Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Adding test cases using another stride size. #2146

Open
wants to merge 1 commit into
base: develop
Choose a base branch
from
Open

Adding test cases using another stride size. #2146

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 20 additions & 7 deletions ci_test/unit_tests/test_unit_layer_convolution_distconv.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,19 @@ def num_samples():
def sample_dims():
return (_sample_size,)

# Convolution results computed using PyTorch beforehand.
def reference_value(num_dims, stride):
if num_dims == 2 and stride == 1:
return 11913.852660080756
elif num_dims == 2 and stride == 2:
return 2850.678372506634
elif num_dims == 3 and stride == 1:
return 9952.365297083174
elif num_dims == 3 and stride == 2:
return 1222.3680410607026

raise ValueError('Does not have a pre-computed reference value for this configuration')

# ==============================================
# PyTorch convolution
# ==============================================
Expand Down Expand Up @@ -165,12 +178,10 @@ def construct_model(lbann):
print('Skip - ' + e)
pytest.skip(e)

for num_dims, reference_val in [
(2, 11913.852660080756),
(3, 9952.365297083174)]:
for num_dims, s in [(2, 1), (2, 2), (3, 1), (3, 2)]:
# Convolution settings
kernel_dims = [5, _sample_dims[0] if num_dims == 2 else _sample_dims_3d[0],] + [3]*num_dims
strides = [1]*num_dims
strides = [s]*num_dims
pads = [1]*num_dims
dilations = [1]*num_dims
kernel = make_random_array(kernel_dims, 11)
Expand All @@ -179,7 +190,7 @@ def construct_model(lbann):
kernel_weights = lbann.Weights(
optimizer=lbann.SGD(),
initializer=lbann.ValueInitializer(values=np.nditer(kernel)),
name='kernel1_{}d'.format(num_dims)
name='kernel1_{}d-stride{}'.format(num_dims, s)
)
x = x_lbann
if num_dims == 3:
Expand All @@ -198,7 +209,9 @@ def construct_model(lbann):
num_height_groups))
z = lbann.L2Norm2(y)
obj.append(z)
metrics.append(lbann.Metric(z, name='basic {}D 3^n convolution'.format(num_dims)))
metrics.append(
lbann.Metric(z, name='basic {}D 3^n convolution with stride{}'.format(num_dims, s))
)

# PyTorch implementation
try:
Expand All @@ -214,7 +227,7 @@ def construct_model(lbann):
val = z
except:
# Precomputed value
val = reference_val
val = reference_value(num_dims, s)
# val = 398.6956458317758 # _num_samples=8, 6 channels
# val = 381.7401227915947 # _num_samples=23, 6 channels
tol = 8 * val * np.finfo(np.float32).eps
Expand Down