forked from mindspore-Ecosystem/mindspore
267 lines
10 KiB
Python
267 lines
10 KiB
Python
|
# Copyright 2022 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.
|
||
|
|
# ============================================================================
|
||
|
|
""" test scatter update """
|
||
|
|
import numpy as np
|
||
|
|
import mindspore.nn as nn
|
||
|
|
from mindspore import Tensor, Parameter
|
||
|
|
from mindspore.ops import operations as P
|
||
|
|
from mindspore import context
|
||
|
|
from mindspore.common.api import _cell_graph_executor
|
||
|
|
from parallel.utils.utils import ParallelValidator
|
||
|
|
|
||
|
|
scatter_ops_dict = {"Add": P.ScatterAdd(), "Mul": P.ScatterMul(), "Div": P.ScatterDiv(),
|
||
|
|
"Min": P.ScatterMin(), "Max": P.ScatterMax(), "Sub": P.ScatterSub()}
|
||
|
|
|
||
|
|
|
||
|
|
class Net(nn.Cell):
|
||
|
|
"""Net definition"""
|
||
|
|
def __init__(self, input_shape, indices_shape, updates_shape, strategy1=None, strategy2=None, ops="Add"):
|
||
|
|
super(Net, self).__init__()
|
||
|
|
self.inputs = Parameter(Tensor(np.ones(input_shape).astype(np.float32)), "input")
|
||
|
|
self.indices = Tensor(np.ones(indices_shape).astype(np.int32))
|
||
|
|
self.updates = Tensor(np.ones(updates_shape).astype(np.float32))
|
||
|
|
self.scatter_ops = scatter_ops_dict.get(ops)
|
||
|
|
self.scatter_ops.shard(strategy1)
|
||
|
|
self.add = P.TensorAdd().shard(strategy2)
|
||
|
|
self.relu = P.ReLU()
|
||
|
|
|
||
|
|
def construct(self, x):
|
||
|
|
out = self.scatter_ops(self.inputs, self.indices, self.updates)
|
||
|
|
out = self.add(x, out)
|
||
|
|
out = self.relu(out)
|
||
|
|
return out
|
||
|
|
|
||
|
|
|
||
|
|
def compile_net(net, *inputs):
|
||
|
|
net.set_auto_parallel()
|
||
|
|
net.set_train()
|
||
|
|
phase, _ = _cell_graph_executor.compile(net, *inputs)
|
||
|
|
context.reset_auto_parallel_context()
|
||
|
|
return phase
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_add_column_split():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter add column split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=8, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32, 64, 128]).astype(np.float32))
|
||
|
|
input_shape = [32, 64, 128]
|
||
|
|
indices_shape = [4, 8]
|
||
|
|
updates_shape = [4, 8, 64, 128]
|
||
|
|
strategy1 = ((1, 2, 4), (1, 1), (1, 1, 2, 4))
|
||
|
|
strategy2 = ((1, 2, 4), (1, 2, 4))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2)
|
||
|
|
compile_net(net, inputs)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_min_column_split():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter add column split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=8, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32, 64, 128]).astype(np.float32))
|
||
|
|
input_shape = [32, 64, 128]
|
||
|
|
indices_shape = [4, 8]
|
||
|
|
updates_shape = [4, 8, 64, 128]
|
||
|
|
strategy1 = ((1, 2, 4), (1, 1), (1, 1, 2, 4))
|
||
|
|
strategy2 = ((1, 2, 4), (1, 2, 4))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2, "Min")
|
||
|
|
compile_net(net, inputs)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_max_column_split():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter add column split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=8, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32, 64, 128]).astype(np.float32))
|
||
|
|
input_shape = [32, 64, 128]
|
||
|
|
indices_shape = [4, 8]
|
||
|
|
updates_shape = [4, 8, 64, 128]
|
||
|
|
strategy1 = ((1, 2, 4), (1, 1), (1, 1, 2, 4))
|
||
|
|
strategy2 = ((1, 2, 4), (1, 2, 4))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2, "Max")
|
||
|
|
compile_net(net, inputs)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_add_row_split():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter add row split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=8, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32, 64, 128]).astype(np.float32))
|
||
|
|
input_shape = [32, 64, 128]
|
||
|
|
indices_shape = [4, 8]
|
||
|
|
updates_shape = [4, 8, 64, 128]
|
||
|
|
strategy1 = ((8, 1, 1), (1, 1), (1, 1, 1, 1))
|
||
|
|
strategy2 = ((1, 1, 1), (1, 1, 1))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2)
|
||
|
|
phase = compile_net(net, inputs)
|
||
|
|
validator = ParallelValidator(net, phase)
|
||
|
|
# check sub_graph
|
||
|
|
sub_graph = {
|
||
|
|
'ScatterAdd-0': ['input', 'Minimum-0', 'Mul-0'],
|
||
|
|
'Mul-0': ['_GetTensorSlice-1', 'Reshape-0'],
|
||
|
|
}
|
||
|
|
assert validator.check_graph_structure(sub_graph)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_add_mix_split():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter add mix split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=24, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32 * 3, 64, 128]).astype(np.float32))
|
||
|
|
input_shape = [32 * 3, 64, 128]
|
||
|
|
indices_shape = [4, 8]
|
||
|
|
updates_shape = [4, 8, 64, 128]
|
||
|
|
strategy1 = ((3, 2, 4), (1, 1), (1, 1, 2, 4))
|
||
|
|
strategy2 = ((1, 1, 1), (1, 1, 1))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2)
|
||
|
|
compile_net(net, inputs)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_mul_mix_split():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter mul mix split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=64, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32 * 4, 64]).astype(np.float32))
|
||
|
|
input_shape = [32 * 4, 64]
|
||
|
|
indices_shape = [4]
|
||
|
|
updates_shape = [4, 64]
|
||
|
|
strategy1 = ((8, 8), (1,), (1, 8))
|
||
|
|
strategy2 = ((1, 1), (1, 1))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2, "Mul")
|
||
|
|
phase = compile_net(net, inputs)
|
||
|
|
validator = ParallelValidator(net, phase)
|
||
|
|
# check sub_graph
|
||
|
|
sub_graph = {
|
||
|
|
'ScatterMul-0': ['input', 'Minimum-0', 'Add-0'],
|
||
|
|
'Add-0': ['Mul-0', 'Sub-1'],
|
||
|
|
}
|
||
|
|
assert validator.check_graph_structure(sub_graph)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_div_mix_split():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter div mix split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=64, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32 * 4, 64]).astype(np.float32))
|
||
|
|
input_shape = [32 * 4, 64]
|
||
|
|
indices_shape = [4, 4]
|
||
|
|
updates_shape = [4, 4, 64]
|
||
|
|
strategy1 = ((8, 8), (1, 1), (1, 1, 8))
|
||
|
|
strategy2 = ((1, 1), (1, 1))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2, "Div")
|
||
|
|
compile_net(net, inputs)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_add_mix_split_auto_parallel_sharding_prop():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter add mix split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=24, full_batch=True,
|
||
|
|
search_mode="sharding_propagation")
|
||
|
|
inputs = Tensor(np.ones([32 * 3, 64, 128]).astype(np.float32))
|
||
|
|
input_shape = [32 * 3, 64, 128]
|
||
|
|
indices_shape = [4, 8]
|
||
|
|
updates_shape = [4, 8, 64, 128]
|
||
|
|
strategy1 = None
|
||
|
|
strategy2 = ((3, 2, 4), (3, 2, 4))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2)
|
||
|
|
compile_net(net, inputs)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_mul_mix_split_auto_parallel():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter mul mix split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=64, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32 * 4, 64]).astype(np.float32))
|
||
|
|
input_shape = [32 * 4, 64]
|
||
|
|
indices_shape = [4]
|
||
|
|
updates_shape = [4, 64]
|
||
|
|
strategy1 = None
|
||
|
|
strategy2 = None
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2, "Mul")
|
||
|
|
compile_net(net, inputs)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_div_mix_split_auto_parallel_rec():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter div mix split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=64, full_batch=True,
|
||
|
|
search_mode="recursive_programming")
|
||
|
|
inputs = Tensor(np.ones([32 * 4, 64]).astype(np.float32))
|
||
|
|
input_shape = [32 * 4, 64]
|
||
|
|
indices_shape = [4, 4]
|
||
|
|
updates_shape = [4, 4, 64]
|
||
|
|
strategy1 = None
|
||
|
|
strategy2 = None
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2, "Div")
|
||
|
|
compile_net(net, inputs)
|
||
|
|
|
||
|
|
|
||
|
|
def test_scatter_sub_mix_split():
|
||
|
|
"""
|
||
|
|
Feature: test scatter ops auto parallel
|
||
|
|
Description: test scatter sub mix split
|
||
|
|
Expectation: compile success
|
||
|
|
"""
|
||
|
|
context.set_context(mode=context.GRAPH_MODE)
|
||
|
|
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=64, full_batch=True)
|
||
|
|
inputs = Tensor(np.ones([32 * 4, 64]).astype(np.float32))
|
||
|
|
input_shape = [32 * 4, 64]
|
||
|
|
indices_shape = [4, 4]
|
||
|
|
updates_shape = [4, 4, 64]
|
||
|
|
strategy1 = ((8, 8), (1, 1), (1, 1, 8))
|
||
|
|
strategy2 = ((1, 1), (1, 1))
|
||
|
|
net = Net(input_shape, indices_shape, updates_shape, strategy1, strategy2, "Sub")
|
||
|
|
compile_net(net, inputs)
|