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 run_node(cls,
node,
inputs,
device='CPU',
opset_version=_known_opset_version):
super(Caffe2Backend, cls).run_node(node, inputs, device)
device_option = get_device_option(Device(device))
with Workspace(), core.DeviceScope(device_option): # temporary!
if isinstance(inputs, dict):
for key, value in inputs.items():
workspace.FeedBlob(key, value)
else:
assert len(node.input) == len(
inputs), "{}: expected {} but got {}".format(
node.op_type, len(node.input), len(inputs))
for key, value in zip(node.input, inputs):
workspace.FeedBlob(key, value)
cls._inplace_rewrite([node])
init_ops, ops, _ = cls._onnx_node_to_caffe2_op(
None, None, node, opset_version or cls._known_opset_version)
ops = init_ops + ops
for op in ops:
op.device_option.CopyFrom(device_option)
workspace.RunOperatorsOnce(ops)
output_values = [workspace.FetchBlob(name) for name in node.output]
return namedtupledict('Outputs', node.output)(*output_values)
def prepare(cls, predict_model, device='CPU',
init_model=None, **kwargs):
'''
For Onnx Caffe2Backend, we require that init_graph don't initialize the actual input of the predict_graph,
for example, if "img" is the input blob for the predict_net, we require that in init_graph and in
initializer of the predict_graph, "img" is not initalized. We don't have a check for this, since
there is no way we can know which blob is the input of the predict_graph.
'''
super(Caffe2Backend, cls).prepare(predict_model, device, **kwargs)
if init_model:
checker.check_model(init_model)
init_net, predict_net = cls.onnx_graph_to_caffe2_net(predict_model.graph)
predict_net.device_option.CopyFrom(get_device_option(Device(device)))
ws = Workspace()
with ws, core.DeviceScope(predict_net.device_option):
if init_model:
_, init_net_from_model = cls.onnx_graph_to_caffe2_net(init_model.graph)
init_net.op.extend(init_net_from_model.op)
workspace.RunNetOnce(init_net)
uninitialized = [x
for x in predict_net.external_input
if not workspace.HasBlob(x)]
return Caffe2Rep(predict_net, ws, uninitialized)
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 prepare(cls, model, device='CPU', **kwargs):
'''
For Onnx Caffe2Backend, we require that init_graph don't initialize the actual input of the predict_graph,
for example, if "img" is the input blob for the predict_net, we require that in init_graph and in
initializer of the predict_graph, "img" is not initalized. We don't have a check for this, since
there is no way we can know which blob is the input of the predict_graph.
'''
super(Caffe2Backend, cls).prepare(model, device, **kwargs)
opset_version = None
for imp in model.opset_import:
if not imp.HasField("domain") or imp.domain == "":
opset_version = imp.version
if imp.version > cls._known_opset_version:
warnings.warn(
"This version of onnx-caffe2 targets ONNX operator set version {}, but the model we are trying to import uses version {}. We will try to import it anyway, but if the model uses operators which had BC-breaking changes in the intervening versions, import will fail."
.format(cls._known_opset_version, imp.version))
else:
warnings.warn("Unrecognized operator set {}".format(
imp.domain))
if opset_version is None:
if model.ir_version >= 0x00000003:
raise RuntimeError(
"Model with IR version >= 3 did not specify ONNX operator set version (onnx-caffe2 requires it)"
)
else:
opset_version = 1
ws = Workspace()
device_option = get_device_option(Device(device))
# Directly load initializer data into blobs in workspace
cls._direct_initialize_parameters(
model.graph.initializer,
ws,
device_option,
)
initialized = {init.name for init in model.graph.initializer}
cls._direct_initialize_inputs(
model.graph.input,
initialized,
ws,
device_option,
)
uninitialized = [
value_info.name for value_info in model.graph.input
if value_info.name not in initialized
]
init_net, predict_net = cls._onnx_model_to_caffe2_net(
model, device, opset_version, False)
retval = Caffe2Rep(init_net, predict_net, ws, uninitialized)
return retval
def prepare(cls, model, device='CPU', **kwargs):
'''
For Onnx Caffe2Backend, we require that init_graph don't initialize the actual input of the predict_graph,
for example, if "img" is the input blob for the predict_net, we require that in init_graph and in
initializer of the predict_graph, "img" is not initalized. We don't have a check for this, since
there is no way we can know which blob is the input of the predict_graph.
'''
super(Caffe2Backend, cls).prepare(model, device, **kwargs)
ws = Workspace()
device_option = get_device_option(Device(device))
# Directly load initializer data into blobs in workspace
cls._direct_initialize_parameters(
model.graph.initializer,
ws,
device_option,
)
# Need to pull this out before we delete model.graph.initializer
initialized = {init.name for init in model.graph.initializer}
initializer = model.graph.initializer[:]
# Delete the initializers so they aren't serialized
del model.graph.initializer[:]
cls._direct_initialize_inputs(
model.graph.input,
initialized,
ws,
device_option,
)
# Pull this out to manually add external inputs
external_inputs = model.graph.input[:]
del model.graph.input[:]
_, predict_net = cls.onnx_graph_to_caffe2_net(model.graph)
predict_net.device_option.CopyFrom(device_option)
predict_net.external_input.extend(value_info.name
for value_info in external_inputs)
predict_net.device_option.CopyFrom(get_device_option(Device(device)))
# Restore these so as not to mutate input
model.graph.initializer.extend(initializer)
model.graph.input.extend(external_inputs)
uninitialized = [
x for x in predict_net.external_input if x not in initialized
]
retval = Caffe2Rep(predict_net, ws, uninitialized)
return retval
def c2_native_run_op(op_def, inputs):
with Workspace():
if isinstance(inputs, dict):
for key, value in inputs.items():
workspace.FeedBlob(key, value)
else:
assert (len(op_def.input) == len(inputs))
for key, value in zip(op_def.input, inputs):
workspace.FeedBlob(key, value)
workspace.RunOperatorOnce(op_def)
output_names = op_def.output
output_values = [workspace.FetchBlob(name) for name in output_names]
return namedtupledict('Outputs', output_names)(*output_values)
def run_node(cls, node, inputs, device='CPU'):
super(Caffe2Backend, cls).run_node(node, inputs, device)
device_option = get_device_option(Device(device))
with Workspace(), core.DeviceScope(device_option): # temporary!
if isinstance(inputs, dict):
for key, value in inputs.items():
workspace.FeedBlob(key, value)
else:
assert(len(node.input) == len(inputs))
for key, value in zip(node.input, inputs):
workspace.FeedBlob(key, value)
cls._inplace_rewrite([node])
ops = cls._onnx_node_to_caffe2_op(node)
for op in ops:
workspace.RunOperatorOnce(op)
output_values = [workspace.FetchBlob(name) for name in node.output]
return namedtupledict('Outputs', node.output)(*output_values)
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)
else:
uninitialized = [
input_name for input_name in predict_net.external_input
if not ws.HasBlob(input_name)
]
assert len(uninitialized) == len(inputs)
for key, value in zip(uninitialized, inputs):
ws.FeedBlob(key, value)
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 c2_native_run_op(op_def, inputs):
ws = Workspace()
if isinstance(inputs, dict):
for key, value in inputs.items():
ws.FeedBlob(key, value, op_def.device_option)
else:
assert (len(op_def.input) == len(inputs))
for key, value in zip(op_def.input, inputs):
ws.FeedBlob(key, value, op_def.device_option)
ws.RunOperatorOnce(op_def)
output_names = op_def.output
output_values = [ws.FetchBlob(name) for name in output_names]
return ws, namedtupledict('Outputs', output_names)(*output_values)
def prepare(cls, model, device='CPU', **kwargs):
'''
For Onnx Caffe2Backend, we require that init_graph don't initialize the actual input of the predict_graph,
for example, if "img" is the input blob for the predict_net, we require that in init_graph and in
initializer of the predict_graph, "img" is not initalized. We don't have a check for this, since
there is no way we can know which blob is the input of the predict_graph.
'''
super(Caffe2Backend, cls).prepare(model, device, **kwargs)
init_net, predict_net = cls.onnx_graph_to_caffe2_net(model.graph)
predict_net.device_option.CopyFrom(get_device_option(Device(device)))
initialized = {init.name for init in model.graph.initializer}
uninitialized = [x for x in predict_net.external_input
if x not in initialized]
ws = Workspace()
with ws, core.DeviceScope(predict_net.device_option):
workspace.RunNetOnce(init_net)
return Caffe2Rep(predict_net, ws, uninitialized)
