def test_import_entrez(self):
TEST_RECORD_ID = "6273291"
TEST_RECORD_LABEL = "testRecord"
if internet_on():
import_reference_genome_from_ncbi(
self.test_project, TEST_RECORD_LABEL, TEST_RECORD_ID,
'genbank')
def update_net_status():
global net_status
if utils.internet_on():
net_status = "ON"
else:
net_status = "OFF"
return True
def test_run_pipeline__genbank_from_ncbi_with_spaces_in_label(self):
"""Tests the pipeline where the genome is imported from NCBI with
spaces in the name.
"""
if not internet_on():
return
MG1655_ACCESSION = 'NC_000913.3'
MG1655_LABEL = 'mg1655 look a space'
ref_genome = import_reference_genome_from_ncbi(self.project,
MG1655_LABEL,
MG1655_ACCESSION,
'genbank')
sample_list = [self.experiment_sample]
alignment_group_obj, async_result = run_pipeline(
'name_placeholder', ref_genome, sample_list)
# Block until pipeline finishes.
while not async_result.ready():
time.sleep(1)
if async_result.status == 'FAILURE':
self.fail('Async task failed.')
alignment_group_obj = AlignmentGroup.objects.get(
id=alignment_group_obj.id)
self.assertEqual(
1, len(alignment_group_obj.experimentsampletoalignment_set.all()))
self.assertEqual(AlignmentGroup.STATUS.COMPLETED,
alignment_group_obj.status)
def test_import_entrez(self):
TEST_RECORD_ID = "6273291"
TEST_RECORD_LABEL = "testRecord"
if internet_on():
import_reference_genome_from_ncbi(
self.test_project, TEST_RECORD_LABEL, TEST_RECORD_ID,
'genbank')
def wait_internet_conn():
if internet_on():
GPIOSetup.ready()
return
time.sleep(1)
GPIOSetup.no_internet()
wait_internet_conn()
class TestIntegration(TestCase):
@unittest.skipIf(
not utils.internet_on(), "Cannot reach internet to download reference model."
)
def test_segmentation_pretrained_test_only(self):
t1 = torch.randn(3, 2, 2, requires_grad=True)
t2 = torch.randn(3, 2, 2, requires_grad=True)
tr1 = torch.randn(2, 2, requires_grad=True)
tr2 = torch.randn(2, 2, requires_grad=True)
model_name = "fcn_resnet101"
num_classes = 21
aux_loss = "store_true"
model = torchvision.models.segmentation.__dict__[model_name](
num_classes=num_classes, aux_loss=aux_loss, pretrained=True
)
model.eval()
# tensor run
t_input = torch.stack([t1, t2])
t_target = torch.stack([tr1, tr2])
confmat = ConfusionMatrix(num_classes)
output1 = model(t_input)
output1 = output1["out"]
confmat.update(t_target.flatten(), output1.argmax(1).flatten())
confmat.reduce_from_all_processes()
# nt run
nt_t1 = t1.clone().detach()
nt_t2 = t2.clone().detach()
nt_tr1 = tr1.clone().detach()
nt_tr2 = tr2.clone().detach()
nt_input = nestedtensor.nested_tensor([nt_t1, nt_t2], requires_grad=True)
nt_target = nestedtensor.nested_tensor([nt_tr1, nt_tr2], requires_grad=True)
confmat2 = ConfusionMatrix(num_classes)
output2 = model(nt_input)
output2 = output2["out"]
for a, b in zip(nt_target, output2):
confmat2.update(a.flatten(), b.argmax(0).flatten())
confmat2.reduce_from_all_processes()
self.assertEqual(confmat.mat, confmat2.mat)
# grad test
output1_sum = output1[0].sum()
output2_sum = output2[0].sum()
self.assertEqual(output1_sum, output2_sum)
output1_sum.backward()
output2_sum.backward()
self.assertEqual(t1.grad, nt_input[0].grad)
self.assertEqual(t2.grad, nt_input[1].grad)
def main():
while True:
initcode = input("Selector Barcode: ")
if utils.internet_on():
pass
if initcode == "QUIT":
utils.quit_log()
exit()
elif bool(initcode == "") | bool(initcode == " "):
print("Blank input. Try again.")
main()
elif initcode == "SHOP":
itemcode = get_barcode(initcode)
history.check_in(itemcode)
current.update_location(itemcode, initcode)
print(itemcode + " checked into SHOP.")
elif initcode == "RETIRE":
itemcode = get_barcode(initcode)
history.check_out(itemcode, initcode)
current.update_location(itemcode, initcode)
print(itemcode + " retired. Thank you for your service, " +
itemcode + ".")
elif initcode == "CHKLOC":
itemcode = get_barcode(initcode)
print(itemcode + " is checked out to " +
current.where_is(itemcode) + ".")
else:
itemcode = get_barcode(initcode)
if current.where_is(itemcode) == initcode:
print(itemcode + " is already checked out to " + initcode +
". Please scan another.")
main()
else:
history.check_out(itemcode, initcode)
current.update_location(itemcode, initcode)
print(itemcode + " checked out to " + initcode + ".")
class TestIntegration(TestCase):
@unittest.skipIf(
not utils.internet_on(), "Cannot reach internet to download reference model."
)
def test_segmentation_pretrained_test_only(self):
def _test(seed, model_factory, use_confmat, num_classes=21):
torch.manual_seed(seed)
t1 = torch.randn(3, 3, 4, requires_grad=True)
t2 = torch.randn(3, 3, 4, requires_grad=True)
tr1 = torch.randn(3, 4, requires_grad=True)
tr2 = torch.randn(3, 4, requires_grad=True)
model1 = model_factory()
# model1 = torchvision.models.segmentation.__dict__[model_name](
# num_classes=num_classes, aux_loss=aux_loss, pretrained=True
# )
# model1.eval()
# tensor run
t_input = torch.stack([t1, t2])
t_target = torch.stack([tr1, tr2])
if use_confmat:
confmat = ConfusionMatrix(num_classes)
output1 = model1(t_input)
if use_confmat:
output1 = output1["out"]
else:
output1 = output1["0"]
if use_confmat:
confmat.update(t_target.flatten(), output1.argmax(1).flatten())
confmat.reduce_from_all_processes()
# nt run
# model2 = torchvision.models.segmentation.__dict__[model_name](
# num_classes=num_classes, aux_loss=aux_loss, pretrained=True
# )
# model2.eval()
model2 = model_factory()
nt_t1 = t1.clone().detach()
nt_t2 = t2.clone().detach()
nt_tr1 = tr1.clone().detach()
nt_tr2 = tr2.clone().detach()
nt_input = nestedtensor.nested_tensor(
[nt_t1, nt_t2], requires_grad=True)
nt_target = nestedtensor.nested_tensor(
[nt_tr1, nt_tr2], requires_grad=True)
if use_confmat:
confmat2 = ConfusionMatrix(num_classes)
output2 = model2(nt_input)
if use_confmat:
output2 = output2["out"]
else:
output2 = output2["0"]
if use_confmat:
for a, b in zip(nt_target, output2):
confmat2.update(a.flatten(), b.argmax(0).flatten())
if use_confmat:
confmat2.reduce_from_all_processes()
self.assertEqual(confmat.mat, confmat2.mat)
# grad test
output1_sum = output1.sum()
output2_sum = output2.sum()
self.assertEqual(output1_sum, output2_sum)
output1_sum.backward()
output2_sum.backward()
for (n1, p1), (n2, p2) in zip(model1.named_parameters(), model2.named_parameters()):
if p1.grad is not None:
self.assertEqual(p1.grad, p2.grad)
else:
self.assertIsNone(p2.grad)
# TODO: Re-enable under autograd support
self.assertEqual(t1.grad, nt_input.grad[0])
self.assertEqual(t2.grad, nt_input.grad[1])
_test(1010, lambda: torchvision.models.segmentation.__dict__["fcn_resnet101"](
num_classes=21, aux_loss="store_true", pretrained=True
).eval(), True)