wehub-resource-sync
1009 lines
32 KiB
Python
1009 lines
32 KiB
Python
# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
import unittest
|
|
|
|
import numpy as np
|
|
from op_test import (
|
|
OpTest,
|
|
convert_float_to_uint16,
|
|
get_device_place,
|
|
is_custom_device,
|
|
)
|
|
|
|
import paddle
|
|
from paddle import base
|
|
from paddle.base import Program, core, program_guard
|
|
from paddle.framework import in_pir_mode
|
|
|
|
|
|
class TestBaddBmmOp(OpTest):
|
|
# test basic
|
|
def setUp(self):
|
|
self.op_type = "baddbmm"
|
|
self.prim_op_type = "comp"
|
|
self.python_api = paddle.baddbmm
|
|
self.public_python_api = paddle.baddbmm
|
|
self.init_dtype_type()
|
|
self.inputs = {
|
|
'Input': np.random.random((2, 10, 5)).astype(self.dtype),
|
|
'X': np.random.random((2, 10, 10)).astype(self.dtype),
|
|
'Y': np.random.random((2, 10, 5)).astype(self.dtype),
|
|
}
|
|
self.outputs = {
|
|
'Out': self.inputs['Input']
|
|
+ np.matmul(self.inputs['X'], self.inputs['Y'])
|
|
}
|
|
|
|
def init_dtype_type(self):
|
|
self.dtype = np.float64
|
|
|
|
def test_check_output(self):
|
|
self.check_output(check_pir=True, check_prim_pir=True)
|
|
|
|
def test_check_grad_normal(self):
|
|
self.check_grad(
|
|
['Input', 'X', 'Y'],
|
|
'Out',
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
def test_check_grad_x(self):
|
|
self.check_grad(
|
|
['X'],
|
|
'Out',
|
|
no_grad_set=None,
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
def test_check_grad_y(self):
|
|
self.check_grad(
|
|
['Y'],
|
|
'Out',
|
|
no_grad_set=None,
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
def test_check_grad_input(self):
|
|
self.check_grad(
|
|
['Input'],
|
|
'Out',
|
|
no_grad_set=None,
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
|
|
@unittest.skipIf(
|
|
not (core.is_compiled_with_cuda() or is_custom_device())
|
|
or not core.is_float16_supported(get_device_place()),
|
|
"core is not compiled with CUDA or not support float16",
|
|
)
|
|
class TestBaddBmmFP16Op(OpTest):
|
|
def setUp(self):
|
|
self.op_type = "baddbmm"
|
|
self.python_api = paddle.baddbmm
|
|
self.init_dtype_type()
|
|
self.inputs = {
|
|
'Input': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
'X': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
'Y': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
}
|
|
self.outputs = {
|
|
'Out': self.inputs['Input']
|
|
+ np.matmul(self.inputs['X'], self.inputs['Y'])
|
|
}
|
|
|
|
self.place = get_device_place()
|
|
|
|
def init_dtype_type(self):
|
|
self.dtype = np.float16
|
|
|
|
def test_check_output(self):
|
|
self.check_output_with_place(self.place)
|
|
|
|
def test_check_grad_normal(self):
|
|
self.check_grad_with_place(self.place, ['Input', 'X', 'Y'], 'Out')
|
|
|
|
def test_check_grad_x(self):
|
|
self.check_grad_with_place(self.place, ['X'], 'Out', no_grad_set=None)
|
|
|
|
def test_check_grad_y(self):
|
|
self.check_grad_with_place(self.place, ['Y'], 'Out', no_grad_set=None)
|
|
|
|
def test_check_grad_input(self):
|
|
self.check_grad_with_place(
|
|
self.place, ['Input'], 'Out', no_grad_set=None
|
|
)
|
|
|
|
|
|
@unittest.skipIf(
|
|
not (core.is_compiled_with_cuda() or is_custom_device())
|
|
or not core.is_bfloat16_supported(get_device_place()),
|
|
"core is not compiled with CUDA or not support bfloat16",
|
|
)
|
|
class TestBaddBmmBF16Op(OpTest):
|
|
def setUp(self):
|
|
self.op_type = "baddbmm"
|
|
self.python_api = paddle.baddbmm
|
|
self.init_dtype_type()
|
|
self.inputs = {
|
|
'Input': np.random.random((2, 50, 1)).astype(self.dtype),
|
|
'X': np.random.random((2, 50, 5)).astype(self.dtype),
|
|
'Y': np.random.random((2, 5, 10)).astype(self.dtype),
|
|
}
|
|
self.outputs = {
|
|
'Out': self.inputs['Input']
|
|
+ np.matmul(self.inputs['X'], self.inputs['Y'])
|
|
}
|
|
|
|
self.inputs['Input'] = convert_float_to_uint16(self.inputs['Input'])
|
|
self.inputs['X'] = convert_float_to_uint16(self.inputs['X'])
|
|
self.inputs['Y'] = convert_float_to_uint16(self.inputs['Y'])
|
|
self.outputs['Out'] = convert_float_to_uint16(self.outputs['Out'])
|
|
self.place = get_device_place()
|
|
|
|
def init_dtype_type(self):
|
|
self.dtype = np.uint16
|
|
self.np_dtype = np.float32
|
|
|
|
def test_check_output(self):
|
|
self.check_output_with_place(self.place)
|
|
|
|
def test_check_grad_normal(self):
|
|
self.check_grad_with_place(self.place, ['Input', 'X', 'Y'], 'Out')
|
|
|
|
def test_check_grad_x(self):
|
|
self.check_grad_with_place(self.place, ['X'], 'Out', no_grad_set=None)
|
|
|
|
def test_check_grad_y(self):
|
|
self.check_grad_with_place(self.place, ['Y'], 'Out', no_grad_set=None)
|
|
|
|
def test_check_grad_input(self):
|
|
self.check_grad_with_place(
|
|
self.place, ['Input'], 'Out', no_grad_set=None
|
|
)
|
|
|
|
|
|
class TestBaddBmmOpError(unittest.TestCase):
|
|
# test error
|
|
def test_errors(self):
|
|
with program_guard(Program(), Program()):
|
|
# The input type of baddbmm_op must be Variable.
|
|
|
|
input = base.create_lod_tensor(
|
|
np.array([[[-1, -1], [-1, -1]], [[-1, -1], [-1, -1]]]),
|
|
[[2]],
|
|
base.CPUPlace(),
|
|
)
|
|
x1 = base.create_lod_tensor(
|
|
np.array([[[-1, -1], [-1, -1]], [[-1, -1], [-1, -1]]]),
|
|
[[2]],
|
|
base.CPUPlace(),
|
|
)
|
|
x2 = base.create_lod_tensor(
|
|
np.array([[[-1, -1], [-1, -1]], [[-1, -1], [-1, -1]]]),
|
|
[[2]],
|
|
base.CPUPlace(),
|
|
)
|
|
# After code sinking to C++, the error type changed from TypeError to ValueError
|
|
self.assertRaises(
|
|
(TypeError, ValueError), paddle.baddbmm, input, x1, x2
|
|
)
|
|
|
|
paddle.enable_static()
|
|
# The input dtype of baddbmm_op must be float32 or float64.
|
|
input = paddle.static.data(
|
|
name='input',
|
|
shape=[2, 4, 4],
|
|
dtype="int32",
|
|
)
|
|
x3 = paddle.static.data(name='x3', shape=[2, 4, 4], dtype="int32")
|
|
x4 = paddle.static.data(name='x4', shape=[2, 4, 4], dtype="int32")
|
|
self.assertRaises(TypeError, paddle.baddbmm, input, x3, x4)
|
|
# x and y dimension mismatch
|
|
x5 = paddle.static.data(
|
|
name='x5',
|
|
shape=[2, 4, 5],
|
|
dtype="float32",
|
|
)
|
|
x6 = paddle.static.data(
|
|
name='x6',
|
|
shape=[2, 4, 4],
|
|
dtype="float32",
|
|
)
|
|
self.assertRaises(ValueError, paddle.baddbmm, input, x5, x6)
|
|
# input and x are not broadcastable
|
|
x7 = paddle.static.data(
|
|
name='x7',
|
|
shape=[2, 4, 4],
|
|
dtype="float32",
|
|
)
|
|
x8 = paddle.static.data(
|
|
name='x8',
|
|
shape=[2, 4, 4],
|
|
dtype="float32",
|
|
)
|
|
input1 = paddle.static.data(
|
|
name='input1',
|
|
shape=[2, 2, 4],
|
|
dtype="float32",
|
|
)
|
|
self.assertRaises(ValueError, paddle.baddbmm, input1, x7, x8)
|
|
# input and x are not broadcastable
|
|
x9 = paddle.static.data(
|
|
name='x9',
|
|
shape=[2, 4, 4],
|
|
dtype="float32",
|
|
)
|
|
x10 = paddle.static.data(
|
|
name='x10',
|
|
shape=[2, 4, 4],
|
|
dtype="float32",
|
|
)
|
|
input2 = paddle.static.data(
|
|
name='input2',
|
|
shape=[2, 1, 2],
|
|
dtype="float32",
|
|
)
|
|
self.assertRaises(ValueError, paddle.baddbmm, input2, x9, x10)
|
|
x11 = paddle.static.data(
|
|
name='x11',
|
|
shape=[2, 4, 4],
|
|
dtype="float32",
|
|
)
|
|
x12 = paddle.static.data(
|
|
name='x12', shape=[2, 4, 4], dtype="float32"
|
|
)
|
|
input3 = paddle.static.data(
|
|
name='input3',
|
|
shape=[2, 4, 2],
|
|
dtype="float32",
|
|
)
|
|
self.assertRaises(ValueError, paddle.baddbmm, input3, x11, x12)
|
|
x13 = paddle.static.data(
|
|
name='x13',
|
|
shape=[2, 4, 4],
|
|
dtype="float32",
|
|
)
|
|
x14 = paddle.static.data(
|
|
name='x14',
|
|
shape=[2, 4, 4],
|
|
dtype="float32",
|
|
)
|
|
input4 = paddle.static.data(
|
|
name='input4',
|
|
shape=[2, 3, 1],
|
|
dtype="float32",
|
|
)
|
|
self.assertRaises(ValueError, paddle.baddbmm, input4, x13, x14)
|
|
|
|
|
|
class TestBaddBmmOp2(TestBaddBmmOp):
|
|
# test alpha and beta
|
|
def setUp(self):
|
|
self.op_type = "baddbmm"
|
|
self.prim_op_type = "comp"
|
|
self.python_api = paddle.baddbmm
|
|
self.public_python_api = paddle.baddbmm
|
|
self.dtype = np.float64
|
|
self.init_dtype_type()
|
|
self.inputs = {
|
|
'Input': np.random.random((2, 10, 5)).astype(self.dtype),
|
|
'X': np.random.random((2, 10, 10)).astype(self.dtype),
|
|
'Y': np.random.random((2, 10, 5)).astype(self.dtype),
|
|
}
|
|
self.attrs = {
|
|
'Alpha': 0.1,
|
|
'Beta': 1.0,
|
|
}
|
|
self.outputs = {
|
|
'Out': self.attrs['Beta'] * self.inputs['Input']
|
|
+ self.attrs['Alpha']
|
|
* np.matmul(self.inputs['X'], self.inputs['Y'])
|
|
}
|
|
|
|
|
|
class TestBaddBmmOp3(OpTest):
|
|
def setUp(self):
|
|
self.op_type = "baddbmm"
|
|
self.prim_op_type = "comp"
|
|
self.python_api = paddle.baddbmm
|
|
self.public_python_api = paddle.baddbmm
|
|
self.dtype = np.float64
|
|
self.init_dtype_type()
|
|
self.inputs = {
|
|
'Input': np.random.random((2, 10, 5)).astype(self.dtype),
|
|
'X': np.random.random((2, 10, 10)).astype(self.dtype),
|
|
'Y': np.random.random((2, 10, 5)).astype(self.dtype),
|
|
}
|
|
self.attrs = {
|
|
'Alpha': 0.5,
|
|
'Beta': 2.0,
|
|
}
|
|
self.outputs = {
|
|
'Out': self.attrs['Beta'] * self.inputs['Input']
|
|
+ self.attrs['Alpha']
|
|
* np.matmul(self.inputs['X'], self.inputs['Y'])
|
|
}
|
|
|
|
def init_dtype_type(self):
|
|
pass
|
|
|
|
def test_check_output(self):
|
|
self.check_output(check_pir=True, check_prim_pir=True)
|
|
|
|
def test_check_grad_normal(self):
|
|
self.check_grad(
|
|
['Input', 'X', 'Y'], 'Out', check_pir=True, check_prim_pir=True
|
|
)
|
|
|
|
def test_check_grad_x(self):
|
|
self.check_grad(
|
|
['X'], 'Out', no_grad_set=None, check_pir=True, check_prim_pir=True
|
|
)
|
|
|
|
def test_check_grad_y(self):
|
|
self.check_grad(
|
|
['Y'], 'Out', no_grad_set=None, check_pir=True, check_prim_pir=True
|
|
)
|
|
|
|
def test_check_grad_input(self):
|
|
self.check_grad(
|
|
['Input'],
|
|
'Out',
|
|
no_grad_set=None,
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
|
|
class TestBaddBmmOp4(OpTest):
|
|
# test broadcast
|
|
def setUp(self):
|
|
self.op_type = "baddbmm"
|
|
self.prim_op_type = "comp"
|
|
self.python_api = paddle.baddbmm
|
|
self.public_python_api = paddle.baddbmm
|
|
self.dtype = np.float64
|
|
self.init_dtype_type()
|
|
self.inputs = {
|
|
'Input': np.random.random((1, 15)).astype(self.dtype),
|
|
'X': np.random.random((1, 50, 10)).astype(self.dtype),
|
|
'Y': np.random.random((1, 10, 15)).astype(self.dtype),
|
|
}
|
|
self.attrs = {
|
|
'Alpha': 0.5,
|
|
'Beta': 2.0,
|
|
}
|
|
|
|
self.outputs = {
|
|
'Out': self.attrs['Beta']
|
|
* np.broadcast_to(
|
|
self.inputs['Input'][:, np.newaxis, :], (1, 50, 15)
|
|
)
|
|
+ self.attrs['Alpha']
|
|
* np.matmul(self.inputs['X'], self.inputs['Y'])
|
|
}
|
|
|
|
def init_dtype_type(self):
|
|
pass
|
|
|
|
def test_check_output(self):
|
|
self.check_output(check_pir=True, check_prim_pir=True)
|
|
|
|
def test_check_grad_normal(self):
|
|
self.inputs['Input'] = np.broadcast_to(
|
|
self.inputs['Input'][:, np.newaxis, :], (1, 50, 15)
|
|
)
|
|
self.check_grad(
|
|
['Input', 'X', 'Y'], 'Out', check_pir=True, check_prim_pir=True
|
|
)
|
|
|
|
def test_check_grad_x(self):
|
|
self.check_grad(['X'], 'Out', no_grad_set=None, check_pir=True)
|
|
|
|
def test_check_grad_y(self):
|
|
self.check_grad(['Y'], 'Out', no_grad_set=None, check_pir=True)
|
|
|
|
def test_check_grad_input(self):
|
|
self.inputs['Input'] = np.broadcast_to(
|
|
self.inputs['Input'][:, np.newaxis, :], (1, 50, 15)
|
|
)
|
|
self.check_grad(
|
|
['Input'],
|
|
'Out',
|
|
no_grad_set=None,
|
|
check_pir=True,
|
|
)
|
|
|
|
|
|
class TestBaddBmmAPI(unittest.TestCase):
|
|
def test_api_error(self):
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
|
|
paddle.disable_static()
|
|
|
|
def test_error1():
|
|
data_x_wrong = np.ones((2, 2, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x_wrong)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error1)
|
|
|
|
def test_error2():
|
|
data_x_wrong = np.ones((2, 2)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x_wrong)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error2)
|
|
|
|
def test_error3():
|
|
data_input_wrong = np.ones((2, 2, 2, 2)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error3)
|
|
|
|
def test_error4():
|
|
data_input_wrong = np.ones((2, 5)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error4)
|
|
|
|
def test_error5():
|
|
data_input_wrong = np.ones((3, 2)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error5)
|
|
|
|
def test_error6():
|
|
data_input_wrong = np.ones((3, 2, 1)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error6)
|
|
|
|
def test_error7():
|
|
data_input_wrong = np.ones((1, 2, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error7)
|
|
|
|
def test_error_y1():
|
|
data_y_wrong = np.ones((2, 2, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y_wrong)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error_y1)
|
|
|
|
def test_error_y2():
|
|
data_y_wrong = np.ones((2, 2)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y_wrong)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error_y2)
|
|
|
|
def test_error_y3():
|
|
data_y_wrong = np.ones((1, 2, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y_wrong)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=0.5, alpha=5.0
|
|
)
|
|
|
|
self.assertRaises(ValueError, test_error_y3)
|
|
|
|
paddle.enable_static()
|
|
|
|
def test_api_normal_1(self):
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_alpha = 0.1
|
|
data_beta = 1.0
|
|
|
|
paddle.disable_static()
|
|
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
paddle_output = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=data_beta, alpha=data_alpha
|
|
)
|
|
numpy_output = data_beta * data_input + data_alpha * np.matmul(
|
|
data_x, data_y
|
|
)
|
|
|
|
np.testing.assert_allclose(
|
|
numpy_output, paddle_output.numpy(), rtol=1e-05
|
|
)
|
|
|
|
paddle.enable_static()
|
|
|
|
def test_api_out(self):
|
|
if in_pir_mode():
|
|
self.skipTest("PIR not support out tensor")
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_alpha = 0.1
|
|
data_beta = 1.0
|
|
|
|
paddle.disable_static()
|
|
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.zeros((2, 2, 2), dtype='float32')
|
|
paddle_output = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=data_beta, alpha=data_alpha, out=out
|
|
)
|
|
numpy_output = data_beta * data_input + data_alpha * np.matmul(
|
|
data_x, data_y
|
|
)
|
|
|
|
# Check that the returned tensor is the same as the out tensor
|
|
self.assertIs(paddle_output, out)
|
|
# Check that the values are correct
|
|
np.testing.assert_allclose(numpy_output, out.numpy(), rtol=1e-05)
|
|
|
|
paddle.enable_static()
|
|
|
|
def test_api_alias(self):
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_alpha = 0.1
|
|
data_beta = 1.0
|
|
|
|
paddle.disable_static()
|
|
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
|
|
# Test using original parameter names
|
|
paddle_output_original = paddle.tensor.baddbmm(
|
|
input=input, x=x, y=y, beta=data_beta, alpha=data_alpha
|
|
)
|
|
|
|
# Test using aliases
|
|
paddle_output_alias = paddle.tensor.baddbmm(
|
|
input=input, batch1=x, batch2=y, beta=data_beta, alpha=data_alpha
|
|
)
|
|
|
|
# Check that both outputs are the same
|
|
np.testing.assert_allclose(
|
|
paddle_output_original.numpy(),
|
|
paddle_output_alias.numpy(),
|
|
rtol=1e-05,
|
|
)
|
|
|
|
paddle.enable_static()
|
|
|
|
def test_api_out_dtype(self):
|
|
"""Test out_dtype parameter for baddbmm"""
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_alpha = 0.1
|
|
data_beta = 1.0
|
|
|
|
paddle.disable_static()
|
|
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
|
|
# Test with out_dtype=float64
|
|
paddle_output = paddle.tensor.baddbmm(
|
|
input=input,
|
|
x=x,
|
|
y=y,
|
|
beta=data_beta,
|
|
alpha=data_alpha,
|
|
out_dtype=paddle.float64,
|
|
)
|
|
numpy_output = data_beta * data_input + data_alpha * np.matmul(
|
|
data_x, data_y
|
|
)
|
|
|
|
# Check that output dtype is float64
|
|
self.assertEqual(paddle_output.dtype, paddle.float64)
|
|
# Check that the values are correct
|
|
np.testing.assert_allclose(
|
|
numpy_output, paddle_output.numpy(), rtol=1e-05
|
|
)
|
|
|
|
# Test with out_dtype=None (should use input dtype)
|
|
paddle_output_none = paddle.tensor.baddbmm(
|
|
input=input,
|
|
x=x,
|
|
y=y,
|
|
beta=data_beta,
|
|
alpha=data_alpha,
|
|
out_dtype=None,
|
|
)
|
|
self.assertEqual(paddle_output_none.dtype, paddle.float32)
|
|
|
|
paddle.enable_static()
|
|
|
|
def test_api_out_dtype_fp16(self):
|
|
"""Test out_dtype parameter with float16"""
|
|
if not (core.is_compiled_with_cuda() or is_custom_device()):
|
|
self.skipTest("CUDA is not available")
|
|
if not core.is_float16_supported(paddle.CUDAPlace(0)):
|
|
self.skipTest("Float16 is not supported")
|
|
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_alpha = 0.5
|
|
data_beta = 1.0
|
|
|
|
paddle.disable_static()
|
|
paddle.set_device('gpu')
|
|
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
|
|
# Test with out_dtype=float16
|
|
paddle_output = paddle.tensor.baddbmm(
|
|
input=input,
|
|
x=x,
|
|
y=y,
|
|
beta=data_beta,
|
|
alpha=data_alpha,
|
|
out_dtype=paddle.float16,
|
|
)
|
|
|
|
# Check that output dtype is float16
|
|
self.assertEqual(paddle_output.dtype, paddle.float16)
|
|
|
|
numpy_output = data_beta * data_input + data_alpha * np.matmul(
|
|
data_x, data_y
|
|
)
|
|
np.testing.assert_allclose(
|
|
numpy_output, paddle_output.numpy(), rtol=1e-02, atol=1e-02
|
|
)
|
|
|
|
paddle.enable_static()
|
|
|
|
|
|
class TestBaddBmmBatch1Op(OpTest):
|
|
# test basic
|
|
def setUp(self):
|
|
self.op_type = "baddbmm"
|
|
self.prim_op_type = "comp"
|
|
self.python_api = paddle.baddbmm
|
|
self.public_python_api = paddle.baddbmm
|
|
self.init_dtype_type()
|
|
self.inputs = {
|
|
'Input': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
'X': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
'Y': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
}
|
|
self.outputs = {
|
|
'Out': self.inputs['Input']
|
|
+ np.matmul(self.inputs['X'], self.inputs['Y'])
|
|
}
|
|
|
|
def init_dtype_type(self):
|
|
self.dtype = np.float64
|
|
|
|
def test_check_output(self):
|
|
self.check_output(check_pir=True, check_prim_pir=True)
|
|
|
|
def test_check_grad_normal(self):
|
|
self.check_grad(
|
|
['Input', 'X', 'Y'],
|
|
'Out',
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
def test_check_grad_x(self):
|
|
self.check_grad(
|
|
['X'],
|
|
'Out',
|
|
no_grad_set=None,
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
def test_check_grad_y(self):
|
|
self.check_grad(
|
|
['Y'],
|
|
'Out',
|
|
no_grad_set=None,
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
def test_check_grad_input(self):
|
|
self.check_grad(
|
|
['Input'],
|
|
'Out',
|
|
no_grad_set=None,
|
|
check_pir=True,
|
|
check_prim_pir=True,
|
|
)
|
|
|
|
|
|
class TestBaddBmmBatch1FP16Op(TestBaddBmmBatch1Op):
|
|
def init_dtype_type(self):
|
|
self.dtype = np.float16
|
|
|
|
def test_check_output(self):
|
|
self.check_output(atol=1e-2)
|
|
|
|
|
|
class TestBaddBmmBatch1Op2(TestBaddBmmBatch1Op):
|
|
# test alpha and beta
|
|
def setUp(self):
|
|
self.op_type = "baddbmm"
|
|
self.prim_op_type = "comp"
|
|
self.python_api = paddle.baddbmm
|
|
self.public_python_api = paddle.baddbmm
|
|
self.dtype = np.float64
|
|
self.init_dtype_type()
|
|
self.inputs = {
|
|
'Input': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
'X': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
'Y': np.random.random((1, 10, 10)).astype(self.dtype),
|
|
}
|
|
self.attrs = {
|
|
'Alpha': 0.1,
|
|
'Beta': 1.0,
|
|
}
|
|
self.outputs = {
|
|
'Out': self.attrs['Beta'] * self.inputs['Input']
|
|
+ self.attrs['Alpha']
|
|
* np.matmul(self.inputs['X'], self.inputs['Y'])
|
|
}
|
|
|
|
|
|
class TestBaddBmmUnderlineAPI(unittest.TestCase):
|
|
def test_api_error(self):
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
|
|
paddle.disable_static()
|
|
|
|
def test_error1():
|
|
data_x_wrong = np.ones((2, 2, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x_wrong)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error1)
|
|
|
|
def test_error2():
|
|
data_x_wrong = np.ones((2, 2)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x_wrong)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error2)
|
|
|
|
def test_error3():
|
|
data_input_wrong = np.ones((2, 2, 2, 2)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error3)
|
|
|
|
def test_error4():
|
|
data_input_wrong = np.ones((2, 5)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error4)
|
|
|
|
def test_error5():
|
|
data_input_wrong = np.ones((3, 2)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error5)
|
|
|
|
def test_error6():
|
|
data_input_wrong = np.ones((3, 2, 1)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error6)
|
|
|
|
def test_error7():
|
|
data_input_wrong = np.ones((1, 2, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error7)
|
|
|
|
def test_error8():
|
|
data_input_wrong = np.ones((2, 3, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error8)
|
|
|
|
def test_error9():
|
|
data_input_wrong = np.ones((2, 1, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input_wrong)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error9)
|
|
|
|
def test_error_y1():
|
|
data_y_wrong = np.ones((2, 2, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y_wrong)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error_y1)
|
|
|
|
def test_error_y2():
|
|
data_y_wrong = np.ones((2, 2)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y_wrong)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error_y2)
|
|
|
|
def test_error_y3():
|
|
data_y_wrong = np.ones((1, 2, 3)).astype(np.float32)
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y_wrong)
|
|
input = paddle.to_tensor(data_input)
|
|
out = paddle.baddbmm_(input=input, x=x, y=y, beta=0.5, alpha=5.0)
|
|
|
|
self.assertRaises(ValueError, test_error_y3)
|
|
|
|
paddle.enable_static()
|
|
|
|
def test_api_normal_1(self):
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_alpha = 0.1
|
|
data_beta = 1.0
|
|
|
|
paddle.disable_static()
|
|
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
|
|
numpy_output = data_beta * data_input + data_alpha * np.matmul(
|
|
data_x, data_y
|
|
)
|
|
|
|
paddle_output = paddle.baddbmm_(
|
|
input=input, x=x, y=y, beta=data_beta, alpha=data_alpha
|
|
)
|
|
|
|
np.testing.assert_allclose(
|
|
numpy_output, paddle_output.numpy(), rtol=1e-05
|
|
)
|
|
|
|
paddle.enable_static()
|
|
|
|
def test_api_alias(self):
|
|
data_x = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_y = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_input = np.ones((2, 2, 2)).astype(np.float32)
|
|
data_alpha = 0.1
|
|
data_beta = 1.0
|
|
|
|
paddle.disable_static()
|
|
|
|
x = paddle.to_tensor(data_x)
|
|
y = paddle.to_tensor(data_y)
|
|
input = paddle.to_tensor(data_input)
|
|
|
|
# Test using original parameter names
|
|
paddle_output_original = paddle.baddbmm_(
|
|
input=input.clone(), x=x, y=y, beta=data_beta, alpha=data_alpha
|
|
)
|
|
|
|
# Test using aliases
|
|
paddle_output_alias = paddle.baddbmm_(
|
|
input=input.clone(),
|
|
batch1=x,
|
|
batch2=y,
|
|
beta=data_beta,
|
|
alpha=data_alpha,
|
|
)
|
|
|
|
# Check that both outputs are the same
|
|
np.testing.assert_allclose(
|
|
paddle_output_original.numpy(),
|
|
paddle_output_alias.numpy(),
|
|
rtol=1e-05,
|
|
)
|
|
|
|
paddle.enable_static()
|
|
|
|
|
|
if __name__ == "__main__":
|
|
paddle.enable_static()
|
|
unittest.main()
|