Skip to content

Commit 1233745

Browse files
kawakami-masaki0kawakami-masaki0
committed
Modified test for 1d tensor
1 parent b4f2543 commit 1233745

File tree

1 file changed

+38
-39
lines changed

1 file changed

+38
-39
lines changed

‎tests_pytest/pytorch_tests/e2e_tests/test_quantization_for_1d_tensor.py‎

Lines changed: 38 additions & 39 deletions
Original file line numberDiff line numberDiff line change
@@ -27,78 +27,77 @@ def __init__(self, name):
2727
self.name = name
2828
self.conv = nn.Conv2d(3, 3, kernel_size=3, padding=1)
2929
self.relu = nn.ReLU()
30-
self.tensor = nn.Parameter(2.0 * torch.ones([192])) # 1D tensor
3130

3231
def forward(self, x):
3332
x = self.conv(x)
3433
x = self.relu(x)
3534
x = torch.reshape(x, (-1,))
3635

3736
if self.name == 'add':
38-
const = torch.add(self.tensor, 1)
37+
y = torch.add(x, 1)
3938
elif self.name == 'relu6':
40-
const = torch.nn.functional.relu6(self.tensor)
39+
y = torch.nn.functional.relu6(x)
4140
elif self.name == 'relu':
42-
const = torch.nn.functional.relu(self.tensor)
41+
y = torch.nn.functional.relu(x)
4342
elif self.name == 'sigmoid':
44-
const = torch.nn.functional.sigmoid(self.tensor)
43+
y = torch.nn.functional.sigmoid(x)
4544
elif self.name == 'leaky_relu':
46-
const = torch.nn.functional.leaky_relu(self.tensor)
45+
y = torch.nn.functional.leaky_relu(x)
4746
elif self.name == 'mul':
48-
const = torch.mul(self.tensor, 1)
47+
y = torch.mul(x, 1)
4948
elif self.name == 'sub':
50-
const = torch.sub(self.tensor, 1)
49+
y = torch.sub(x, 1)
5150
elif self.name == 'div':
52-
const = torch.div(self.tensor, 1)
51+
y = torch.div(x, 1)
5352
elif self.name == 'softmax':
54-
const = torch.nn.functional.softmax(self.tensor)
53+
y = torch.nn.functional.softmax(x)
5554
elif self.name == 'tanh':
56-
const = torch.nn.functional.tanh(self.tensor)
55+
y = torch.nn.functional.tanh(x)
5756
elif self.name == 'negative':
58-
const = torch.negative(self.tensor)
57+
y = torch.negative(x)
5958
elif self.name == 'abs':
60-
const = torch.abs(self.tensor)
59+
y = torch.abs(x)
6160
elif self.name == 'sqrt':
62-
const = torch.sqrt(self.tensor)
61+
y = torch.sqrt(torch.clamp(x, min=1e-6))
6362
elif self.name == 'rsqrt':
64-
const = torch.rsqrt(self.tensor)
63+
y = torch.rsqrt(torch.clamp(x, min=1e-6))
6564
elif self.name == 'silu':
66-
const = torch.nn.functional.silu(self.tensor)
65+
y = torch.nn.functional.silu(x)
6766
elif self.name == 'hardswish':
68-
const = torch.nn.functional.hardswish(self.tensor)
67+
y = torch.nn.functional.hardswish(x)
6968
elif self.name == 'hardsigmoid':
70-
const = torch.nn.functional.hardsigmoid(self.tensor)
69+
y = torch.nn.functional.hardsigmoid(x)
7170
elif self.name == 'pow':
72-
const = torch.pow(self.tensor, 1)
71+
y = torch.pow(x, 1)
7372
elif self.name == 'gelu':
74-
const = torch.nn.functional.gelu(self.tensor)
73+
y = torch.nn.functional.gelu(x)
7574
elif self.name == 'cos':
76-
const = torch.cos(self.tensor)
75+
y = torch.cos(x)
7776
elif self.name == 'sin':
78-
const = torch.sin(self.tensor)
77+
y = torch.sin(x)
7978
elif self.name == 'exp':
80-
const = torch.exp(self.tensor)
79+
y = torch.exp(x)
8180
elif self.name == 'mean':
82-
const = torch.mean(self.tensor, dim=0, keepdim=True)
81+
y = torch.mean(x, dim=0, keepdim=True)
8382
elif self.name == 'amax':
84-
const = torch.amax(self.tensor, dim=0, keepdim=True)
83+
y = torch.amax(x, dim=0, keepdim=True)
8584
elif self.name == 'maximum':
86-
const = torch.maximum(self.tensor, torch.tensor(0.0))
85+
y = torch.maximum(x, torch.tensor(0.0))
8786
elif self.name == 'minimum':
88-
const = torch.minimum(self.tensor, torch.tensor(0.0))
87+
y = torch.minimum(x, torch.tensor(0.0))
8988
elif self.name == 'sum':
90-
const = torch.sum(self.tensor, dim=0, keepdim=True)
91-
92-
y = x + const
89+
y = torch.sum(x, dim=0, keepdim=True)
90+
elif self.name == 'linalg_norm':
91+
y = torch.linalg.norm(x, dim=0, keepdim=True)
9392
return y
9493

9594
def representative_data_gen():
9695
yield [torch.randn(1, 3, 8, 8)]
9796

9897
@pytest.mark.parametrize("layer", [
99-
'add', 'relu6', 'relu', 'sigmoid', 'leaky_relu', 'mul', 'sub', 'div', 'softmax',
100-
'tanh', 'negative', 'abs', 'sqrt', 'rsqrt', 'silu', 'hardswish', 'hardsigmoid',
101-
'pow', 'gelu', 'cos', 'sin', 'exp', 'mean', 'amax', 'maximum', 'minimum', 'sum'
98+
'add', 'relu6', 'relu', 'sigmoid', 'leaky_relu', 'mul', 'sub', 'div', 'mean', 'amax', 'softmax',
99+
'tanh', 'negative', 'maximum', 'minimum', 'abs', 'sqrt', 'sum', 'rsqrt', 'silu', 'hardswish', 'hardsigmoid',
100+
'linalg_norm', 'pow', 'gelu', 'cos', 'sin', 'exp',
102101
])
103102
def test_ptq_1d_tensor(layer):
104103

@@ -118,9 +117,9 @@ def test_ptq_1d_tensor(layer):
118117

119118

120119
@pytest.mark.parametrize("layer", [
121-
'add', 'relu6', 'relu', 'sigmoid', 'leaky_relu', 'mul', 'sub', 'div', 'softmax',
122-
'tanh', 'negative', 'abs', 'sqrt', 'rsqrt', 'silu', 'hardswish', 'hardsigmoid',
123-
'pow', 'gelu', 'cos', 'sin', 'exp', 'mean', 'amax', 'maximum', 'minimum', 'sum'
120+
'add', 'relu6', 'relu', 'sigmoid', 'leaky_relu', 'mul', 'sub', 'div', 'mean', 'amax', 'softmax',
121+
'tanh', 'negative', 'maximum', 'minimum', 'abs', 'sqrt', 'sum', 'rsqrt', 'silu', 'hardswish', 'hardsigmoid',
122+
'linalg_norm', 'pow', 'gelu', 'cos', 'sin', 'exp',
124123
])
125124
def test_ptq_mixed_precision_1d_tensor(layer):
126125

@@ -149,9 +148,9 @@ def test_ptq_mixed_precision_1d_tensor(layer):
149148

150149

151150
@pytest.mark.parametrize("layer", [
152-
'add', 'relu6', 'relu', 'sigmoid', 'leaky_relu', 'mul', 'sub', 'div', 'softmax',
153-
'tanh', 'negative', 'abs', 'sqrt', 'sum', 'rsqrt', 'silu', 'hardswish', 'hardsigmoid',
154-
'pow', 'gelu', 'cos', 'sin', 'exp', 'mean', 'amax', 'maximum', 'minimum'
151+
'add', 'relu6', 'relu', 'sigmoid', 'leaky_relu', 'mul', 'sub', 'div', 'mean', 'amax', 'softmax',
152+
'tanh', 'negative', 'maximum', 'minimum', 'abs', 'sqrt', 'sum', 'rsqrt', 'silu', 'hardswish', 'hardsigmoid',
153+
'linalg_norm', 'pow', 'gelu', 'cos', 'sin', 'exp',
155154
])
156155
def test_gptq_1d_tensor(layer):
157156

0 commit comments

Comments
 (0)