Skip to content

Commit

Permalink
consistent formatting
Browse files Browse the repository at this point in the history
  • Loading branch information
@csteppa
csteppa committed Feb 9, 2019
1 parent 356d501 commit 5fecd30
Show file tree
Hide file tree
Showing 8 changed files with 2,190 additions and 1,268 deletions.
699 changes: 461 additions & 238 deletions src/hexagdly.py
View file

Large diffs are not rendered by default.

394 changes: 230 additions & 164 deletions tests/test_Conv2d.py
View file

Large diffs are not rendered by default.

392 changes: 226 additions & 166 deletions tests/test_Conv2d_Custom.py
View file

Large diffs are not rendered by default.

409 changes: 238 additions & 171 deletions tests/test_Conv2d_CustomKernel.py
View file

Large diffs are not rendered by default.

474 changes: 323 additions & 151 deletions tests/test_Conv3d.py
View file

Large diffs are not rendered by default.

487 changes: 328 additions & 159 deletions tests/test_Conv3d_CustomKernel.py
View file

Large diffs are not rendered by default.

170 changes: 101 additions & 69 deletions tests/test_MaxPool2d.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,51 +3,77 @@
import hexagdly as hex
import pytest


class TestMaxPool2d(object):
def get_array(self):
return np.array([[j*5+1+i for j in range(8)] for i in range(5)], dtype=np.float32)
return np.array(
[[j * 5 + 1 + i for j in range(8)] for i in range(5)], dtype=np.float32
)

def get_array_maxpool2d_size1_stride1(self):
return np.array([[6,12,16,22,26,32,36,37],
[7,13,17,23,27,33,37,38],
[8,14,18,24,28,34,38,39],
[9,15,19,25,29,35,39,40],
[10,15,20,25,30,35,40,40]], dtype=np.float32)
return np.array(
[
[6, 12, 16, 22, 26, 32, 36, 37],
[7, 13, 17, 23, 27, 33, 37, 38],
[8, 14, 18, 24, 28, 34, 38, 39],
[9, 15, 19, 25, 29, 35, 39, 40],
[10, 15, 20, 25, 30, 35, 40, 40],
],
dtype=np.float32,
)

def get_array_maxpool2d_size2_stride1(self):
return np.array([[12,17,22,27,32,37,37,38],
[13,18,23,28,33,38,38,39],
[14,19,24,29,34,39,39,40],
[15,20,25,30,35,40,40,40],
[15,20,25,30,35,40,40,40]], dtype=np.float32)
return np.array(
[
[12, 17, 22, 27, 32, 37, 37, 38],
[13, 18, 23, 28, 33, 38, 38, 39],
[14, 19, 24, 29, 34, 39, 39, 40],
[15, 20, 25, 30, 35, 40, 40, 40],
[15, 20, 25, 30, 35, 40, 40, 40],
],
dtype=np.float32,
)

def get_array_stride_2(self, array_stride_1):
array_stride_2 = np.zeros((2,4), dtype=np.float32)
stride_2_pos = [(0,0,0,0), (0,1,1,2), (0,2,0,4), (0,3,1,6),
(1,0,2,0), (1,1,3,2), (1,2,2,4), (1,3,3,6)]
array_stride_2 = np.zeros((2, 4), dtype=np.float32)
stride_2_pos = [
(0, 0, 0, 0),
(0, 1, 1, 2),
(0, 2, 0, 4),
(0, 3, 1, 6),
(1, 0, 2, 0),
(1, 1, 3, 2),
(1, 2, 2, 4),
(1, 3, 3, 6),
]
for pos in stride_2_pos:
array_stride_2[pos[0],pos[1]] = array_stride_1[pos[2],pos[3]]
array_stride_2[pos[0], pos[1]] = array_stride_1[pos[2], pos[3]]
return array_stride_2

def get_array_stride_3(self, array_stride_1):
array_stride_3 = np.zeros((2,3), dtype=np.float32)
stride_3_pos = [(0,0,0,0), (0,1,1,3), (0,2,0,6),
(1,0,3,0), (1,1,4,3), (1,2,3,6)]
array_stride_3 = np.zeros((2, 3), dtype=np.float32)
stride_3_pos = [
(0, 0, 0, 0),
(0, 1, 1, 3),
(0, 2, 0, 6),
(1, 0, 3, 0),
(1, 1, 4, 3),
(1, 2, 3, 6),
]
for pos in stride_3_pos:
array_stride_3[pos[0],pos[1]] = array_stride_1[pos[2],pos[3]]
array_stride_3[pos[0], pos[1]] = array_stride_1[pos[2], pos[3]]
return array_stride_3



def get_tensors(self, in_channels, kernel_size, stride):
channel_dist = 1000

# input tensor
array = self.get_array()
array = np.expand_dims(np.stack([j * channel_dist + array
for j in range(in_channels)]),
0)
array = np.expand_dims(
np.stack([j * channel_dist + array for j in range(in_channels)]), 0
)
tensor = torch.FloatTensor(array)

# expected output tensor
if kernel_size == 1:
pooled_array = self.get_array_maxpool2d_size1_stride1()
Expand All @@ -57,120 +83,126 @@ def get_tensors(self, in_channels, kernel_size, stride):
pooled_array = self.get_array_stride_2(pooled_array)
elif stride == 3:
pooled_array = self.get_array_stride_3(pooled_array)
pooled_array = np.expand_dims(np.stack([channel * channel_dist + pooled_array
for channel in range(in_channels)]),
0)
pooled_array = np.expand_dims(
np.stack(
[
channel * channel_dist + pooled_array
for channel in range(in_channels)
]
),
0,
)
pooled_tensor = torch.FloatTensor(pooled_array)

# output tensor of test method
maxpool2d = hex.MaxPool2d(kernel_size,stride)
maxpool2d = hex.MaxPool2d(kernel_size, stride)

return maxpool2d(tensor), pooled_tensor

def test_in_channels_1_kernel_size_1_stride_1(self):
in_channels = 1
kernel_size = 1
stride = 1

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_1_kernel_size_1_stride_2(self):
in_channels = 1
kernel_size = 1
stride = 2

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_1_kernel_size_1_stride_3(self):
in_channels = 1
kernel_size = 1
stride = 3

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_1_kernel_size_2_stride_1(self):
in_channels = 1
kernel_size = 2
stride = 1

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_1_kernel_size_2_stride_2(self):
in_channels = 1
kernel_size = 2
stride = 2

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_1_kernel_size_2_stride_3(self):
in_channels = 1
kernel_size = 2
stride = 3

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_5_kernel_size_1_stride_1(self):
in_channels = 5
kernel_size = 1
stride = 1

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_5_kernel_size_1_stride_2(self):
in_channels = 5
kernel_size = 1
stride = 2

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_5_kernel_size_1_stride_3(self):
in_channels = 5
kernel_size = 1
stride = 3

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_5_kernel_size_2_stride_1(self):
in_channels = 5
kernel_size = 2
stride = 1

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_5_kernel_size_2_stride_2(self):
in_channels = 5
kernel_size = 2
stride = 2

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)

assert torch.equal(test_ouput, expectation)

def test_in_channels_5_kernel_size_2_stride_3(self):
in_channels = 5
kernel_size = 2
stride = 3

test_ouput, expectation = self.get_tensors(in_channels, kernel_size, stride)
assert torch.equal(test_ouput, expectation)

assert torch.equal(test_ouput, expectation)
Loading

0 comments on commit 5fecd30

Please sign in to comment.