def test_hoppy_v1():
nb_entities = 10
nb_predicates = 5
embedding_size = 10
rs = np.random.RandomState(0)
for _ in range(16):
for nb_hops in range(6):
for use_attention in [True, False]:
with torch.no_grad():
entity_embeddings = nn.Embedding(nb_entities,
embedding_size * 2,
sparse=True)
predicate_embeddings = nn.Embedding(nb_predicates,
embedding_size * 2,
sparse=True)
base = ComplEx(entity_embeddings)
if use_attention:
reformulator = AttentiveReformulator(
nb_hops, predicate_embeddings)
else:
reformulator = LinearReformulator(
nb_hops, embedding_size * 2)
model = SimpleHoppy(base,
entity_embeddings,
hops=reformulator)
xs = torch.from_numpy(rs.randint(nb_entities, size=32))
xp = torch.from_numpy(rs.randint(nb_predicates, size=32))
xo = torch.from_numpy(rs.randint(nb_entities, size=32))
xs_emb = entity_embeddings(xs)
xp_emb = predicate_embeddings(xp)
xo_emb = entity_embeddings(xo)
scores = model.forward(xp_emb, xs_emb, xo_emb)
inf = model.score(xp_emb, xs_emb, xo_emb)
scores_sp, scores_po = scores
inf = inf.cpu().numpy()
scores_sp = scores_sp.cpu().numpy()
scores_po = scores_po.cpu().numpy()
for i in range(xs.shape[0]):
np.testing.assert_allclose(inf[i],
scores_sp[i, xo[i]],
rtol=1e-3,
atol=1e-3)
np.testing.assert_allclose(inf[i],
scores_po[i, xs[i]],
rtol=1e-3,
atol=1e-3)
def test_multi():
nb_entities = 10
nb_predicates = 5
embedding_size = 10
init_size = 1.0
rs = np.random.RandomState(0)
for _ in range(8):
for nb_hops in range(1, 6):
for use_attention in [True, False]:
for pt in {'max', 'min', 'sum', 'mixture'}:
with torch.no_grad():
entity_embeddings = nn.Embedding(nb_entities,
embedding_size * 2,
sparse=True)
predicate_embeddings = nn.Embedding(nb_predicates,
embedding_size * 2,
sparse=True)
entity_embeddings.weight.data *= init_size
predicate_embeddings.weight.data *= init_size
base = ComplEx(entity_embeddings)
models = []
for i in range(nb_hops):
if use_attention:
reformulator = AttentiveReformulator(
i, predicate_embeddings)
else:
reformulator = LinearReformulator(
i, embedding_size * 2)
h_model = SimpleHoppy(base,
entity_embeddings,
hops=reformulator)
models += [h_model]
model = Multi(models=models,
pooling_type=pt,
embedding_size=embedding_size * 2)
xs = torch.from_numpy(rs.randint(nb_entities, size=32))
xp = torch.from_numpy(
rs.randint(nb_predicates, size=32))
xo = torch.from_numpy(rs.randint(nb_entities, size=32))
xs_emb = entity_embeddings(xs)
xp_emb = predicate_embeddings(xp)
xo_emb = entity_embeddings(xo)
scores = model.forward(xp_emb, xs_emb, xo_emb)
inf = model.score(xp_emb, xs_emb, xo_emb)
scores_sp, scores_po = scores
inf = inf.cpu().numpy()
scores_sp = scores_sp.cpu().numpy()
scores_po = scores_po.cpu().numpy()
for i in range(xs.shape[0]):
np.testing.assert_allclose(inf[i],
scores_sp[i, xo[i]],
rtol=1e-3,
atol=1e-3)
np.testing.assert_allclose(inf[i],
scores_po[i, xs[i]],
rtol=1e-3,
atol=1e-3)
def _test_reasoning_v5(_st):
torch.set_num_threads(multiprocessing.cpu_count())
nb_entities = 10
nb_predicates = 5
embedding_size = 20
rs = np.random.RandomState(0)
triples = [('a', 'p', 'b'), ('b', 'q', 'c'), ('c', 'r', 'd'),
('d', 's', 'e')]
entity_to_index = {'a': 0, 'b': 1, 'c': 2, 'd': 3, 'e': 4}
predicate_to_index = {'p': 0, 'q': 1, 'r': 2, 's': 3}
for st in [_st]:
with torch.no_grad():
kernel = GaussianKernel()
entity_embeddings = nn.Embedding(nb_entities,
embedding_size * 2,
sparse=True)
predicate_embeddings = nn.Embedding(nb_predicates,
embedding_size * 2,
sparse=True)
fact_rel = torch.from_numpy(
np.array([predicate_to_index[p] for (_, p, _) in triples]))
fact_arg1 = torch.from_numpy(
np.array([entity_to_index[s] for (s, _, _) in triples]))
fact_arg2 = torch.from_numpy(
np.array([entity_to_index[o] for (_, _, o) in triples]))
facts = [fact_rel, fact_arg1, fact_arg2]
model = NeuralKB(entity_embeddings=entity_embeddings,
predicate_embeddings=predicate_embeddings,
kernel=kernel,
facts=facts,
scoring_type=st)
indices = torch.from_numpy(
np.array([
predicate_to_index['p'], predicate_to_index['q'],
predicate_to_index['r'], predicate_to_index['s']
]))
reformulator = SymbolicReformulator(predicate_embeddings, indices)
hoppy = SimpleHoppy(model, entity_embeddings, hops=reformulator)
rhoppy = RecursiveHoppy(model,
entity_embeddings,
hops=reformulator,
depth=1)
xs_np = rs.randint(nb_entities, size=32)
xp_np = rs.randint(nb_predicates, size=32)
xo_np = rs.randint(nb_entities, size=32)
xs_np[0] = 0
xp_np[0] = 0
xo_np[0] = 1
xs_np[1] = 1
xp_np[1] = 1
xo_np[1] = 2
xs_np[2] = 0
xp_np[2] = 3
xo_np[2] = 4
xs = torch.from_numpy(xs_np)
xp = torch.from_numpy(xp_np)
xo = torch.from_numpy(xo_np)
xs_emb = entity_embeddings(xs)
xp_emb = predicate_embeddings(xp)
xo_emb = entity_embeddings(xo)
scores = hoppy.forward(xp_emb, xs_emb, xo_emb)
inf = hoppy.score(xp_emb, xs_emb, xo_emb)
scores_h = rhoppy.depth_r_forward(xp_emb, xs_emb, xo_emb, depth=1)
inf_h = rhoppy.depth_r_score(xp_emb, xs_emb, xo_emb, depth=1)
print(inf)
print(inf_h)
assert inf[2] > 0.95
scores_sp, scores_po = scores
scores_h_sp, scores_h_po = scores_h
inf = inf.cpu().numpy()
scores_sp = scores_sp.cpu().numpy()
scores_po = scores_po.cpu().numpy()
inf_h = inf_h.cpu().numpy()
scores_h_sp = scores_h_sp.cpu().numpy()
scores_h_po = scores_h_po.cpu().numpy()
np.testing.assert_allclose(inf, inf_h)
np.testing.assert_allclose(scores_sp, scores_h_sp)
np.testing.assert_allclose(scores_po, scores_h_po)
for i in range(xs.shape[0]):
np.testing.assert_allclose(inf[i],
scores_sp[i, xo[i]],
rtol=1e-2,
atol=1e-2)
np.testing.assert_allclose(inf[i],
scores_po[i, xs[i]],
rtol=1e-2,
atol=1e-2)
np.testing.assert_allclose(inf_h[i],
scores_h_sp[i, xo[i]],
rtol=1e-2,
atol=1e-2)
np.testing.assert_allclose(inf_h[i],
scores_h_po[i, xs[i]],
rtol=1e-2,
atol=1e-2)
def test_masking_v2():
nb_entities = 10
nb_predicates = 5
embedding_size = 10
rs = np.random.RandomState(0)
for _ in range(1):
for position in [0, 1, 2]:
for st in ['min', 'concat']:
with torch.no_grad():
triples = [('a', 'p', 'b'), ('b', 'q', 'c'),
('a', 'p', 'c')]
entity_to_index = {'a': 0, 'b': 1, 'c': 2, 'd': 3}
predicate_to_index = {'p': 0, 'q': 1}
kernel = GaussianKernel()
entity_emb = nn.Embedding(nb_entities,
embedding_size * 2,
sparse=True)
predicate_emb = nn.Embedding(nb_predicates,
embedding_size * 2,
sparse=True)
fact_rel = torch.from_numpy(
np.array(
[predicate_to_index[p] for (_, p, _) in triples]))
fact_arg1 = torch.from_numpy(
np.array([entity_to_index[s]
for (s, _, _) in triples]))
fact_arg2 = torch.from_numpy(
np.array([entity_to_index[o]
for (_, _, o) in triples]))
facts = [fact_rel, fact_arg1, fact_arg2]
base = NeuralKB(entity_embeddings=entity_emb,
predicate_embeddings=predicate_emb,
kernel=kernel,
facts=facts,
scoring_type=st)
indices = torch.from_numpy(
np.array(
[predicate_to_index['p'],
predicate_to_index['q']]))
reformulator = SymbolicReformulator(predicate_emb, indices)
model = SimpleHoppy(base, entity_emb, hops=reformulator)
xs_np = rs.randint(nb_entities, size=32)
xp_np = rs.randint(nb_predicates, size=32)
xo_np = rs.randint(nb_entities, size=32)
xi_np = np.array([position] * xs_np.shape[0])
xs_np[0] = 0
xp_np[0] = 0
xo_np[0] = 1
xs_np[1] = 1
xp_np[1] = 1
xo_np[1] = 2
xs_np[2] = 0
xp_np[2] = 0
xo_np[2] = 2
xs = torch.from_numpy(xs_np)
xp = torch.from_numpy(xp_np)
xo = torch.from_numpy(xo_np)
xi = torch.from_numpy(xi_np)
xs_emb = entity_emb(xs)
xp_emb = predicate_emb(xp)
xo_emb = entity_emb(xo)
# xi = None
base.mask_indices = xi
scores = model.forward(xp_emb, xs_emb, xo_emb)
inf = model.score(xp_emb, xs_emb, xo_emb)
if position in {0, 1}:
assert inf[2] < 0.5
else:
assert inf[2] > 0.9
scores_sp, scores_po = scores
inf = inf.cpu().numpy()
scores_sp = scores_sp.cpu().numpy()
scores_po = scores_po.cpu().numpy()
for i in range(xs.shape[0]):
np.testing.assert_allclose(inf[i],
scores_sp[i, xo[i]],
rtol=1e-2,
atol=1e-2)
np.testing.assert_allclose(inf[i],
scores_po[i, xs[i]],
rtol=1e-2,
atol=1e-2)
def test_learning_v2():
embedding_size = 100
torch.manual_seed(0)
triples, hops = [], []
for i in range(16):
triples += [(f'a{i}', 'p', f'b{i}'), (f'b{i}', 'q', f'c{i}')]
hops += [(f'a{i}', 'r', f'c{i}')]
entity_lst = sorted({e
for (e, _, _) in triples + hops}
| {e
for (e, _, e) in triples + hops})
predicate_lst = sorted({p for (_, p, _) in triples + hops})
nb_entities, nb_predicates = len(entity_lst), len(predicate_lst)
entity_to_index = {e: i for i, e in enumerate(entity_lst)}
predicate_to_index = {p: i for i, p in enumerate(predicate_lst)}
kernel = GaussianKernel()
entity_embeddings = nn.Embedding(nb_entities,
embedding_size * 2,
sparse=True)
predicate_embeddings = nn.Embedding(nb_predicates,
embedding_size * 2,
sparse=True)
fact_rel = torch.from_numpy(
np.array([predicate_to_index[p] for (_, p, _) in triples]))
fact_arg1 = torch.from_numpy(
np.array([entity_to_index[s] for (s, _, _) in triples]))
fact_arg2 = torch.from_numpy(
np.array([entity_to_index[o] for (_, _, o) in triples]))
facts = [fact_rel, fact_arg1, fact_arg2]
model = NeuralKB(entity_embeddings=entity_embeddings,
predicate_embeddings=predicate_embeddings,
kernel=kernel,
facts=facts)
reformulator = AttentiveReformulator(2, predicate_embeddings)
hoppy = SimpleHoppy(model, entity_embeddings, hops=reformulator)
for s, p, o in hops:
xs_np = np.array([entity_to_index[s]])
xp_np = np.array([predicate_to_index[p]])
xo_np = np.array([entity_to_index[o]])
with torch.no_grad():
xs = torch.from_numpy(xs_np)
xp = torch.from_numpy(xp_np)
xo = torch.from_numpy(xo_np)
xs_emb = entity_embeddings(xs)
xp_emb = predicate_embeddings(xp)
xo_emb = entity_embeddings(xo)
inf = hoppy.score(xp_emb, xs_emb, xo_emb)
inf_np = inf.cpu().numpy()
assert inf_np < 0.5
def test_learning_v3():
embedding_size = 10
batch_size = 16
triples, hops = [], []
for i in range(16):
triples += [(f'a{i}', 'p', f'b{i}'), (f'b{i}', 'q', f'c{i}')]
hops += [(f'a{i}', 'r', f'c{i}')]
entity_lst = sorted({e
for (e, _, _) in triples + hops}
| {e
for (e, _, e) in triples + hops})
predicate_lst = sorted({p for (_, p, _) in triples + hops})
nb_entities, nb_predicates = len(entity_lst), len(predicate_lst)
entity_to_index = {e: i for i, e in enumerate(entity_lst)}
predicate_to_index = {p: i for i, p in enumerate(predicate_lst)}
torch.manual_seed(0)
kernel = GaussianKernel()
entity_embeddings = nn.Embedding(nb_entities,
embedding_size * 2,
sparse=True)
predicate_embeddings = nn.Embedding(nb_predicates,
embedding_size * 2,
sparse=True)
fact_rel = torch.from_numpy(
np.array([predicate_to_index[p] for (_, p, _) in triples]))
fact_arg1 = torch.from_numpy(
np.array([entity_to_index[s] for (s, _, _) in triples]))
fact_arg2 = torch.from_numpy(
np.array([entity_to_index[o] for (_, _, o) in triples]))
facts = [fact_rel, fact_arg1, fact_arg2]
model = NeuralKB(entity_embeddings=entity_embeddings,
predicate_embeddings=predicate_embeddings,
kernel=kernel,
facts=facts)
reformulator = AttentiveReformulator(2, predicate_embeddings)
hoppy = SimpleHoppy(model, entity_embeddings, hops=reformulator)
N3_reg = N3()
params = [
p for p in hoppy.parameters()
if not torch.equal(p, entity_embeddings.weight)
and not torch.equal(p, predicate_embeddings.weight)
]
loss_function = nn.CrossEntropyLoss(reduction='mean')
p_emb = predicate_embeddings(
torch.from_numpy(np.array([predicate_to_index['p']])))
q_emb = predicate_embeddings(
torch.from_numpy(np.array([predicate_to_index['q']])))
# r_emb = predicate_embeddings(torch.from_numpy(np.array([predicate_to_index['r']])))
optimizer = optim.Adagrad(params, lr=0.1)
hops_data = []
for i in range(128):
hops_data += hops
batches = make_batches(len(hops_data), batch_size)
c, d = 0.0, 0.0
for batch_start, batch_end in batches:
hops_batch = hops_data[batch_start:batch_end]
s_lst = [s for (s, _, _) in hops_batch]
p_lst = [p for (_, p, _) in hops_batch]
o_lst = [o for (_, _, o) in hops_batch]
xs_np = np.array([entity_to_index[s] for s in s_lst])
xp_np = np.array([predicate_to_index[p] for p in p_lst])
xo_np = np.array([entity_to_index[o] for o in o_lst])
xs = torch.from_numpy(xs_np)
xp = torch.from_numpy(xp_np)
xo = torch.from_numpy(xo_np)
xs_emb = entity_embeddings(xs)
xp_emb = predicate_embeddings(xp)
xo_emb = entity_embeddings(xo)
sp_scores, po_scores = hoppy.forward(xp_emb, xs_emb, xo_emb)
loss = loss_function(sp_scores, xo) + loss_function(po_scores, xs)
factors = [hoppy.factor(e) for e in [xp_emb, xs_emb, xo_emb]]
loss += 0.1 * N3_reg(factors)
tmp = hoppy.hops(xp_emb)
hop_1_emb = tmp[0]
hop_2_emb = tmp[1]
c = kernel.pairwise(p_emb, hop_1_emb).mean().cpu().detach().numpy()
d = kernel.pairwise(q_emb, hop_2_emb).mean().cpu().detach().numpy()
loss.backward()
optimizer.step()
optimizer.zero_grad()
assert c > 0.95
assert d > 0.95