def c2_native_run_net(init_net, predict_net, inputs):
ws = Workspace()
if init_net:
ws.RunNetOnce(init_net)
if isinstance(inputs, dict):
for key, value in inputs.items():
ws.FeedBlob(key, value, predict_net.device_option)
else:
uninitialized = [
input_name for input_name in predict_net.external_input
if not ws.HasBlob(input_name)
]
if len(uninitialized) == len(inputs):
for key, value in zip(uninitialized, inputs):
ws.FeedBlob(key, value, predict_net.device_option)
else:
# If everything is initialized,
# we just initialized the first len(inputs) external_input.
assert (len(inputs) <= len(predict_net.external_input))
for i in range(len(inputs)):
ws.FeedBlob(predict_net.external_input[i], inputs[i],
predict_net.device_option)
ws.RunNetOnce(predict_net)
output_names = predict_net.external_output
output_values = [ws.FetchBlob(name) for name in output_names]
return ws, namedtupledict('Outputs', output_names)(*output_values)
def benchmark_caffe2_model(init_net, predict_net, warmup_iters=3, main_iters=10, layer_details=True):
'''
Run the benchmark net on the target model.
Return the execution time per iteration (millisecond).
'''
ws = Workspace()
if init_net:
ws.RunNetOnce(init_net)
ws.CreateNet(predict_net)
results = ws.BenchmarkNet(predict_net.name, warmup_iters, main_iters, layer_details)
del ws
return results[0]
def c2_native_run_net(init_net, predict_net, inputs, debug_arg=None):
ws = Workspace()
if init_net:
ws.RunNetOnce(init_net)
if isinstance(inputs, dict):
for key, value in inputs.items():
ws.FeedBlob(key, value, predict_net.device_option)
else:
uninitialized = [
input_name for input_name in predict_net.external_input
if not ws.HasBlob(input_name)
]
if len(uninitialized) == len(inputs):
for key, value in zip(uninitialized, inputs):
ws.FeedBlob(key, value, predict_net.device_option)
else:
# If everything is initialized,
# we just initialized the first len(inputs) external_input.
# Added some extra logging to help debug sporadic sandcastle fails
if len(inputs) > len(predict_net.external_input):
print("c2_native_run_net assert. len(inputs)=", len(inputs),
"len(predict_net.external_input)=",
len(predict_net.external_input))
print("debug_arg: ", debug_arg)
print("predict_net ", type(predict_net), ":", predict_net)
print("inputs ", type(inputs), ":", inputs)
assert (len(inputs) <= len(predict_net.external_input))
for i in range(len(inputs)):
ws.FeedBlob(predict_net.external_input[i], inputs[i],
predict_net.device_option)
ws.RunNetOnce(predict_net)
output_names = predict_net.external_output
output_values = [ws.FetchBlob(name) for name in output_names]
return ws, namedtupledict('Outputs', output_names)(*output_values)
def test_resnet50_core(self):
N = 2
warmup = 20
repeat = 100
print("Batch size: {}, repeat inference {} times, warmup {} times".
format(N, repeat, warmup))
init_net, pred_net, _ = self._get_c2_model('resnet50')
self._add_head_tail(pred_net, 'real_data', 'real_softmax')
input_blob_dims = (N, 3, 224, 224)
input_name = "real_data"
device_option = core.DeviceOption(caffe2_pb2.CUDA, 0)
init_net.device_option.CopyFrom(device_option)
pred_net.device_option.CopyFrom(device_option)
for op in pred_net.op:
op.device_option.CopyFrom(device_option)
op.engine = 'CUDNN'
net_outputs = pred_net.external_output
Y_c2 = None
data = np.random.randn(*input_blob_dims).astype(np.float32)
c2_time = 1
ws = Workspace()
with core.DeviceScope(device_option):
ws.FeedBlob(input_name, data)
ws.RunNetOnce(init_net)
ws.CreateNet(pred_net)
for _ in range(warmup):
ws.RunNet(pred_net.name)
start = time.time()
for _ in range(repeat):
ws.RunNet(pred_net.name)
end = time.time()
c2_time = end - start
output_values = [ws.FetchBlob(name) for name in net_outputs]
Y_c2 = namedtupledict('Outputs', net_outputs)(*output_values)
ws.ResetWorkspace()
# Cut the graph
init_net_cut, pred_net_cut = transform_caffe2_net(
init_net, pred_net, {input_name: input_blob_dims})
del init_net, pred_net
#print_net(pred_net_cut)
Y_trt = None
input_name = pred_net_cut.external_input[0]
print("C2 runtime: {}s".format(c2_time))
ws = Workspace()
with core.DeviceScope(device_option):
ws.FeedBlob(input_name, data)
ws.RunNetOnce(init_net_cut)
ws.CreateNet(pred_net_cut)
for _ in range(warmup):
ws.RunNet(pred_net_cut.name)
start = time.time()
for _ in range(repeat):
ws.RunNet(pred_net_cut.name)
end = time.time()
trt_time = end - start
print("TRT runtime: {}s, improvement: {}%".format(
trt_time, (c2_time - trt_time) / c2_time * 100))
output_values = [ws.FetchBlob(name) for name in net_outputs]
Y_trt = namedtupledict('Outputs', net_outputs)(*output_values)
np.testing.assert_allclose(Y_c2, Y_trt, rtol=1e-3)
