mindspore/tests/st/ops/gpu/test_adam_op.py

202 lines
7.4 KiB
Python

# Copyright 2020-2022 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.
# ============================================================================
import math
import pytest
import numpy as np
import mindspore.context as context
import mindspore.nn as nn
import mindspore as ms
from mindspore import Tensor, Parameter
from mindspore.nn import Dense
from mindspore.nn import TrainOneStepCell, WithLossCell
from mindspore.nn.optim import Adam
from mindspore.ops import operations as P
from mindspore.ops.functional import vmap
context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
class NetAdam(nn.Cell):
def __init__(self):
super(NetAdam, self).__init__()
self.batch_size = 1
self.reshape = P.Reshape()
weight = Tensor(np.ones([10, 16]).astype(np.float32) * 0.01)
self.fc1 = Dense(16, 10, weight_init=weight)
def construct(self, input_x):
output = self.reshape(input_x, (self.batch_size, -1))
output = self.fc1(output)
return output
class NetWithSparseGatherV2(nn.Cell):
def __init__(self):
super(NetWithSparseGatherV2, self).__init__()
self.weight1 = Parameter(Tensor(np.ones([3, 1, 2]).astype(np.float32)), name="weight1")
self.weight2 = Parameter(Tensor(np.ones([3, 1, 2]).astype(np.float32)), name="weight2")
self.axis = 0
self.gather = P.SparseGatherV2()
def construct(self, indices, label):
return self.gather(self.weight1, indices, self.axis) + self.weight2
@pytest.mark.level0
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
def test_adam():
"""
Feature: Adam optimizer
Description: Verify if the loss is converged
Expectation: success
"""
epoch = 3
net = NetAdam()
optimizer = Adam(filter(lambda x: x.requires_grad, net.get_parameters()), learning_rate=0.01)
criterion = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean')
net_with_criterion = WithLossCell(net, criterion)
train_network = TrainOneStepCell(net_with_criterion, optimizer)
train_network.set_train()
context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
losses1 = []
for _ in range(epoch):
data = Tensor(np.arange(0, 16).reshape((1, 1, 4, 4)).astype(np.float32) * 0.01)
label = Tensor(np.array([0]).astype(np.int32))
loss = train_network(data, label)
losses1.append(loss.asnumpy())
assert losses1[0] > losses1[1]
assert losses1[1] > losses1[2]
context.set_context(mode=context.PYNATIVE_MODE, device_target="GPU")
losses2 = []
for _ in range(epoch):
data = Tensor(np.arange(0, 16).reshape((1, 1, 4, 4)).astype(np.float32) * 0.01)
label = Tensor(np.array([0]).astype(np.int32))
loss = train_network(data, label)
losses2.append(loss.asnumpy())
assert losses2[0] > losses2[1]
assert losses2[1] > losses2[2]
@pytest.mark.level0
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
def test_lazy_adam():
"""
Feature: LazyAdam optimizer
Description: Verify if the result is correct
Expectation: success
"""
indices = Tensor(np.array([0, 0, 1]).astype(np.int32))
label = Tensor(np.zeros([2, 1, 2]).astype(np.float32))
net = NetWithSparseGatherV2()
output = []
optimizer = Adam(net.trainable_params(), learning_rate=0.1, use_lazy=True)
optimizer.target = 'CPU'
for _ in range(2):
train_network = TrainOneStepCell(net, optimizer)
output = train_network(indices, label)
expected_output = np.array([[[1.8000001, 1.8000001]], [[1.8000001, 1.8000001]],
[[1.8000001, 1.8000001]]]).astype(np.float32)
assert np.allclose(output.asnumpy(), expected_output)
@pytest.mark.level0
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
def test_adam_offload_acc():
"""
Feature: AdamOffload optimizer
Description: Verify if the loss is the same as the original AdamOffload
Expectation: success
"""
epoch = 3
net = NetAdam()
optimizer = Adam(filter(lambda x: x.requires_grad,
net.get_parameters()), learning_rate=0.01, use_offload=True)
criterion = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean')
net_with_criterion = WithLossCell(net, criterion)
train_network = TrainOneStepCell(net_with_criterion, optimizer)
train_network.set_train()
context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
losses1 = []
for _ in range(epoch):
data = Tensor(np.arange(0, 16).reshape((1, 1, 4, 4)).astype(np.float32) * 0.01)
label = Tensor(np.array([0]).astype(np.int32))
loss = train_network(data, label)
losses1.append(loss.asnumpy())
assert np.array_equal(losses1[-1], np.array(2.2237475, np.float32))
def numpy_apply_adam(var, m, v, grad, beta1=0.9, beta2=0.999, eps=1e-8, lr=0.01):
new_lr = lr * math.sqrt(1 - beta2) / (1 - beta1)
m = m * beta1 + grad * (1 - beta1)
v = v * beta2 + grad * grad * (1 - beta2)
var = var - new_lr * m / (np.sqrt(v) + eps)
return var
class AdamNetVmap(nn.Cell):
def __init__(self, net):
super(AdamNetVmap, self).__init__()
shape = (8, 9, 6, 10, 5)
self.net = net
self.var_np = np.random.randn(*shape).astype(np.float32)
self.m_np = np.random.randn(*shape).astype(np.float32)
self.v_np = np.random.randn(*shape).astype(np.float32)
self.var = Parameter(Tensor(self.var_np), name="var")
self.m = Parameter(Tensor(self.m_np), name="m")
self.v = Parameter(Tensor(self.v_np), name="v")
self.vmap_adam = vmap(self.net, in_axes=(
0, 0, 0, None, None, None, None, None, None, 0), out_axes=0)
def construct(self, beta1_power, beta2_power, lr, beta1, beta2, epsilon, grad):
return self.vmap_adam(self.var, self.m, self.v, beta1_power, beta2_power, lr, beta1, beta2, epsilon, grad)
@pytest.mark.level1
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
def test_apply_adam_witm_adam_op_vmap():
"""
Feature: Adam gpu kernel
Description: test the Adam vmap.
Expectation: match to np benchmark.
"""
shape = (8, 9, 6, 10, 5)
def cal_amsgrad(var, m, v, beta1_power, beta2_power, lr, beta1, beta2, epsilon, grad):
return P.Adam()(var, m, v, beta1_power, beta2_power, lr, beta1, beta2, epsilon, grad)
error = 1e-4
grad_np = np.random.randn(*shape).astype(np.float32)
grad = Tensor(grad_np)
vmap_adam = AdamNetVmap(cal_amsgrad)
_ = vmap_adam(Tensor(0.9, ms.float32), Tensor(0.999, ms.float32), Tensor(0.01, ms.float32), Tensor(
0.9, ms.float32), Tensor(0.999, ms.float32), Tensor(1e-8, ms.float32), grad)
ms_var = vmap_adam.var.asnumpy()
np_var = numpy_apply_adam(vmap_adam.var_np, vmap_adam.m_np,
vmap_adam.v_np, grad_np)
np.testing.assert_allclose(ms_var, np_var, rtol=error, atol=error)