paddlepaddle--paddle/test/legacy_test/test_parallel_dygraph_dataparallel.py

# Copyright (c) 2021 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 copy
import os
import subprocess
import time
import unittest

from paddle import base
from paddle.distributed.utils.launch_utils import (
    TrainerProc,
    find_free_ports,
    get_cluster,
    watch_local_trainers,
)


def get_cluster_from_args(selected_devices):
    cluster_node_ips = '127.0.0.1'
    node_ip = '127.0.0.1'

    node_ips = [x.strip() for x in cluster_node_ips.split(',')]

    node_ips.index(node_ip)

    free_ports = None

    free_ports = find_free_ports(len(selected_devices))
    if free_ports is not None:
        free_ports = list(free_ports)

    trainer_endpoints = []
    for ip in node_ips:
        trainer_endpoints.append([f"{ip}:{port}" for port in free_ports])
    return get_cluster(node_ips, node_ip, trainer_endpoints, selected_devices)


def get_devices(selected_devices):
    selected_devices = [x.strip() for x in selected_devices.split(',')]
    return selected_devices


def start_local_trainers_cpu(
    trainer_endpoints, training_script, training_script_args, log_dir=None
):
    current_env = copy.copy(os.environ.copy())
    current_env.pop("http_proxy", None)
    current_env.pop("https_proxy", None)

    procs = []
    n_rank = len(trainer_endpoints)
    print(trainer_endpoints)
    for rank_id, endpoint in enumerate(trainer_endpoints):
        proc_env = {
            "PADDLE_DISTRI_BACKEND": "gloo",
            "PADDLE_TRAINER_ID": str(rank_id),
            "PADDLE_CURRENT_ENDPOINT": str(endpoint),
            "PADDLE_TRAINERS_NUM": str(n_rank),
            "PADDLE_TRAINER_ENDPOINTS": ",".join(trainer_endpoints),
        }

        current_env.update(proc_env)

        print(f"trainer proc env:{current_env}")

        assert os.getenv('WITH_COVERAGE', 'OFF') == 'OFF', (
            "Gloo don't support WITH_COVERAGE."
        )
        cmd = "python -u " + training_script

        print(f"start trainer proc:{cmd} env:{proc_env}")

        fn = None

        proc = subprocess.Popen(cmd.split(" "), env=current_env)

        tp = TrainerProc()
        tp.proc = proc
        tp.rank = rank_id
        tp.log_fn = fn
        tp.cmd = cmd

        procs.append(tp)

    return procs


def start_local_trainers(
    cluster,
    pod,
    training_script,
    training_script_args,
    allocator_strategy="auto_growth",
    log_dir=None,
    need_envs={},
    accelerator_type="gpu",
):
    current_env = copy.copy(os.environ.copy())
    # paddle broadcast ncclUniqueId use socket, and
    # proxy maybe make trainers unreachable, so delete them.
    # if we set them to "", grpc will log error message "bad uri"
    # so just delete them.
    current_env.pop("http_proxy", None)
    current_env.pop("https_proxy", None)

    procs = []
    for t in pod.trainers:
        proc_env = {
            f"FLAGS_selected_{accelerator_type}s": "{}".format(
                ",".join([str(g) for g in t.gpus])
            ),
            "PADDLE_TRAINER_ID": str(t.rank),
            "PADDLE_CURRENT_ENDPOINT": str(t.endpoint),
            "PADDLE_TRAINERS_NUM": str(cluster.trainers_nranks()),
            "PADDLE_TRAINER_ENDPOINTS": ",".join(cluster.trainers_endpoints()),
        }

        proc_env["FLAGS_allocator_strategy"] = allocator_strategy
        if allocator_strategy == "auto_growth":
            proc_env["FLAGS_fraction_of_gpu_memory_to_use"] = "0.1"

        current_env.update(proc_env)
        current_env.update(need_envs)

        print(f"trainer proc env:{current_env}")

        if os.getenv('WITH_COVERAGE', 'OFF') == 'ON':
            cmd = "python -m coverage run --branch -p " + training_script
        else:
            cmd = "python -u " + training_script

        print(f"start trainer proc:{cmd} env:{proc_env}")

        fn = None

        proc = subprocess.Popen(cmd.split(" "), env=current_env)

        tp = TrainerProc()
        tp.proc = proc
        tp.rank = t.rank
        tp.log_fn = fn
        tp.cmd = cmd

        procs.append(tp)

    return procs


class TestMultipleAccelerators(unittest.TestCase):
    def run_mnist_2accelerators(
        self,
        target_file_name,
        allocator_strategy="auto_growth",
        need_envs={},
        accelerator_type="xpu" if base.core.is_compiled_with_xpu() else "gpu",
    ):
        if accelerator_type == "gpu":
            if (
                not base.core.is_compiled_with_cuda()
                or base.core.get_cuda_device_count() == 0
            ):
                return
        elif accelerator_type == "xpu":
            if (
                not base.core.is_compiled_with_xpu()
                or base.core.get_xpu_device_count() == 0
            ):
                return
        else:
            if (
                not base.core.is_compiled_with_custom_device(accelerator_type)
                or base.core.get_custom_device_count(accelerator_type) == 0
            ):
                return

        selected_devices = get_devices('0,1')
        cluster = None
        pod = None

        cluster, pod = get_cluster_from_args(selected_devices)

        procs = start_local_trainers(
            cluster,
            pod,
            allocator_strategy=allocator_strategy,
            training_script=target_file_name,
            training_script_args=[],
            need_envs=need_envs,
            accelerator_type=accelerator_type,
        )

        while True:
            alive = watch_local_trainers(procs, cluster.trainers_endpoints())

            if not alive:
                print(f"Local procs complete, POD info:{pod}")
                break
            time.sleep(3)


class TestMultipleWithGloo(unittest.TestCase):
    def run_mnist_2cpu(self, target_file_name):
        cluster, pod = get_cluster_from_args(
            [0, 1]
        )  # tmp use. for getting trainer_nranks()

        procs = start_local_trainers_cpu(
            cluster.trainers_endpoints(),
            training_script=target_file_name,
            training_script_args=[],
        )

        while True:
            alive = watch_local_trainers(procs, cluster.trainers_nranks())

            if not alive:
                print(f"Local procs complete, POD info:{pod}")
                break
            time.sleep(3)


class TestDataParallelWithPyLayer(TestMultipleAccelerators):
    def test_parallel_dygraph_dataparallel_with_pylayer(self):
        self.run_mnist_2accelerators(
            'parallel_dygraph_dataparallel_with_pylayer.py'
        )
        self.run_mnist_2accelerators(
            'parallel_dygraph_dataparallel_with_pylayer.py',
            allocator_strategy="naive_best_fit",
        )


class TestGradientCheckInEagerMode(TestMultipleAccelerators):
    def test_multiple_gpus_dynamic(self):
        self.run_mnist_2accelerators(
            'parallel_dygraph_gradient_check_in_eager_mode.py'
        )


if __name__ == "__main__":
    unittest.main()