def test_wrong_shape_values(self):
over_shape_values = [[32, 33]]
try:
iu.infer_shape_tensor(
self,
'plan',
np.float32,
over_shape_values, [[8, 4, 4]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
batch_size=8)
# InferenceServerException will be raised from different namespace,
# use dynamic type characteristic to catch both ex
except Exception as ex:
self.assertTrue(
"The shape value at index 2 is expected to be in range from 1 to 32, Got: 33"
in ex.message())
def test_wrong_shape_values(self):
over_shape_values = [[32, 33]]
try:
iu.infer_shape_tensor(
self,
'plan',
8,
np.float32,
over_shape_values, [[4, 4]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
use_cuda_shared_memory=TEST_CUDA_SHARED_MEMORY)
except InferenceServerException as ex:
self.assertEqual("inference:0", ex.server_id())
self.assertTrue(
"The shape value at index 2 is expected to be in range from 1 to 32, Got: 33"
in ex.message())
def check_response(self,
bs,
thresholds,
shape_values,
dummy_input_shapes,
shm_region_names=None,
precreated_shm_regions=None,
shm_suffix=""):
try:
# Add batch size to shape as full shape is expected
for i in range(len(dummy_input_shapes)):
dummy_input_shapes[i] = [
bs,
] + dummy_input_shapes[i]
start_ms = int(round(time.time() * 1000))
iu.infer_shape_tensor(
self,
'plan',
np.float32,
shape_values,
dummy_input_shapes,
use_grpc=False,
use_streaming=False,
shm_suffix=shm_suffix,
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
use_cuda_shared_memory=TEST_CUDA_SHARED_MEMORY,
batch_size=bs)
end_ms = int(round(time.time() * 1000))
lt_ms = thresholds[0]
gt_ms = thresholds[1]
if lt_ms is not None:
self.assertTrue(
(end_ms - start_ms) < lt_ms,
"expected less than " + str(lt_ms) +
"ms response time, got " + str(end_ms - start_ms) + " ms")
if gt_ms is not None:
self.assertTrue(
(end_ms - start_ms) > gt_ms,
"expected greater than " + str(gt_ms) +
"ms response time, got " + str(end_ms - start_ms) + " ms")
except Exception as ex:
self.add_deferred_exception(ex)
def test_nobatch(self):
iu.infer_shape_tensor(
self,
'plan_nobatch',
np.float32, [[32, 32]], [[4, 4]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY)
iu.infer_shape_tensor(
self,
'plan_nobatch',
np.float32, [[4, 4]], [[32, 32]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY)
iu.infer_shape_tensor(
self,
'plan_nobatch',
np.float32, [[4, 4]], [[4, 4]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY)
def test_static_batch(self):
iu.infer_shape_tensor(
self,
'plan',
np.float32, [[32, 32]], [[8, 4, 4]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
batch_size=8)
iu.infer_shape_tensor(
self,
'plan',
np.float32, [[4, 4]], [[8, 32, 32]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
batch_size=8)
iu.infer_shape_tensor(
self,
'plan',
np.float32, [[4, 4]], [[8, 4, 4]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
batch_size=8)
def test_static_batch(self):
iu.infer_shape_tensor(
self,
'plan',
8,
np.float32, [[32, 32]], [[4, 4]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
use_cuda_shared_memory=TEST_CUDA_SHARED_MEMORY)
iu.infer_shape_tensor(
self,
'plan',
8,
np.float32, [[4, 4]], [[32, 32]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
use_cuda_shared_memory=TEST_CUDA_SHARED_MEMORY)
iu.infer_shape_tensor(
self,
'plan',
8,
np.float32, [[4, 4]], [[4, 4]],
use_system_shared_memory=TEST_SYSTEM_SHARED_MEMORY,
use_cuda_shared_memory=TEST_CUDA_SHARED_MEMORY)