wehub-resource-sync
111 lines
3.6 KiB
Python
111 lines
3.6 KiB
Python
# Copyright (c) 2020 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
|
|
|
|
from auto_checkpoint_utils import get_logger
|
|
from test_auto_checkpoint import AutoCheckPointACLBase
|
|
|
|
import paddle
|
|
import paddle.base.incubate.checkpoint.auto_checkpoint as acp
|
|
from paddle import base
|
|
from paddle.distributed.fleet.utils.fs import HDFSClient, LocalFS
|
|
from paddle.incubate.distributed.fleet import role_maker
|
|
from paddle.incubate.distributed.fleet.collective import fleet
|
|
|
|
paddle.enable_static()
|
|
logger = get_logger()
|
|
|
|
|
|
class AutoCheckpointTestDist(AutoCheckPointACLBase):
|
|
def setUp(self):
|
|
get_logger()
|
|
logger.info("enter tests")
|
|
|
|
self._old_environ = dict(os.environ)
|
|
proc_env = {
|
|
"PADDLE_RUNNING_ENV": "PADDLE_EDL_AUTO_CHECKPOINT",
|
|
"PADDLE_TRAINER_ID": "0",
|
|
"PADDLE_RUNNING_PLATFORM": "PADDLE_CLOUD",
|
|
"PADDLE_JOB_ID": "test_job_auto_dist_basic",
|
|
"PADDLE_EDL_HDFS_HOME": "/usr/local/hadoop-2.7.7",
|
|
"PADDLE_EDL_HDFS_NAME": "",
|
|
"PADDLE_EDL_HDFS_UGI": "",
|
|
"PADDLE_EDL_HDFS_CHECKPOINT_PATH": "auto_checkpoint_dist_basic",
|
|
"PADDLE_EDL_ONLY_FOR_CE_TEST": "1",
|
|
"PADDLE_EDL_FS_CACHE": ".auto_checkpoint_test_dist_basic",
|
|
"PADDLE_EDL_SAVE_CHECKPOINT_INTER": "0",
|
|
}
|
|
os.environ.update(proc_env)
|
|
|
|
def test_distributed_basic(self):
|
|
checker = acp._get_checker()
|
|
fs = HDFSClient(checker.hdfs_home, None)
|
|
fs.delete(checker.hdfs_checkpoint_path)
|
|
self._reset_generator()
|
|
|
|
logger.info("begin test_distributed_basic")
|
|
fs = LocalFS()
|
|
save_dir = "./run_save_0"
|
|
fs.delete(save_dir)
|
|
|
|
# basic
|
|
exe, main_prog, startup_prog = self._generate()
|
|
|
|
compiled, data_loader, optimizer, loss, image, label = self._init_env(
|
|
exe, main_prog, startup_prog, minimize=False
|
|
)
|
|
|
|
# fleet
|
|
os.environ["TRAINING_ROLE"] = "TRAINER"
|
|
os.environ["PADDLE_TRAINER_ID"] = "0"
|
|
os.environ["PADDLE_TRAINER_ENDPOINTS"] = "127.0.0.1:6070"
|
|
|
|
role = role_maker.PaddleCloudRoleMaker(is_collective=True)
|
|
fleet.init(role)
|
|
|
|
with base.program_guard(main_prog, startup_prog):
|
|
dist_optimizer = fleet.distributed_optimizer(optimizer)
|
|
dist_optimizer.minimize(loss)
|
|
|
|
exe.run(startup_prog)
|
|
|
|
o = None
|
|
i = 0
|
|
name = None
|
|
for i in acp.train_epoch_range(3, 0):
|
|
o = acp._get_train_epoch_range()
|
|
name = o.name
|
|
logger.info(f"_run_save_0 name:{o.name} epoch_no:{i}")
|
|
|
|
for data in data_loader():
|
|
fetch = exe.run(
|
|
fleet.main_program, feed=data, fetch_list=[loss]
|
|
)
|
|
|
|
self.assertEqual(len(o._exe_status), 1)
|
|
|
|
o = acp._get_train_epoch_range()
|
|
assert o is None, "now train epoch must not exits now"
|
|
self.assertEqual(i, 2)
|
|
|
|
fs.delete(save_dir)
|
|
|
|
logger.info("end test_distributed_basic")
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|