fix selected_gpus bug for local_cluster (#205)

* fix selected_gpus bug for local_cluster

* fix bugs about single machine multicards training for local_cluster

* fix python -m config.yaml -b backend.yaml

* re add selected_gpus_num config for single train mode

* fix bug for submit for paddlecloud k8s

Co-authored-by: wuzhihua <[email protected]>

Author: vslyu

core/engine/cluster/cluster.py

@@ -19,10 +19,16 @@
 import os
 import subprocess
 import warnings
+import sys
+import logging
 
 from paddlerec.core.engine.engine import Engine
 from paddlerec.core.factory import TrainerFactory
 from paddlerec.core.utils import envs
+import paddlerec.core.engine.cluster_utils as cluster_utils
+
+logger = logging.getLogger("root")
+logger.propagate = False
 
 
 class ClusterEngine(Engine):
@@ -47,8 +53,38 @@ def __init_impl__(self):
                 self.backend))
 
     def start_worker_procs(self):
-        trainer = TrainerFactory.create(self.trainer)
-        trainer.run()
+        if (envs.get_runtime_environ("fleet_mode") == "COLLECTIVE"):
+            #trainer_ports = os.getenv("TRAINER_PORTS", None).split(",")                                            
+            cuda_visible_devices = os.getenv("CUDA_VISIBLE_DEVICES")
+            if cuda_visible_devices is None or cuda_visible_devices == "":
+                selected_gpus = range(int(os.getenv("TRAINER_GPU_CARD_COUNT")))
+            else:
+                # change selected_gpus into relative values                                                         
+                # e.g. CUDA_VISIBLE_DEVICES=4,5,6,7; args.selected_gpus=4,5,6,7;                                    
+                # therefore selected_gpus=0,1,2,3                                                                   
+                cuda_visible_devices_list = cuda_visible_devices.split(',')
+                for x in range(int(os.getenv("TRAINER_GPU_CARD_COUNT"))):
+                    assert x in cuda_visible_devices_list, "Can't find "\
+                    "your selected_gpus %s in CUDA_VISIBLE_DEVICES[%s]."\
+                    % (x, cuda_visible_devices)
+                selected_gpus = [cuda_visible_devices_list.index(x)]
+            print("selected_gpus:{}".format(selected_gpus))
+
+            factory = "paddlerec.core.factory"
+            cmd = [sys.executable, "-u", "-m", factory, self.trainer]
+            logs_dir = envs.get_runtime_environ("log_dir")
+            print("use_paddlecloud_flag:{}".format(
+                cluster_utils.use_paddlecloud()))
+            if cluster_utils.use_paddlecloud():
+                cluster, pod = cluster_utils.get_cloud_cluster(selected_gpus)
+                logger.info("get cluster from cloud:{}".format(cluster))
+                procs = cluster_utils.start_local_trainers(
+                    cluster, pod, cmd, log_dir=logs_dir)
+                print("cluster:{}".format(cluster))
+                print("pod:{}".format(pod))
+        else:
+            trainer = TrainerFactory.create(self.trainer)
+            trainer.run()
 
     def start_master_procs(self):
         if self.backend == "PADDLECLOUD":

core/engine/cluster_utils.py

@@ -0,0 +1,324 @@
+# 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 functools
+import logging
+import socket
+import time
+import os
+import signal
+import copy
+import sys
+import subprocess
+from contextlib import closing
+import socket
+
+logger = logging.getLogger("root")
+logger.propagate = False
+
+
+class Cluster(object):
+    def __init__(self, hdfs):
+        self.job_server = None
+        self.pods = []
+        self.hdfs = None
+        self.job_stage_flag = None
+
+    def __str__(self):
+        return "job_server:{} pods:{} job_stage_flag:{} hdfs:{}".format(
+            self.job_server, [str(pod) for pod in self.pods],
+            self.job_stage_flag, self.hdfs)
+
+    def __eq__(self, cluster):
+        if len(self.pods) != len(cluster.pods):
+            return False
+
+        for a, b in zip(self.pods, cluster.pods):
+            if a != b:
+                return False
+
+        if self.job_stage_flag != cluster.job_stage_flag:
+            return False
+
+        return True
+
+    def __ne__(self, cluster):
+        return not self.__eq__(cluster)
+
+    def update_pods(cluster):
+        self.pods = copy.copy(cluster.pods)
+
+    def trainers_nranks(self):
+        return len(self.trainers_endpoints())
+
+    def pods_nranks(self):
+        return len(self.pods)
+
+    def trainers_endpoints(self):
+        r = []
+        for pod in self.pods:
+            for t in pod.trainers:
+                r.append(t.endpoint)
+        return r
+
+    def pods_endpoints(self):
+        r = []
+        for pod in self.pods:
+            ep = "{}:{}".format(pod.addr, pod.port)
+            assert pod.port != None and pod.addr != None, "{} not a valid endpoint".format(
+                ep)
+            r.append(ep)
+
+        return r
+
+    def get_pod_by_id(self, pod_id):
+        for pod in self.pods:
+            if str(pod_id) == str(pod.id):
+                return pod
+
+        return None
+
+
+class JobServer(object):
+    def __init__(self):
+        self.endpoint = None
+
+    def __str__(self):
+        return "{}".format(self.endpoint)
+
+    def __eq__(self, j):
+        return self.endpint == j.endpoint
+
+    def __ne__(self, j):
+        return not self == j
+
+
+class Trainer(object):
+    def __init__(self):
+        self.gpus = []
+        self.endpoint = None
+        self.rank = None
+
+    def __str__(self):
+        return "gpu:{} endpoint:{} rank:{}".format(self.gpus, self.endpoint,
+                                                   self.rank)
+
+    def __eq__(self, t):
+        if len(self.gpus) != len(t.gpus):
+            return False
+
+        if self.endpoint != t.endpoint or \
+                self.rank != t.rank:
+            return False
+
+        for a, b in zip(self.gpus, t.gpus):
+            if a != b:
+                return False
+
+        return True
+
+    def __ne__(self, t):
+        return not self == t
+
+    def rank(self):
+        return self.rank
+
+
+class Pod(object):
+    def __init__(self):
+        self.rank = None
+        self.id = None
+        self.addr = None
+        self.port = None
+        self.trainers = []
+        self.gpus = []
+
+    def __str__(self):
+        return "rank:{} id:{} addr:{} port:{} visible_gpu:{} trainers:{}".format(
+            self.rank, self.id, self.addr, self.port, self.gpus,
+            [str(t) for t in self.trainers])
+
+    def __eq__(self, pod):
+        if self.rank != pod.rank or \
+                self.id != pod.id or \
+                self.addr != pod.addr or \
+                self.port != pod.port:
+            logger.debug("pod {} != pod".format(self, pod))
+            return False
+
+        if len(self.trainers) != len(pod.trainers):
+            logger.debug("trainers {} != {}".format(self.trainers,
+                                                    pod.trainers))
+            return False
+
+        for i in range(len(self.trainers)):
+            if self.trainers[i] != pod.trainers[i]:
+                logger.debug("trainer {} != {}".format(self.trainers[i],
+                                                       pod.trainers[i]))
+                return False
+
+        return True
+
+    def __ne__(self, pod):
+        return not self == pod
+
+    def parse_response(self, res_pods):
+        pass
+
+    def rank(self):
+        return self.rank
+
+    def get_visible_gpus(self):
+        r = ""
+        for g in self.gpus:
+            r += "{},".format(g)
+
+        assert r != "", "this pod {} can't see any gpus".format(self)
+
+        r = r[:-1]
+        return r
+
+
+def get_cluster(node_ips, node_ip, paddle_ports, selected_gpus):
+    assert type(paddle_ports) is list, "paddle_ports must be list"
+    cluster = Cluster(hdfs=None)
+    trainer_rank = 0
+    for node_rank, ip in enumerate(node_ips):
+        pod = Pod()
+        pod.rank = node_rank
+        pod.addr = ip
+        for i in range(len(selected_gpus)):
+            trainer = Trainer()
+            trainer.gpus.append(selected_gpus[i])
+            trainer.endpoint = "%s:%d" % (ip, paddle_ports[i])
+            trainer.rank = trainer_rank
+            trainer_rank += 1
+
+            pod.trainers.append(trainer)
+        cluster.pods.append(pod)
+
+    pod_rank = node_ips.index(node_ip)
+    return cluster, cluster.pods[pod_rank]
+
+
+def get_cloud_cluster(selected_gpus, args_port=None):
+    #you can automatically get ip info while using paddlecloud multi nodes mode.
+    node_ips = os.getenv("PADDLE_TRAINERS")
+    assert node_ips is not None, "PADDLE_TRAINERS should not be None"
+    print("node_ips:{}".format(node_ips))
+    node_ip = os.getenv("POD_IP")
+    assert node_ip is not None, "POD_IP should not be None"
+    print("node_ip:{}".format(node_ip))
+    node_rank = os.getenv("PADDLE_TRAINER_ID")
+    assert node_rank is not None, "PADDLE_TRAINER_ID should not be None"
+    print("node_rank:{}".format(node_rank))
+    node_ips = node_ips.split(",")
+    num_nodes = len(node_ips)
+    node_rank = int(node_rank)
+
+    started_port = args_port
+    print("num_nodes:", num_nodes)
+    if num_nodes > 1:
+        try:
+            paddle_port = int(os.getenv("PADDLE_PORT", ""))
+            paddle_port_num = int(os.getenv("TRAINER_PORTS_NUM", ""))
+
+            if paddle_port_num >= len(
+                    selected_gpus) and paddle_port != args_port:
+                logger.warning("Use Cloud specified port:{}.".format(
+                    paddle_port))
+                started_port = paddle_port
+
+        except Exception as e:
+            print(e)
+            pass
+
+    if started_port is None:
+        started_port = 6170
+
+    logger.debug("parsed from args:node_ips:{} \
+        node_ip:{} node_rank:{} started_port:{}"
+                 .format(node_ips, node_ip, node_rank, started_port))
+
+    ports = [x for x in range(started_port, started_port + len(selected_gpus))]
+    cluster, pod = get_cluster(node_ips, node_ip, ports, selected_gpus)
+    return cluster, cluster.pods[node_rank]
+
+
+def use_paddlecloud():
+    node_ips = os.getenv("PADDLE_TRAINERS", None)
+    node_ip = os.getenv("POD_IP", None)
+    node_rank = os.getenv("PADDLE_TRAINER_ID", None)
+    if node_ips is None or node_ip is None or node_rank is None:
+        return False
+    else:
+        return True
+
+
+class TrainerProc(object):
+    def __init__(self):
+        self.proc = None
+        self.log_fn = None
+        self.log_offset = None
+        self.rank = None
+        self.local_rank = None
+        self.cmd = None
+
+
+def start_local_trainers(cluster, pod, cmd, log_dir=None):
+    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 idx, t in enumerate(pod.trainers):
+        proc_env = {
+            "FLAGS_selected_gpus": "%s" % ",".join([str(g) for g in t.gpus]),
+            "PADDLE_TRAINER_ID": "%d" % t.rank,
+            "PADDLE_CURRENT_ENDPOINT": "%s" % t.endpoint,
+            "PADDLE_TRAINERS_NUM": "%d" % cluster.trainers_nranks(),
+            "PADDLE_TRAINER_ENDPOINTS": ",".join(cluster.trainers_endpoints())
+        }
+
+        current_env.update(proc_env)
+
+        logger.debug("trainer proc env:{}".format(current_env))
+
+        # cmd = [sys.executable, "-u", training_script]
+
+        logger.info("start trainer proc:{} env:{}".format(cmd, proc_env))
+
+        fn = None
+        if log_dir is not None:
+            os.system("mkdir -p {}".format(log_dir))
+            fn = open("%s/workerlog.%d" % (log_dir, idx), "a")
+            proc = subprocess.Popen(cmd, env=current_env, stdout=fn, stderr=fn)
+        else:
+            proc = subprocess.Popen(cmd, env=current_env)
+
+        tp = TrainerProc()
+        tp.proc = proc
+        tp.rank = t.rank
+        tp.local_rank = idx
+        tp.log_fn = fn
+        tp.log_offset = fn.tell() if fn else None
+        tp.cmd = cmd
+
+        procs.append(proc)
+
+    return procs