wehub-resource-sync
145 lines
4.8 KiB
Python
145 lines
4.8 KiB
Python
# Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
|
|
#
|
|
# 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 os
|
|
import unittest
|
|
|
|
import numpy as np
|
|
|
|
import paddle
|
|
import paddle.distributed as dist
|
|
from paddle.base import core
|
|
from paddle.distributed.auto_parallel.placement_type import (
|
|
check_placements_equal,
|
|
)
|
|
|
|
|
|
class TestDistTensorSRP(unittest.TestCase):
|
|
def setUp(self):
|
|
self._shape = eval(os.getenv("shape"))
|
|
self._dtype = os.getenv("dtype")
|
|
self._seed = 2023
|
|
self._backend = os.getenv("backend")
|
|
self._mesh = dist.ProcessMesh([0, 1], dim_names=["x"])
|
|
paddle.seed(self._seed)
|
|
np.random.seed(self._seed)
|
|
|
|
def run_test_placements(self):
|
|
self.placements = [core.Replicate(), core.Replicate()]
|
|
shard = core.Shard(1)
|
|
replicate = core.Replicate()
|
|
partial = core.Partial()
|
|
|
|
self.assertEqual(shard.get_dim(), 1)
|
|
self.assertEqual(shard.is_shard(), True)
|
|
self.assertEqual(shard.is_replicated(), False)
|
|
self.assertEqual(shard.is_partial(), False)
|
|
self.assertEqual(str(shard), "Shard(dim=1)")
|
|
|
|
self.assertEqual(replicate.is_shard(), False)
|
|
self.assertEqual(replicate.is_replicated(), True)
|
|
self.assertEqual(replicate.is_partial(), False)
|
|
self.assertEqual(str(replicate), "Replicate()")
|
|
|
|
self.assertEqual(partial.is_shard(), False)
|
|
self.assertEqual(partial.is_replicated(), False)
|
|
self.assertEqual(partial.is_partial(), True)
|
|
self.assertEqual(str(partial), "Partial(reduce_type=SUM)")
|
|
|
|
shard_1 = core.Shard(1)
|
|
replicate_1 = core.Replicate()
|
|
partial_1 = core.Partial()
|
|
|
|
self.assertEqual(shard_1, shard)
|
|
self.assertEqual(replicate_1, replicate)
|
|
self.assertEqual(partial_1, partial)
|
|
self.assertNotEqual(shard_1, replicate)
|
|
self.assertNotEqual(shard_1, partial)
|
|
|
|
self.assertEqual(hash(shard_1), hash(shard))
|
|
self.assertEqual(hash(replicate_1), hash(replicate))
|
|
self.assertEqual(hash(partial_1), hash(partial))
|
|
|
|
def run_test_check_placements_equal(self):
|
|
this_placements = [dist.Shard(1), dist.Replicate()]
|
|
that_placements = [dist.Shard(1)]
|
|
self.assertTrue(
|
|
check_placements_equal(this_placements, that_placements)
|
|
)
|
|
self.assertTrue(
|
|
check_placements_equal(that_placements, this_placements)
|
|
)
|
|
that_placements = [dist.Shard(1), dist.Replicate(), dist.Shard(0)]
|
|
self.assertFalse(
|
|
check_placements_equal(this_placements, that_placements)
|
|
)
|
|
that_placements = [dist.Shard(0), dist.Replicate()]
|
|
self.assertFalse(
|
|
check_placements_equal(this_placements, that_placements)
|
|
)
|
|
that_placements = [dist.Replicate(), dist.Shard(1)]
|
|
self.assertFalse(
|
|
check_placements_equal(this_placements, that_placements)
|
|
)
|
|
|
|
def run_test_dist_tensor(self):
|
|
if self._backend == "cpu":
|
|
paddle.set_device("cpu")
|
|
place = paddle.CPUPlace()
|
|
elif self._backend == "gpu":
|
|
place = paddle.CUDAPlace(dist.get_rank())
|
|
|
|
tensor = paddle.rand([2, 10])
|
|
dist_tensor = paddle.Tensor(
|
|
tensor, process_mesh=self._mesh, placements=[dist.Shard(0)]
|
|
)
|
|
|
|
self.assertEqual(dist_tensor.num_shard, 2)
|
|
|
|
np.testing.assert_equal(
|
|
dist_tensor.numpy(),
|
|
tensor.numpy(),
|
|
)
|
|
|
|
def run_test_dist_tensor_with_local_tensor(self):
|
|
if self._backend == "cpu":
|
|
paddle.set_device("cpu")
|
|
place = paddle.CPUPlace()
|
|
elif self._backend == "gpu":
|
|
place = paddle.CUDAPlace(dist.get_rank())
|
|
|
|
tensor = paddle.rand([2, 10])
|
|
global_dims = [4, 10]
|
|
dist_tensor = paddle.Tensor(
|
|
tensor,
|
|
dims=global_dims,
|
|
process_mesh=self._mesh,
|
|
placements=core.Shard(0),
|
|
)
|
|
|
|
np.testing.assert_equal(
|
|
dist_tensor._local_value().numpy(),
|
|
tensor.numpy(),
|
|
)
|
|
|
|
def test_case(self):
|
|
self.run_test_placements()
|
|
self.run_test_check_placements_equal()
|
|
self.run_test_dist_tensor()
|
|
self.run_test_dist_tensor_with_local_tensor()
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|