def test_np_empty():
dtypes = [np.int8, np.int32, np.float16, np.float32, np.float64, None]
expected_dtypes = [
np.int8, np.int32, np.float16, np.float32, np.float64, np.float32
]
orders = ['C', 'F', 'A']
shapes = [
(),
0,
(0, ),
(0, 0),
2,
(2, ),
(3, 0),
(4, 5),
(1, 1, 1, 1),
]
ctxes = [npx.current_context(), None]
for dtype, expected_dtype in zip(dtypes, expected_dtypes):
for shape in shapes:
for order in orders:
for ctx in ctxes:
if order == 'C':
ret = np.empty(shape, dtype, order, ctx)
assert ret.dtype == expected_dtype
assert ret.shape == shape if isinstance(
shape, tuple) else (shape, )
assert ret.ctx == npx.current_context()
else:
assert_exception(np.empty, NotImplementedError, shape,
dtype, order, ctx)
def __init__(self, user_ids, item_ids, L, num_users, num_items,
candidates):
user_ids, item_ids = np.array(user_ids), np.array(item_ids)
sort_idx = np.array(sorted(range(len(user_ids)),
key=lambda k: user_ids[k]))
u_ids, i_ids = user_ids[sort_idx], item_ids[sort_idx]
temp, u_ids, self.cand = {}, u_ids.asnumpy(), candidates
self.all_items = set([i for i in range(num_items)])
# 按照u_id-item编号,一个u的所有item完了，才下一个u的item编号
[temp.setdefault(u_ids[i], []).append(i) for i, _ in enumerate(u_ids)]
temp = sorted(temp.items(), key=lambda x: x[0])
u_ids = np.array([i[0] for i in temp])
idx = np.array([i[1][0] for i in temp])
# ns是什么意思
self.ns = ns = int(sum([c - L if c >= L + 1 else 1 for c
in np.array([len(i[1]) for i in temp])]))
self.seq_items = np.zeros((ns, L))
self.seq_users = np.zeros(ns, dtype='int32')
self.seq_tgt = np.zeros((ns, 1)) # target序列
self.test_seq = np.zeros((num_users, L))
test_users, _uid = np.empty(num_users), None
for i, (uid, i_seq) in enumerate(self._seq(u_ids, i_ids, idx, L + 1)):
# 最后一个序列为test集
if uid != _uid:
# 这里也感觉不对
self.test_seq[uid][:] = i_seq[-L:]
test_users[uid], _uid = uid, uid
# 这里总感觉不对
self.seq_tgt[i][:] = i_seq[-1:]
self.seq_items[i][:], self.seq_users[i] = i_seq[:L], uid
def get_conv_data_mxnet(oc, ic, n, k, p, s):
mpx.random.seed(0)
data = mp.random.normal(size=(1, ic, n, n))
weight = mp.random.normal(size=(oc, ic, k, k))
bias = mp.zeros((oc, ))
on = conv_out_size(n, k, p, s)
out = mp.empty((1, oc, on, on))
# Wait data are generated to make later benchmarking accurate
mpx.waitall()
return data, weight, bias, out
def test_np_empty():
# (input dtype, expected output dtype)
dtype_pairs = [
(np.int8, np.int8),
(np.int32, np.int32),
(np.float16, np.float16),
(np.float32, np.float32),
(np.float64, np.float64),
(np.bool_, np.bool_),
(np.bool, np.bool_),
('int8', np.int8),
('int32', np.int32),
('float16', np.float16),
('float32', np.float32),
('float64', np.float64),
('bool', np.bool_),
(None, np.float32),
]
orders = ['C', 'F', 'A']
shapes = [
(),
0,
(0, ),
(0, 0),
2,
(2, ),
(3, 0),
(4, 5),
(1, 1, 1, 1),
]
ctxes = [npx.current_context(), None]
for dtype, expected_dtype in dtype_pairs:
for shape in shapes:
for order in orders:
for ctx in ctxes:
if order == 'C':
ret = np.empty(shape, dtype, order, ctx)
assert ret.dtype == expected_dtype
assert ret.shape == shape if isinstance(
shape, tuple) else (shape, )
assert ret.ctx == npx.current_context()
else:
assert_exception(np.empty, NotImplementedError, shape,
dtype, order, ctx)
# vector 생성 x = np.arange(12) x x.shape # shape 확인 x.size # size 확인 # shape chahe x = x.reshape(3, 4) x x.shape # 자동으로 남은 값들의 dimention 작동 x.reshape(-1, 4) x.reshape(3, -1) # empty 메모리의 덩어리를 가져오는 mothod np.empty((3, 4)) # 0으로 값을 채우기 np.zeros((2, 3, 4)) # 1로 값을 채우기 np.ones((2, 3, 4)) # 평균이 0이고, 표준편차가 1인 (3, 4) dimention 만들기 np.random.normal(0, 1, size=(3, 4)) # 직접 리스트로 dimention 만들기 np.array([[2, 1, 4, 3], [1, 2, 3, 4], [4, 3, 2, 1]])