forked from mindspore-Ecosystem/mindspore
!17240 add parallel layernorm test case
From: @hanyang001 Reviewed-by: @stsuteng,@yangzhenzhang Signed-off-by: @stsuteng
This commit is contained in:
Gitee
commit
5e87097d15
|
|
@ -0,0 +1,410 @@
|
|||
# Copyright 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.
|
||||
# ============================================================================
|
||||
import numpy as np
|
||||
import pytest
|
||||
|
||||
import mindspore as ms
|
||||
from mindspore import context, Tensor, Parameter
|
||||
from mindspore.common.api import _executor
|
||||
from mindspore.common.initializer import initializer
|
||||
from mindspore.nn import Cell, TrainOneStepCell, Momentum
|
||||
from mindspore.ops import operations as P
|
||||
|
||||
|
||||
class Net(Cell):
|
||||
def __init__(self, begin_norm_axis, begin_params_axis, mul_weight, normalized_shape,
|
||||
strategy1=None, strategy2=None, strategy3=None):
|
||||
super().__init__()
|
||||
self.begin_norm_axis = begin_norm_axis
|
||||
self.begin_params_axis = begin_params_axis
|
||||
self.mul = P.Mul().shard(strategy1)
|
||||
self.layer_norm = P.LayerNorm(
|
||||
self.begin_norm_axis, self.begin_params_axis).shard(strategy2)
|
||||
self.relu = P.ReLU().shard(strategy3)
|
||||
self.mul_weight = Parameter(mul_weight, "w1")
|
||||
self.normalized_shape = normalized_shape
|
||||
self.gamma = Parameter(initializer(
|
||||
'ones', self.normalized_shape), name="gamma")
|
||||
self.beta = Parameter(initializer(
|
||||
'zeros', self.normalized_shape), name="beta")
|
||||
|
||||
def construct(self, x, b):
|
||||
out = self.mul(x, self.mul_weight)
|
||||
out, _, _ = self.layer_norm(out, self.gamma, self.beta)
|
||||
out = self.relu(out)
|
||||
return out
|
||||
|
||||
|
||||
class Net2(Cell):
|
||||
def __init__(self, begin_norm_axis, begin_params_axis, mul_weight, normalized_shape,
|
||||
strategy1=None, strategy2=None, strategy3=None):
|
||||
super().__init__()
|
||||
self.begin_norm_axis = begin_norm_axis
|
||||
self.begin_params_axis = begin_params_axis
|
||||
self.mul = P.Mul().shard(strategy1)
|
||||
self.layer_norm = P.LayerNorm(
|
||||
self.begin_norm_axis, self.begin_params_axis).shard(strategy2)
|
||||
self.relu = P.ReLU().shard(strategy3)
|
||||
self.mul_weight = Parameter(mul_weight, "w1")
|
||||
self.normalized_shape = normalized_shape
|
||||
self.gamma = Parameter(initializer(
|
||||
'ones', self.normalized_shape), name="gamma")
|
||||
self.beta = Parameter(initializer(
|
||||
'zeros', self.normalized_shape), name="beta")
|
||||
|
||||
def construct(self, x, b):
|
||||
out = self.mul(x, self.mul_weight)
|
||||
_, out, _ = self.layer_norm(out, self.gamma, self.beta)
|
||||
out = self.relu(out)
|
||||
return out
|
||||
|
||||
|
||||
class Net3(Cell):
|
||||
def __init__(self, begin_norm_axis, begin_params_axis, mul_weight, normalized_shape,
|
||||
strategy1=None, strategy2=None, strategy3=None):
|
||||
super().__init__()
|
||||
self.begin_norm_axis = begin_norm_axis
|
||||
self.begin_params_axis = begin_params_axis
|
||||
self.mul = P.Mul().shard(strategy1)
|
||||
self.layer_norm = P.LayerNorm(
|
||||
self.begin_norm_axis, self.begin_params_axis).shard(strategy2)
|
||||
self.relu = P.ReLU().shard(strategy3)
|
||||
self.mul_weight = Parameter(mul_weight, "w1")
|
||||
self.normalized_shape = normalized_shape
|
||||
self.gamma = Parameter(initializer(
|
||||
'ones', self.normalized_shape), name="gamma")
|
||||
self.beta = Parameter(initializer(
|
||||
'zeros', self.normalized_shape), name="beta")
|
||||
|
||||
def construct(self, x, b):
|
||||
out = self.mul(x, self.mul_weight)
|
||||
_, _, out = self.layer_norm(out, self.gamma, self.beta)
|
||||
out = self.relu(out)
|
||||
return out
|
||||
|
||||
|
||||
class Net4(Cell):
|
||||
def __init__(self, begin_norm_axis, begin_params_axis, mul_weight, normalized_shape,
|
||||
strategy1=None, strategy2=None, strategy3=None):
|
||||
super().__init__()
|
||||
self.begin_norm_axis = begin_norm_axis
|
||||
self.begin_params_axis = begin_params_axis
|
||||
self.mul = P.Mul().shard(strategy1)
|
||||
self.layer_norm = P.LayerNorm(
|
||||
self.begin_norm_axis, self.begin_params_axis).shard(strategy2)
|
||||
self.mul_weight = Parameter(mul_weight, "w1")
|
||||
self.normalized_shape = normalized_shape
|
||||
self.gamma = Parameter(initializer(
|
||||
'ones', self.normalized_shape), name="gamma")
|
||||
self.beta = Parameter(initializer(
|
||||
'zeros', self.normalized_shape), name="beta")
|
||||
|
||||
def construct(self, x, b):
|
||||
out = self.mul(x, self.mul_weight)
|
||||
_, _, out = self.layer_norm(out, self.gamma, self.beta)
|
||||
return out
|
||||
|
||||
|
||||
class Net5(Cell):
|
||||
def __init__(self, begin_norm_axis, begin_params_axis, mul_weight, normalized_shape,
|
||||
strategy1=None, strategy2=None, strategy3=None):
|
||||
super().__init__()
|
||||
self.begin_norm_axis = begin_norm_axis
|
||||
self.begin_params_axis = begin_params_axis
|
||||
self.mul = P.Mul().shard(strategy1)
|
||||
self.layer_norm = P.LayerNorm(
|
||||
self.begin_norm_axis, self.begin_params_axis).shard(strategy2)
|
||||
self.relu = P.ReLU().shard(strategy3)
|
||||
self.mul_weight = Parameter(mul_weight, "w1")
|
||||
self.normalized_shape = normalized_shape
|
||||
self.gamma = Parameter(initializer(
|
||||
'ones', self.normalized_shape), name="gamma")
|
||||
self.beta = Parameter(initializer(
|
||||
'zeros', self.normalized_shape), name="beta")
|
||||
|
||||
def construct(self, x, b):
|
||||
out, _, _ = self.layer_norm(x, self.gamma, self.beta)
|
||||
out = self.relu(out)
|
||||
return out
|
||||
|
||||
|
||||
_x = Tensor(np.ones([16, 64, 32, 16]), dtype=ms.float32)
|
||||
_w = Tensor(np.ones([16, 64, 32, 16]), dtype=ms.float32)
|
||||
_b = Tensor(np.ones([16, 64, 32, 16]), dtype=ms.float32)
|
||||
|
||||
|
||||
def compile_net(net):
|
||||
optimizer = Momentum(net.trainable_params(),
|
||||
learning_rate=0.1, momentum=0.9)
|
||||
train_net = TrainOneStepCell(net, optimizer)
|
||||
train_net.set_auto_parallel()
|
||||
train_net.set_train()
|
||||
_executor.compile(train_net, _x, _b)
|
||||
context.reset_auto_parallel_context()
|
||||
|
||||
|
||||
def test_layer_norm_data_parallel():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((16, 1, 1, 1), (16, 1, 1, 1))
|
||||
strategy2 = ((16, 1, 1, 1), (1, 1, 1), (1, 1, 1))
|
||||
strategy3 = ((16, 1, 1, 1),)
|
||||
net = Net(1, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_data_parallel2():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((16, 1, 1, 1), (16, 1, 1, 1))
|
||||
strategy2 = ((16, 1, 1, 1), (1, 1, 1), (1, 1, 1))
|
||||
strategy3 = ((16, 1, 1, 1),)
|
||||
net = Net2(1, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_data_parallel3():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((16, 1, 1, 1), (16, 1, 1, 1))
|
||||
strategy2 = ((16, 1, 1, 1), (1, 1, 1), (1, 1, 1))
|
||||
strategy3 = ((16, 1, 1, 1),)
|
||||
net = Net3(1, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_data_parallel4():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((16, 1, 1, 1), (16, 1, 1, 1))
|
||||
strategy2 = ((16, 1, 1, 1), (1, 1, 1), (1, 1, 1))
|
||||
strategy3 = ((16, 1, 1, 1),)
|
||||
net = Net4(1, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_data_parallel5():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((16, 1, 1, 1), (16, 1, 1, 1))
|
||||
strategy2 = ((16, 1, 1, 1), (1, 1, 1), (1, 1, 1))
|
||||
strategy3 = ((16, 1, 1, 1),)
|
||||
net = Net5(1, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_model_parallel():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((1, 16, 1, 1), (1, 16, 1, 1))
|
||||
strategy2 = ((1, 16, 1, 1), (16, 1, 1), (16, 1, 1))
|
||||
strategy3 = ((1, 16, 1, 1),)
|
||||
net = Net(2, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_model_parallel2():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((1, 16, 1, 1), (1, 16, 1, 1))
|
||||
strategy2 = ((1, 16, 1, 1), (16, 1, 1), (16, 1, 1))
|
||||
strategy3 = ((1, 16, 1, 1),)
|
||||
net = Net2(2, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_model_parallel3():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((1, 16, 1, 1), (1, 16, 1, 1))
|
||||
strategy2 = ((1, 16, 1, 1), (16, 1, 1), (16, 1, 1))
|
||||
strategy3 = ((1, 16, 1, 1),)
|
||||
net = Net3(2, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_model_parallel4():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((1, 16, 1, 1), (1, 16, 1, 1))
|
||||
strategy2 = ((1, 16, 1, 1), (16, 1, 1), (16, 1, 1))
|
||||
strategy3 = ((1, 16, 1, 1),)
|
||||
net = Net4(2, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_model_parallel5():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((1, 16, 1, 1), (1, 16, 1, 1))
|
||||
strategy2 = ((1, 16, 1, 1), (16, 1, 1), (16, 1, 1))
|
||||
strategy3 = ((1, 16, 1, 1),)
|
||||
net = Net5(2, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_hybrid_parallel():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 4, 2, 1), (2, 4, 2, 1))
|
||||
strategy2 = ((2, 4, 2, 1), (4, 2, 1), (4, 2, 1))
|
||||
strategy3 = ((2, 4, 2, 1),)
|
||||
net = Net(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_hybrid_parallel2():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 4, 2, 1), (2, 4, 2, 1))
|
||||
strategy2 = ((2, 4, 2, 1), (4, 2, 1), (4, 2, 1))
|
||||
strategy3 = ((2, 4, 2, 1),)
|
||||
net = Net2(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_hybrid_parallel3():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 4, 2, 1), (2, 4, 2, 1))
|
||||
strategy2 = ((2, 4, 2, 1), (4, 2, 1), (4, 2, 1))
|
||||
strategy3 = ((2, 4, 2, 1),)
|
||||
net = Net3(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_hybrid_parallel4():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 4, 2, 1), (2, 4, 2, 1))
|
||||
strategy2 = ((2, 4, 2, 1), (4, 2, 1), (4, 2, 1))
|
||||
strategy3 = ((2, 4, 2, 1),)
|
||||
net = Net4(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_hybrid_parallel5():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 4, 2, 1), (2, 4, 2, 1))
|
||||
strategy2 = ((2, 4, 2, 1), (4, 2, 1), (4, 2, 1))
|
||||
strategy3 = ((2, 4, 2, 1),)
|
||||
net = Net5(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_repeat_calc():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((2, 2, 1, 1), (2, 1, 1), (2, 1, 1))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_repeat_calc2():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((2, 2, 1, 1), (2, 1, 1), (2, 1, 1))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net2(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_repeat_calc3():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((2, 2, 1, 1), (2, 1, 1), (2, 1, 1))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net3(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_repeat_calc4():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((2, 2, 1, 1), (2, 1, 1), (2, 1, 1))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net4(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_repeat_calc5():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((2, 2, 1, 1), (2, 1, 1), (2, 1, 1))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net5(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_wrong_strategy1():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((1, 2, 1, 2), (2, 1, 2), (2, 1, 2))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net(1, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
with pytest.raises(RuntimeError):
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_wrong_strategy2():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((1, 2, 1, 2), (2, 1, 2), (2, 1, 2))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net(2, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
with pytest.raises(RuntimeError):
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_wrong_strategy3():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((1, 2, 1, 2), (2, 1, 2), (2, 1, 2))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
with pytest.raises(RuntimeError):
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_wrong_strategy4():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((1, 2, 1, 1), (2, 1, 1), (2, 1, 1))
|
||||
strategy3 = ((2, 2, 4, 1),)
|
||||
net = Net2(2, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
with pytest.raises(RuntimeError):
|
||||
compile_net(net)
|
||||
|
||||
|
||||
def test_layer_norm_wrong_strategy5():
|
||||
context.set_auto_parallel_context(
|
||||
parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
|
||||
strategy1 = ((2, 2, 4, 1), (2, 2, 4, 1))
|
||||
strategy2 = ((1, 2, 1, 2), (2, 1, 2), (2, 1, 2))
|
||||
strategy3 = ((2, 2, 1, 4),)
|
||||
net = Net3(3, 1, _w, [64, 32, 16], strategy1, strategy2, strategy3)
|
||||
with pytest.raises(RuntimeError):
|
||||
compile_net(net)
|
||||
Loading…
Reference in New Issue