def Test(num_datapoints):
tokens_to_index = get_token_dict_from_file()
err_count = 0
count = 0
for datapoint in test_dataset():
conjecture = datapoint.conjecture
statement = datapoint.statement
label = datapoint.label
conjecture_statement = [conjecture, statement]
prediction_val = F([conjecture_statement])
_, prediction_label = torch.max(prediction_val, dim=1)
if cuda_available:
prediction_label = prediction_label.cpu()
prediction_label = prediction_label.numpy()
if datapoint.label != prediction_label[0]:
err_count += 1
count += 1
if count % 100 == 0:
print("Count: ", count)
print("Fraction of Incorrect Test Points: ", err_count / count)
return err_count / count
def Validate(num_datapoints):
tokens_to_index = get_token_dict_from_file()
err_count = 0
count = 0
for datapoint in validation_dataset():
conjecture = datapoint.conjecture
statement = datapoint.statement
label = datapoint.label
for node_id, node_obj in conjecture.nodes.items(
): # Find and replace unknowns
if node_obj.token not in tokens_to_index.keys(): # UNKOWN token
node_obj.token = "UNKNOWN"
for node_id, node_obj in statement.nodes.items():
if node_obj.token not in tokens_to_index.keys():
node_obj.token = "UNKNOWN"
prediction_val = F([conjecture], [statement])
_, prediction_label = torch.max(prediction_val, dim=1)
if cuda_available:
prediction_label = prediction_label.cpu()
prediction_label = prediction_label.numpy()
# print(label)
# print(prediction_label)
if datapoint.label != prediction_label[0]:
err_count += 1
count += 1
if count % 100 == 0:
print("Count: ", count)
if count == num_datapoints:
break
print("Fraction of Incorrect Validations: ", err_count / count)
return err_count / count
def __init__(self, num_steps, cuda_available=False):
super(FormulaNet, self).__init__()
# Initialize models
self.dense_map = LinearMap() # maps one_hot -> 256 dimension vector
self.FP = FPClass()
self.FI = FIClass()
self.FO = FOClass()
self.FL = FLClass()
self.FR = FRClass()
self.FH = FHClass()
self.Classifier = CondClassifier()
self.Softmax = nn.Softmax(dim=1)
self.max_pool_dense_graph = max_pool_dense_graph()
self.num_steps = num_steps
self.token_to_index = get_token_dict_from_file()
self.cuda_available = cuda_available