def _EagerRunModelSave(var_blobs, snapshot_path):
path_input_op_conf, path_lbi = _GenModelIOPathInputOpConfAndRetLbi()
path_input_blob_objects = {}
(
BuildModelIOPathInputInstruction,
BuildFeedPathInstruction,
) = _MakeModelIOPathInputBuilds(path_input_op_conf, snapshot_path,
path_input_blob_objects)
model_save_op_conf = _GenModelSaveOpConf(var_blobs, path_lbi)
model_save_blob_objects = {}
def BuildModelSaveInstruction(builder):
path_blob_object = path_input_blob_objects["out"]
model_save_blob_objects["path"] = path_blob_object
for i, blob in enumerate(var_blobs):
model_save_blob_objects["in_{}".format(i)] = blob.blob_object
op_attribute = op_infer_util.Infer(
model_save_op_conf, ibn2blob_object=model_save_blob_objects)
parallel_conf = path_blob_object.parallel_desc_symbol.parallel_conf
builder.StatelessCall(op_attribute,
parallel_conf,
bn_in_op2blob_object=model_save_blob_objects)
sess = session_ctx.GetDefaultSession()
with scope_util.ScopeContext(scope_util.MakeScope(_BuildNotMirroredScope)):
vm_util.LogicalRun(BuildModelIOPathInputInstruction)
vm_util.LogicalRun(BuildFeedPathInstruction)
vm_util.LogicalRun(BuildModelSaveInstruction)
def name_scope(name: str) -> None:
r"""Create a namespace. All variables within the namespace will have a prefix `[SCOPE NAME]-`. This is for convenience only and has no other effect on the system.
Usage::
with oneflow.scope.namespace("scope1"):
...
with oneflow.scope.namespace("scope2"):
...
Args:
name: Name of this namespace
"""
assert isinstance(name, str)
name_scope_stack_push(name)
def BuildScope(old_scope, builder):
return builder.BuildScopeWithNewScopeName(old_scope, name)
sess = session_context.GetDefaultSession()
try:
with scope_util.ScopeContext(scope_util.MakeScope(BuildScope)):
yield
finally:
name_scope_stack_pop()
def _EagerRunModelLoad(var_op_conf, snapshot_path):
assert isinstance(snapshot_path, str)
assert os.path.basename(snapshot_path) == "out"
snapshot_path = os.path.dirname(snapshot_path)
assert os.path.basename(snapshot_path) == var_op_conf.name
snapshot_path = os.path.dirname(snapshot_path)
path_input_op_conf, path_lbi = _GenModelIOPathInputOpConfAndRetLbi()
path_input_blob_objects = {}
(
BuildModelIOPathInputInstruction,
BuildFeedPathInstruction,
) = _MakeModelIOPathInputBuilds(path_input_op_conf, snapshot_path,
path_input_blob_objects)
model_load_op_conf, _ = _GenModelLoadOpConfAndRetLbi(var_op_conf, path_lbi)
model_load_blob_objects = {}
def BuildModelLoadInstruction(builder):
path_blob_object = path_input_blob_objects["out"]
model_load_blob_objects["path"] = path_blob_object
op_attribute = op_infer_util.Infer(
model_load_op_conf, ibn2blob_object=model_load_blob_objects)
parallel_conf = path_blob_object.parallel_desc_symbol.parallel_conf
builder.StatelessCall(op_attribute,
parallel_conf,
bn_in_op2blob_object=model_load_blob_objects)
sess = session_ctx.GetDefaultSession()
with scope_util.ScopeContext(scope_util.MakeScope(_BuildNotMirroredScope)):
vm_util.LogicalRun(BuildModelIOPathInputInstruction)
vm_util.LogicalRun(BuildFeedPathInstruction)
vm_util.LogicalRun(BuildModelLoadInstruction)
return model_load_blob_objects["out_0"]
def _EagerRunModelSave(var_blobs, snapshot_path):
path_input_op_conf, path_lbi = _GenModelIOPathInputOpConfAndRetLbi()
path_input_blob_objects = {}
(
BuildModelIOPathInputInstruction,
BuildFeedPathInstruction,
) = _MakeModelIOPathInputBuilds(path_input_op_conf, snapshot_path,
path_input_blob_objects)
model_save_op_conf = _GenModelSaveOpConf(var_blobs, path_lbi)
model_save_blob_objects = oneflow_api.deprecated.BnInOp2BlobObject()
def BuildModelSaveInstruction(builder):
path_blob_object = path_input_blob_objects["out"]
model_save_blob_objects["path"] = path_blob_object
for i, blob in enumerate(var_blobs):
model_save_blob_objects["in_{}".format(i)] = blob.blob_object
op_attribute = op_infer_util.Infer(
model_save_op_conf, ibn2blob_object=model_save_blob_objects)
parallel_conf = path_blob_object.parallel_desc_symbol.parallel_conf
cfg_op_attribute = oneflow_api.deprecated.MakeOpAttributeByString(
str(op_attribute))
builder.StatelessCall(
cfg_op_attribute,
parallel_conf,
model_save_blob_objects,
boxing_util.BoxingTo,
)
sess = session_ctx.GetDefaultSession()
with scope_util.ScopeContext(scope_util.MakeScope(_BuildNotMirroredScope)):
oneflow_api.deprecated.LogicalRun(BuildModelIOPathInputInstruction)
oneflow_api.deprecated.LogicalRun(BuildFeedPathInstruction)
oneflow_api.deprecated.LogicalRun(BuildModelSaveInstruction)
def _EagerRunModelInit(var_op_conf):
op_conf, _ = _GenModelInitOpConfAndRetLbi(var_op_conf)
bn_in_op2blob_object = oneflow_api.deprecated.BnInOp2BlobObject()
def BuildModelInitInstruction(builder):
upstream_signature = op_node_signature_pb.OpNodeSignature()
op_conf.scope_symbol_id = oneflow.current_scope().symbol_id
op_attribute = c_api_util.InferOpConf(op_conf, upstream_signature)
parallel_conf = (
oneflow.current_scope().device_parallel_desc_symbol.parallel_conf)
cfg_op_attribute = oneflow_api.deprecated.MakeOpAttributeByString(
str(op_attribute))
builder.StatelessCall(
cfg_op_attribute,
parallel_conf,
bn_in_op2blob_object,
boxing_util.BoxingTo,
vm_util._FindOrCreateDelegateBlobObject,
)
sess = session_ctx.GetDefaultSession()
with scope_util.ScopeContext(scope_util.MakeScope(_BuildNotMirroredScope)):
vm_util.LogicalRun(BuildModelInitInstruction)
return bn_in_op2blob_object["out_0"]
def GetNormalModePlacementScope(device_tag, machine_device_ids):
sess = session_ctx.GetDefaultSession()
scope = scope_util.MakeScope(
lambda old_scope, builder: old_scope.BuildWithNewParallelDesc(
builder, device_tag, machine_device_ids
)
)
return scope_util.ScopeContext(scope)
def GetNormalModePlacementScope(device_tag, machine_device_ids):
if isinstance(machine_device_ids, tuple):
machine_device_ids = list(machine_device_ids)
if not isinstance(machine_device_ids, list):
machine_device_ids = [machine_device_ids]
sess = session_ctx.GetDefaultSession()
scope = scope_util.MakeScope(
lambda old_scope, builder: builder.BuildScopeWithNewParallelDesc(
old_scope, device_tag, machine_device_ids))
return scope_util.ScopeContext(scope)
def GetGlobalModePlacementScope(device_tag, machine_device_ids):
if isinstance(machine_device_ids, (list, tuple)) == False:
machine_device_ids = [machine_device_ids]
sess = session_ctx.GetDefaultSession()
def BuildScope(old_scope, builder):
return old_scope.BuildWithNewParallelDesc(builder, device_tag,
machine_device_ids)
scope_ctx = scope_util.ScopeContext(scope_util.MakeScope(BuildScope))
return placement_ctx.GlobalModePlacementScope(scope_ctx)
def __init__(self, is_mirrored):
self.is_mirrored_ = is_mirrored
self.scope_context_ = None
sess = session_ctx.GetDefaultSession()
# bypass the first DistributeStrategy for avoiding None old_scope
if sess.is_running and len(sess.is_mirrored_strategy_enabled_stack) > 0:
def BuildScope(old_scope, builder):
return old_scope.BuildWithNewIsMirrored(builder, is_mirrored)
self.scope_context_ = scope_util.ScopeContext(
scope_util.MakeScope(BuildScope)
)
def GetGlobalModePlacementScope(device_tag, machine_device_ids, hierarchy=None):
if isinstance(machine_device_ids, (list, tuple)) == False:
machine_device_ids = [machine_device_ids]
sess = session_ctx.GetDefaultSession()
if hierarchy is not None:
hierarchy = oneflow_api.Size(tuple(hierarchy))
def BuildScope(old_scope, builder):
return builder.BuildScopeWithNewParallelDesc(
old_scope, device_tag, machine_device_ids, hierarchy
)
scope_ctx = scope_util.ScopeContext(scope_util.MakeScope(BuildScope))
return placement_ctx.GlobalModePlacementScope(scope_ctx)
def GetNormalModePlacementScope(device_tag,
machine_device_ids,
hierarchy=None):
if isinstance(machine_device_ids, tuple):
machine_device_ids = list(machine_device_ids)
if not isinstance(machine_device_ids, list):
machine_device_ids = [machine_device_ids]
sess = session_ctx.GetDefaultSession()
if hierarchy is not None:
hierarchy = oneflow._oneflow_internal.Size(tuple(hierarchy))
scope = scope_util.MakeScope(
lambda old_scope, builder: builder.BuildScopeWithNewParallelDesc(
old_scope, device_tag, machine_device_ids, hierarchy))
return scope_util.ScopeContext(scope)
def _EagerRunModelLoad(var_op_conf, snapshot_path):
assert isinstance(snapshot_path, str)
assert os.path.basename(snapshot_path) == "out"
snapshot_path = os.path.dirname(snapshot_path)
assert os.path.basename(snapshot_path) == var_op_conf.name
snapshot_path = os.path.dirname(snapshot_path)
path_input_op_conf, path_lbi = _GenModelIOPathInputOpConfAndRetLbi()
path_input_blob_objects = {}
(
BuildModelIOPathInputInstruction,
BuildFeedPathInstruction,
) = _MakeModelIOPathInputBuilds(path_input_op_conf, snapshot_path,
path_input_blob_objects)
model_load_op_conf, _ = _GenModelLoadOpConfAndRetLbi(var_op_conf, path_lbi)
model_load_blob_objects = oneflow._oneflow_internal.deprecated.BnInOp2BlobObject(
)
def BuildModelLoadInstruction(builder):
path_blob_object = path_input_blob_objects["out"]
model_load_blob_objects["path"] = path_blob_object
op_attribute = op_infer_util.Infer(
model_load_op_conf, ibn2blob_object=model_load_blob_objects)
parallel_conf = path_blob_object.parallel_desc_symbol.parallel_conf
cfg_op_attribute = oneflow._oneflow_internal.deprecated.MakeOpAttributeByString(
str(op_attribute))
builder.StatelessCall(
cfg_op_attribute,
parallel_conf,
model_load_blob_objects,
boxing_util.BoxingTo,
)
sess = session_ctx.GetDefaultSession()
with scope_util.ScopeContext(scope_util.MakeScope(_BuildNotMirroredScope)):
oneflow._oneflow_internal.deprecated.LogicalRun(
BuildModelIOPathInputInstruction)
oneflow._oneflow_internal.deprecated.LogicalRun(
BuildFeedPathInstruction)
oneflow._oneflow_internal.deprecated.LogicalRun(
BuildModelLoadInstruction)
return model_load_blob_objects["out_0"]
def _EagerRunModelInit(var_op_conf):
op_conf, _ = _GenModelInitOpConfAndRetLbi(var_op_conf)
bn_in_op2blob_object = {}
def BuildModelInitInstruction(builder):
upstream_signature = op_attribute_pb.OpNodeSignature()
op_conf.scope_symbol_id = oneflow.current_scope().symbol_id
op_attribute = c_api_util.InferOpConf(op_conf, upstream_signature)
parallel_conf = (
oneflow.current_scope().device_parallel_desc_symbol.parallel_conf)
builder.StatelessCall(op_attribute,
parallel_conf,
bn_in_op2blob_object=bn_in_op2blob_object)
sess = session_ctx.GetDefaultSession()
with scope_util.ScopeContext(scope_util.MakeScope(_BuildNotMirroredScope)):
vm_util.LogicalRun(BuildModelInitInstruction)
return bn_in_op2blob_object["out_0"]
def _GetReturnOpConfAndOutLbiAndScope(remote_blob, allow_cpu_return_op=True):
op_conf = op_conf_util.OperatorConf()
op_conf.name = id_util.UniqueStr("Return_")
setattr(op_conf.return_conf, "in", remote_blob.unique_name)
op_conf.return_conf.out = "out"
if allow_cpu_return_op:
op_conf.device_tag = "cpu"
lbi = logical_blob_id_util.LogicalBlobId()
lbi.op_name = op_conf.name
lbi.blob_name = "out"
parallel_conf = placement_proto_pb.ParallelConf()
parallel_conf.CopyFrom(remote_blob.parallel_conf)
def BuildScope(old_scope, builder):
return old_scope.BuildWithNewParallelConf(builder, parallel_conf)
sess = session_ctx.GetDefaultSession()
scope = scope_util.MakeScope(BuildScope)
return op_conf, lbi, scope