_size, args.hidden_size, output_size=1)
model.to(device)
criterion = nn.MSELoss()
def function_scaling(x):
return x
# if x < 0:
# return -np.log(-x+1)
# else:
# return np.log(x+1)
dataset_training = 'datasets/arithmetic/fixed_1e2/training.txt'
dataset_testing = 'datasets/arithmetic/fixed_1e2/testing.txt'
decoder_training_percent = 0.9
dataset = Dataset(dataset_training, dataset_testing)
num_layers = 1
hidden_size = 100
num_epochs = 449
input_size = dataset.vector_size
PATH = 'models/arithmetic_1e2_fixed_l_' + str(num_layers) + '_h_' + str(
hidden_size) + '_ep_' + str(num_epochs)
regression_trials = 1
model = GatedGRU(dataset.vector_size, hidden_size, output_size=1)
model.load_state_dict(torch.load(PATH))
model.eval()
temporal_hidden = {
from tqdm import tqdm import matplotlib.pyplot as plt from matplotlib.colors import Normalize from sklearn.decomposition import PCA from sklearn.linear_model import RidgeClassifier from textwrap import wrap from tqdm import tqdm dataset_training = 'datasets/arithmetic/1e2/training.txt' dataset_testing = 'datasets/arithmetic/1e2/testing.txt' dataset = Dataset(dataset_training,dataset_testing) num_layers = 1 hidden_size = 100 num_epochs = 5 input_size = dataset.vector_size PATH = 'models/arithmetic_l_'+str(num_layers)+'_h_'+str(hidden_size)+'_ep_'+str(num_epochs) model = GatedGRU(dataset.vector_size,hidden_size,output_size=1) model.load_state_dict(torch.load(PATH)) model.eval() wiu = model.W_iu.detach() whu = model.W_hu.detach() bu = model.b_u.detach()