input_embeddings,
update_embeddings = update_embeddings,
hidden_dim = hidden_dim,
order = 3,
)
elif args["lstm"]:
net = networks.LSTM(
model,
input_embeddings,
update_embeddings = update_embeddings,
hidden_dim = hidden_dim,
)
elif args["bow"]:
net = networks.BOW(
model,
input_embeddings,
update_embeddings = update_embeddings,
hidden_dim = hidden_dim,
)
elif args["tree-lstm"]:
input_embeddings = np.load("data/dicteval/input_embeddings_parsed.reduced.npy")
net = networks.CYK(
model,
input_embeddings,
update_embeddings = update_embeddings,
hidden_dim = hidden_dim,
)
parsed = True
if args["--restart"]:
restart = int(args["--restart"])
else:
input_embeddings,
update_embeddings=False,
hidden_dim=100,
order=3,
)
elif args["lstm"]:
net = networks.LSTM(
model,
input_embeddings,
update_embeddings=False,
hidden_dim=100,
)
elif args["bow"]:
net = networks.BOW(
model,
input_embeddings,
update_embeddings=False,
hidden_dim=100,
)
elif args["tree-lstm"]:
net = networks.CYK(
model,
input_embeddings,
update_embeddings=False,
hidden_dim=100,
)
parsed = True
classifier = networks.SNLIClassifier(model, input_embeddings.shape[1])
model.load(args["<model-file>"])
print(eval_nli_dataset(net, classifier, test, parsed))