# This is the Python adaptation and derivative work of Myia (https://github.com/mila-iqia/myia/).
#
# Copyright 2020-2021 Huawei Technologies Co., Ltd
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ============================================================================
"""Providing interface methods."""
import types
import sys
import os
import time
from collections import OrderedDict
from functools import wraps
from mindspore import context
from mindspore import log as logger
from mindspore._extends.remote import kernel_build_server
from .tensor import Tensor as MsTensor
from .._c_expression import generate_arguments_key, GraphExecutor_, Tensor, MetaTensor, PynativeExecutor_
from .._c_expression import verify_inputs_signature, init_exec_dataset, _set_dataset_mode_config, init_pipeline
from ..parallel._ps_context import _is_role_pserver
from ..parallel._utils import _get_device_num, _get_global_rank, _need_to_full, _check_full_batch, _to_full_tensor, \
_get_parameter_broadcast, _get_pipeline_stages
# store ms_function class compiled pipeline cache
ms_compile_cache = {}
BROADCAST_PHASE = "_broadcast_"
def _convert_function_arguments(fn, *args):
"""
Process the fn default parameters.
Args:
fn (Function): The function to be parsed.
args (tuple): The parameters of the function.
"""
arguments_dict = OrderedDict()
parse_method = None
if isinstance(fn, (types.FunctionType, types.MethodType)):
parse_method = fn.__name__
index = 0
for value in args:
arguments_dict[f'arg{index}'] = value
index = index + 1
logger.debug("fn(%r) full parameters dict is: %r", fn, arguments_dict)
converted = True
else:
logger.warning("Find error: fn isn't function or method")
converted = False
return converted, arguments_dict, parse_method
def _wrap_func(fn):
"""
Wrapper function, convert return data to tensor or tuple of tensor.
Args:
fn (Function): The function need be wrapped.
Returns:
Function, a new function with return suitable format data.
"""
@wraps(fn)
def wrapper(*arg, **kwargs):
results = fn(*arg, **kwargs)
def _convert_data(data):
if isinstance(data, Tensor) and not isinstance(data, MsTensor):
return MsTensor(data)
if isinstance(data, tuple):
return tuple(_convert_data(x) for x in data)
if isinstance(data, list):
return list(_convert_data(x) for x in data)
return data
return _convert_data(results)
return wrapper
def _exec_init_graph(obj, init_phase):
"""Execute the parameter initializer graph."""
inst_executor = GraphExecutor_.get_instance()
param_dict = OrderedDict()
for name, param in obj.parameters_dict().items():
if not param.is_init:
param_dict[name] = param
param.is_init = True
param.data.init_flag = True
if param_dict:
inst_executor.run_init_graph(param_dict, init_phase)
class _MindsporeFunctionExecutor:
"""
Represents a function compiled by graph compiler.
_MindsporeFunctionExecutor will compile the original function for every combination
of argument types and shapes it is given (as well as their values, optionally).
Args:
fn (Function): The root function to compile.
input_signature (Function): User defines signature to verify input.
ms_create_time(TimeStamp): The time ms_function created
obj (Object): If function is a method, obj is the owner of function,
else, obj is none.
Returns:
The result of pipeline running in graph mode.
"""
def __init__(self, fn, ms_create_time, input_signature=None, obj=None):
self.fn = fn
self.input_signature = input_signature
self.obj = None
if hasattr(obj, fn.__name__):
self.obj = obj
self._graph_executor = GraphExecutor_.get_instance()
self._create_time = ms_create_time
def build_data_init_graph(self, graph_name):
"""Build GE data graph and init graph for the given graph name."""
if self.obj is None:
logger.warning("Make sure parameter should not be used in function")
para_dict = OrderedDict()
self._graph_executor.build_data_graph(para_dict, graph_name)
return
self._graph_executor.build_data_graph(self.obj.parameters_dict(), graph_name,
self.obj.parameters_broadcast_dict())
init_phase = "init_subgraph" + graph_name[graph_name.find("."):]
_exec_init_graph(self.obj, init_phase)
def compile(self, args_list, arg_names, method_name):
"""Returns pipeline for the given args."""
# Verify the signature for both function and method
if self.input_signature is not None:
signatures = []
for sig_spec in self.input_signature:
if not isinstance(sig_spec, MetaTensor):
raise TypeError("Input_signature is not MetaTensor")
signatures.append(sig_spec)
is_valid_input = verify_inputs_signature(signatures, args_list)
if not is_valid_input:
raise ValueError("Inputs is incompatible with input signature!")
dic = dict(zip(arg_names, args_list))
generate_name = self.fn.__module__ + "." + self.fn.__name__ + "." + self.fn.__code__.co_filename + "." + \
str(self.fn.__code__.co_firstlineno) + '.' + str(id(self.fn))
if _pynative_executor.grad_flag():
generate_name = generate_name + ".grad"
self.fn.__parse_method__ = method_name
# Add key with obj
if self.obj is not None:
if self.obj.__module__ != self.fn.__module__:
logger.error(f'`obj` module not equal to `fn` module: {self.obj.__module__}, {self.fn.__module__}')
self.obj.__parse_method__ = method_name
generate_name = generate_name + '.' + str(self.obj.create_time) + '.' + str(id(self.obj))
else:
# Different instance of same class may use same memory(means same obj_id) at diff times.
# To avoid unexpected phase matched, add create_time to generate_name.
generate_name = generate_name + '.' + str(self._create_time)
key = generate_arguments_key(dic)
phase = generate_name + '.' + str(key)
if phase not in ms_compile_cache.keys():
if self.obj is None:
is_compile = self._graph_executor.compile(self.fn, args_list, phase, True, "")
else:
is_compile = self._graph_executor.compile(self.obj, args_list, phase, True, "")
if not is_compile:
raise RuntimeError("Executor compile failed.")
if context.get_context("enable_ge"):
self.build_data_init_graph(phase)
ms_compile_cache[phase] = phase
return phase
return phase
@_wrap_func
def __call__(self, *args):
init_pipeline()
converted, arguments_dict, parse_method = _convert_function_arguments(self.fn, *args)
if not converted:
raise RuntimeError('Process function parameter is failure')
args_list = tuple(arguments_dict.values())
arg_names = tuple(arguments_dict.keys())
if self.obj is not None:
args_list = args_list[1:]
arg_names = arg_names[1:]
phase = self.compile(args_list, arg_names, parse_method)
if context.get_context("precompile_only"):
return None
new_inputs = []
for i in args_list:
if isinstance(i, Tensor):
new_inputs.append(i)
elif context.get_context("grad_for_scalar") and isinstance(i, (int, float)):
new_inputs.append(i)
output = self._graph_executor(tuple(new_inputs), phase)
if context.get_context("mode") == context.PYNATIVE_MODE:
_pynative_executor.set_graph_phase(phase)
output = _pynative_executor.grad_ms_function(output, *new_inputs)
return output
[docs]def ms_function(fn=None, obj=None, input_signature=None):
"""
Create a callable MindSpore graph from a Python function.
This allows the MindSpore runtime to apply optimizations based on graph.
Args:
fn (Function): The Python function that will be run as a graph. Default: None.
obj (Object): The Python object is used to distinguish the compiled function. Default: None.
input_signature (Tensor): The Tensor which describes the input arguments. The shape and dtype of the Tensor
will be supplied to this function. If input_signature is specified, each input to `fn` must be a `Tensor`.
And the input parameters of `fn` cannot accept `**kwargs`. The shape and dtype of actual inputs should
keep the same as input_signature. Otherwise, TypeError will be raised. Default: None.
Returns:
Function, if `fn` is not None, returns a callable function that will execute the compiled function; If `fn` is
None, returns a decorator and when this decorator invokes with a single `fn` argument, the callable function is
equal to the case when `fn` is not None.
Supported Platforms:
``Ascend`` ``GPU`` ``CPU``
Examples:
>>> import numpy as np
>>> from mindspore import Tensor
>>> from mindspore import ms_function
...
>>> x = Tensor(np.ones([1, 1, 3, 3]).astype(np.float32))
>>> y = Tensor(np.ones([1, 1, 3, 3]).astype(np.float32))
...
>>> # create a callable MindSpore graph by calling ms_function
>>> def tensor_add(x, y):
... z = x + y
... return z
...
>>> tensor_add_graph = ms_function(fn=tensor_add)
>>> out = tensor_add_graph(x, y)
...
>>> # create a callable MindSpore graph through decorator @ms_function
>>> @ms_function
... def tensor_add_with_dec(x, y):
... z = x + y
... return z
...
>>> out = tensor_add_with_dec(x, y)
...
>>> # create a callable MindSpore graph through decorator @ms_function with input_signature parameter
>>> @ms_function(input_signature=(Tensor(np.ones([1, 1, 3, 3]).astype(np.float32)),
... Tensor(np.ones([1, 1, 3, 3]).astype(np.float32))))
... def tensor_add_with_sig(x, y):
... z = x + y
... return z
...
>>> out = tensor_add_with_sig(x, y)
"""
def wrap_mindspore(func):
ms_create_time = int(time.time() * 1e9)
@wraps(func)
def staging_specialize(*args):
if obj is not None:
logger.warning("Obj is no longer in use, and the function's own object has been used to \
distinguish whether it has been compiled.")
process_obj = None
if args and not isinstance(args[0], MsTensor) and hasattr(args[0], func.__name__):
process_obj = args[0]
out = _MindsporeFunctionExecutor(func, ms_create_time, input_signature, process_obj)(*args)
return out
return staging_specialize
if fn is not None:
return wrap_mindspore(fn)
return wrap_mindspore
def _generate_pip_args(obj, *args, method="construct"):
"""Generate arguments for pipeline."""
if hasattr(obj, method):
fn = getattr(obj, method)
else:
raise AttributeError('The process method is not exist')
converted, arguments_dict, parse_method = _convert_function_arguments(fn, *args)
if not converted:
raise RuntimeError('Process method parameter is failure')
args_list = tuple(arguments_dict.values())
args_names = tuple(arguments_dict.keys())
obj.__parse_method__ = parse_method
return args_names, args_list
def _get_auto_split_param_names(parameter_layout_dict):
auto_split_param_names = []
for key, value in parameter_layout_dict.items():
for dim in value[1]:
if dim != -1:
auto_split_param_names.append(key)
break
return auto_split_param_names
def _build_broadcast_graph(broadcast_params_dict, broadcast_phase):
"""Build broadcast graph."""
from mindspore.nn.wrap.cell_wrapper import _BroadCastCell
if not broadcast_params_dict:
broadcast_params_dict = {}
broadcast_params = []
for param in broadcast_params_dict.values():
broadcast_params.append(Tensor(param.asnumpy()))
_broadcast_net = _BroadCastCell(broadcast_params)
_broadcast_net.phase = broadcast_phase
broadcasted_params = _broadcast_net()
for param_name, param in zip(broadcast_params_dict.keys(), broadcasted_params):
broadcast_params_dict[param_name].set_data(param)
def _parameter_broadcast(obj, auto_parallel_mode):
"""Parameter broadcast."""
auto_split_param_names = []
if auto_parallel_mode:
auto_split_param_names = _get_auto_split_param_names(obj.parameter_layout_dict)
broadcast_params_dict = obj.parameters_broadcast_dict()
if auto_split_param_names and broadcast_params_dict:
broadcast_params_dict = OrderedDict()
for param_name, param in obj.parameters_broadcast_dict().items():
if param_name not in auto_split_param_names:
broadcast_params_dict[param_name] = param
broadcast_phase = "_broadcast_subgraph"
_build_broadcast_graph(broadcast_params_dict, broadcast_phase)
class _PynativeExecutor:
"""
A pynative executor used to compile/manage/run single op.
The main functions include: construct op graph, compile op graph, auto grad and run op graph.
Args:
obj (Object): The python network that will be run in pynative mode.
args (Tuple(Tensor...)): The inputs of network in tuple form.
Returns:
gradients (Tuple(Tensor...)): The gradients of network parameters and inputs.
Supported Platforms:
``Ascend`` ``GPU`` ``CPU``
"""
def __init__(self):
self._executor = PynativeExecutor_.get_instance()
self._executor.set_py_exe_path(sys.executable)
self._executor.set_kernel_build_server_dir(os.path.split(kernel_build_server.__file__)[0] + os.sep)
def new_graph(self, obj, *args, **kwargs):
self._executor.new_graph(obj, *args, *(kwargs.values()))
def end_graph(self, obj, output, *args, **kwargs):
self._executor.end_graph(obj, output, *args, *(kwargs.values()))
def check_graph(self, obj, *args, **kwargs):
return self._executor.check_graph(obj, *args, *(kwargs.values()))
def check_run(self, grad, obj, *args, **kwargs):
return self._executor.check_run(grad, obj, *args, *(kwargs.values()))
def grad(self, grad, obj, weights, *args, **kwargs):
self._executor.grad_net(grad, obj, weights, *args, *(kwargs.values()))
def del_cell(self, cell_id=""):
self._executor.clear_cell(cell_id)
def clear_res(self):
return self._executor.clear_res()
def clear_grad(self, obj, *args, **kwargs):
self._executor.clear_grad(obj, *args, *(kwargs.values()))
def sync(self):
self._executor.sync()
def set_lazy_build(self, enable):
self._executor.set_lazy_build(enable)
def execute_all_task(self):
self._executor.execute_all_task()
def grad_ms_function(self, output, *args):
return self._executor.grad_ms_function(output, *args)
def set_graph_phase(self, phase):
self._executor.set_graph_phase(phase)
def grad_flag(self):
return self._executor.grad_flag()
def set_grad_flag(self, flag):
self._executor.set_grad_flag(flag)
def enter_construct(self, cell):
self._executor.enter_construct(cell)
def leave_construct(self, cell):
self._executor.leave_construct(cell)
def parameter_broadcast(self, obj, phase, auto_parallel_mode):
if BROADCAST_PHASE not in phase and _get_parameter_broadcast():
_parameter_broadcast(obj, auto_parallel_mode)
def enter_cell(self):
self._executor.enter_cell()
def exit_cell(self):
self._executor.exit_cell()
def is_top_cell(self):
return self._executor.is_top_cell()
def __call__(self, obj, *args, **kwargs):
args = args + tuple(kwargs.values())
return self._executor(obj, args)
class _CellGraphExecutor:
"""
An executor used to compile/manage/run graph for a Cell.
Including data_graph, train_graph, eval_graph and predict graph.
Args:
obj (Function/Cell): The function or cell instance need compile.
args (tuple): Function or cell input arguments.
Returns:
Graph, return the result of pipeline running.
"""
def __init__(self):
# create needed graph by lazy mode
self.is_init = False
self._graph_executor = GraphExecutor_.get_instance()
self.compile_cache = {}
self._graph_executor.set_py_exe_path(sys.executable)
self._graph_executor.set_kernel_build_server_dir(os.path.split(kernel_build_server.__file__)[0] + os.sep)
self.queue_name = ""
def init_dataset(self, queue_name, dataset_size, batch_size, dataset_types, dataset_shapes,
input_indexs, phase='dataset'):
"""
Initialization interface for calling data subgraph.
Args:
queue_name (str): The name of tdt queue on the device.
dataset_size (int): The size of dataset.
batch_size (int): The size of batch.
dataset_types (list): The output types of element in dataset.
dataset_shapes (list): The output shapes of element in dataset.
input_indexs (list): The index of data with net.
phase (str): The name of phase, e.g., train_dataset/eval_dataset. Default: 'dataset'.
Returns:
bool, specifies whether the data subgraph was initialized successfully.
"""
if not init_exec_dataset(queue_name=queue_name,
size=dataset_size,
batch_size=batch_size,
types=dataset_types,
shapes=dataset_shapes,
input_indexs=input_indexs,
phase=phase):
raise RuntimeError("Failure to init and dataset subgraph!")
self.queue_name = queue_name
return True
def _build_data_graph(self, obj, phase):
self._graph_executor.build_data_graph(obj.parameters_dict(), phase, obj.parameters_broadcast_dict())
def _set_dataset_mode(self, args_list):
"""set dataset mode."""
# decide whether to sink based on whether the inputs is virtual or args_list is ()
if (args_list and isinstance(args_list[0], Tensor) and args_list[0].virtual_flag) or \
(args_list is not None and args_list == ()):
_set_dataset_mode_config('sink')
else:
_set_dataset_mode_config('normal')
def compile(self, obj, *args, phase='predict', do_convert=True, auto_parallel_mode=False):
"""
Compiles graph.
Args:
obj (Function/Cell): The function or cell instance need compile.
args (tuple): Function or cell input arguments.
phase (str): The name of compile phase. Default: 'predict'.
do_convert (bool): When set to True, convert ME graph to GE graph after compiling graph.
auto_parallel_mode: When set to True, use auto parallel mode to compile graph.
Return:
Str, the full phase of the cell.
Bool, if the graph has been compiled before, return False, else return True.
"""
args_names, args_list = _generate_pip_args(obj, *args)
dic = dict(zip(args_names, args_list))
key = generate_arguments_key(dic)
obj.arguments_key = str(key)
phase = phase + '.' + str(obj.create_time) + '.' + str(id(obj)) + '.' + obj.arguments_key
if phase in self.compile_cache.keys():
logger.debug("%r graph has existed.", phase)
return phase, False
obj.check_names()
_check_full_batch()
self._set_dataset_mode(args_list)
is_sink_mode = args and isinstance(args[0], Tensor) and args[0].virtual_flag
if auto_parallel_mode and _need_to_full() and not is_sink_mode and obj.auto_parallel_compile_and_run():
args_full = _to_full_tensor(args, _get_device_num(), _get_global_rank())
_, args_list = _generate_pip_args(obj, *args_full)
enable_debug_runtime = context.get_context("enable_debug_runtime")
enable_ge = context.get_context("enable_ge")
use_vm = not enable_ge or (enable_debug_runtime and context.get_context("mode") == context.PYNATIVE_MODE)
result = self._graph_executor.compile(obj, args_list, phase, use_vm, self.queue_name)
self.compile_cache[phase] = phase
if not result:
raise RuntimeError("Executor compile failed.")
graph = self._graph_executor.get_func_graph(phase)
if graph is None:
raise RuntimeError("Compile graph failed for phase {}.".format(phase))
self._auto_parallel_process(obj, phase, is_sink_mode, auto_parallel_mode, *args)
if not do_convert:
return phase, True
# the following GE init process is not needed when use vm or ms backend
if enable_ge:
self._build_data_graph(obj, phase)
if "export" not in phase:
init_phase = "init_subgraph." + str(obj.create_time) + "." + str(id(obj))
_exec_init_graph(obj, init_phase)
elif "export" in phase:
self._build_data_graph(obj, phase)
elif BROADCAST_PHASE not in phase and _get_parameter_broadcast():
_parameter_broadcast(obj, auto_parallel_mode)
return phase, True
def _auto_parallel_process(self, obj, phase, is_sink_mode, auto_parallel_mode, *args):
"""compile graph in auto parallel mode."""
if not auto_parallel_mode:
replace = obj.init_parameters_data(auto_parallel_mode=auto_parallel_mode)
self._update_param_node_default_input(phase, replace)
return
obj.parameter_layout_dict = self._graph_executor.get_parameter_layout(phase)
obj.parallel_parameter_name_list = self._graph_executor.get_parallel_parameter_name_list(phase)
replace = obj.init_parameters_data(auto_parallel_mode=True)
if _get_pipeline_stages() > 1 and (not hasattr(obj, "is_first_iteration") or not obj.is_first_iteration):
obj.remove_redundant_parameters()
if not context.get_context("enable_debug_runtime") or context.get_context("enable_ge"):
obj.load_parameter_slice(None)
self._update_param_node_default_input(phase, replace)
# set parallel inputs in sink mode
if is_sink_mode:
obj.set_parallel_input_with_inputs(*args)
def _update_param_node_default_input(self, phase, replace):
new_param = {x.name: replace[x] for x in replace if id(x) != id(replace[x])}
return self._graph_executor.updata_param_node_default_input(phase, new_param)
def _get_shard_strategy(self, obj):
real_phase = obj.phase + '.' + str(obj.create_time) + '.' + str(id(obj)) + '.' + obj.arguments_key
return self._graph_executor.get_strategy(real_phase)
def _get_num_parallel_ops(self, obj):
real_phase = obj.phase + '.' + str(obj.create_time) + '.' + str(id(obj)) + '.' + obj.arguments_key
return self._graph_executor.get_num_parallel_ops(real_phase)
def _get_allreduce_fusion(self, obj):
real_phase = obj.phase + '.' + str(obj.create_time) + '.' + str(id(obj)) + '.' + obj.arguments_key
return self._graph_executor.get_allreduce_fusion(real_phase)
def has_compiled(self, phase='predict'):
"""
Specify whether have been compiled.
Args:
phase (str): The phase name. Default: 'predict'.
Returns:
bool, specifies whether the specific graph has been compiled.
"""
return self._graph_executor.has_compiled(phase)
def __call__(self, obj, *args, phase='predict'):
if context.get_context("precompile_only") or _is_role_pserver():
return None
return self.run(obj, *args, phase=phase)
@_wrap_func
def _exec_pip(self, obj, *args, phase=''):
"""Execute the generated pipeline."""
fn = obj.construct
converted, arguments_dict, parse_method = _convert_function_arguments(fn, *args)
if not converted:
raise RuntimeError('Process method parameter is failure')
args_list = tuple(arguments_dict.values())
obj.__parse_method__ = parse_method
return self._graph_executor(args_list, phase)
def run(self, obj, *args, phase='predict'):
"""
Run the specific graph.
Args:
phase (str): The phase name. Default: 'predict'.
Returns:
Tensor/Tuple, return execute result.
"""
if phase == 'save':
return self._graph_executor((), phase + '.' + str(obj.create_time) + '.' + str(id(obj)))
phase_real = phase + '.' + str(obj.create_time) + '.' + str(id(obj)) + '.' + obj.arguments_key
if self.has_compiled(phase_real):
return self._exec_pip(obj, *args, phase=phase_real)
raise KeyError('{} graph is not exist.'.format(phase_real))
def del_net_res(self, net_id):
self._graph_executor.del_net_res(net_id)
def _get_func_graph_proto(self, obj, exec_id, ir_type="onnx_ir", use_prefix=False):
"""Get graph proto from pipeline."""
if use_prefix:
exec_id = exec_id + '.' + obj.arguments_key
if self._graph_executor.has_compiled(exec_id) is False:
return None
return self._graph_executor.get_func_graph_proto(exec_id, ir_type)
def export(self, file_name, graph_id):
"""
Export graph.
Args:
file_name (str): File name of model to export
graph_id (str): id of graph to be exported
"""
from .._c_expression import export_graph
export_graph(file_name, 'AIR', graph_id)
def fetch_info_for_quant_export(self, exec_id):
"""Get graph proto from pipeline."""
if self._graph_executor.has_compiled(exec_id) is False:
return None
return self._graph_executor.fetch_info_for_quant_export(exec_id)
_cell_graph_executor = _CellGraphExecutor()
_pynative_executor = _PynativeExecutor()
__all__ = ['ms_function']