a /Sicd'@sdZddlZddlmZddlmZddlmZddlm Z ddl m Z ddl m Z dd lmZdd lmZdd lmZdd lmZdd lmZdZdZeddddZeddddZddZdS)z%Helpers to connect to remote servers.N)logging) ServerDef) pywrap_tfe) device_util)cluster_resolver)context)ops) remote_utils) server_lib)nest) tf_exportzgrpc://)localz#config.experimental_connect_to_hostworkercCs:|s tdt|}t|dd|Di}t|dS)aConnects to a single machine to enable remote execution on it. Will make devices on the remote host available to use. Note that calling this more than once will work, but will invalidate any tensor handles on the old remote devices. Using the default job_name of worker, you can schedule ops to run remotely as follows: ```python # When eager execution is enabled, connect to the remote host. tf.config.experimental_connect_to_host("exampleaddr.com:9876") with ops.device("job:worker/replica:0/task:1/device:CPU:0"): # The following tensors should be resident on the remote device, and the op # will also execute remotely. x1 = array_ops.ones([2, 2]) x2 = array_ops.ones([2, 2]) y = math_ops.matmul(x1, x2) ``` Args: remote_host: a single or a list the remote server addr in host-port format. job_name: The job name under which the new server will be accessible. Raises: ValueError: if remote_host is None. z%Must provide at least one remote_hostcSsg|]}t|tqS) _strip_prefix _GRPC_PREFIX).0hostrrZ/var/www/html/django/DPS/env/lib/python3.9/site-packages/tensorflow/python/eager/remote.py Gz*connect_to_remote_host..N) ValueErrorr flattenr ClusterSpecconnect_to_cluster)Z remote_hostjob_nameZ remote_hosts cluster_specrrrconnect_to_remote_host%s rz&config.experimental_connect_to_cluster localhostTcCststd|pt}t|tjr.|}n.t|tj rT| t vrJdS| }ntdt |}|rt|tjrt |}ntdd}t} | r||jvrt| jjD]\} } | j|kr| jj| =q| dus| j|ks| j|ks| j|krd}||jvr6t} |j} || _d| | jd<tjdurdt |}|rdt!|t"||||tj#|d }|rt$|n t%||rt|tj r| r| }d}d}|jD]F}|&|D]4}|'||}||vs||vr|}|}qΐqܐq|s(td |d ||}t()|}t(*}|rVt()|}|rv||krvtd ||f|st+,d |t-.|/dS)a) Connects to the given cluster. Will make devices on the cluster available to use. Note that calling this more than once will work, but will invalidate any tensor handles on the old remote devices. If the given local job name is not present in the cluster specification, it will be automatically added, using an unused port on the localhost. Device filters can be specified to isolate groups of remote tasks to avoid undesired accesses between workers. Workers accessing resources or launching ops / functions on filtered remote devices will result in errors (unknown devices). For any remote task, if no device filter is present, all cluster devices will be visible; if any device filter is specified, it can only see devices matching at least one filter. Devices on the task itself are always visible. Device filters can be particially specified. For example, for a cluster set up for parameter server training, the following device filters might be specified: ```python cdf = tf.config.experimental.ClusterDeviceFilters() # For any worker, only the devices on PS nodes and itself are visible for i in range(num_workers): cdf.set_device_filters('worker', i, ['/job:ps']) # Similarly for any ps, only the devices on workers and itself are visible for i in range(num_ps): cdf.set_device_filters('ps', i, ['/job:worker']) tf.config.experimental_connect_to_cluster(cluster_def, cluster_device_filters=cdf) ``` Args: cluster_spec_or_resolver: A `ClusterSpec` or `ClusterResolver` describing the cluster. job_name: The name of the local job. task_index: The local task index. protocol: The communication protocol, such as `"grpc"`. If unspecified, will use the default from `python/platform/remote_utils.py`. make_master_device_default: If True and a cluster resolver is passed, will automatically enter the master task device scope, which indicates the master becomes the default device to run ops. It won't do anything if a cluster spec is passed. Will throw an error if the caller is currently already in some device scope. cluster_device_filters: an instance of `tf.train.experimental/ClusterDeviceFilters` that specify device filters to the remote tasks in cluster. zM`tf.config.experimental_connect_to_cluster` can only be called in eager mode.NzJ`cluster_spec_or_resolver` must be a `ClusterSpec` or a `ClusterResolver`.z]`cluster_device_filters` must be an instance of `tf.train.experimental.ClusterDeviceFilters`.FTz localhost:{}r)clusterr task_indexprotocoldefault_session_configcluster_device_filterszT`make_master_device_default` is set to True but cannot find master %s in the clusterz/job:{}/replica:0/task:{}z`connect_to_cluster` is called inside existing device scope %s, which is different from the master device scope %s to enter. This is not allowed.z%Entering into master device scope: %s)0rexecuting_eagerlyrr Z"get_default_communication_protocol isinstancer rrClusterResolvermaster_LOCAL_MASTERSrcopydeepcopyas_cluster_defClusterDeviceFilters_as_cluster_device_filtersget_server_defjobs enumerater jobnamerr!rTF_PickUnusedPortOrDieaddformattaskscoordination_serviceZcoordination_service_typeconfigure_coordination_servicerconfigset_server_defupdate_server_def task_indices task_addressr canonicalizecurrentrinfordevice __enter__)Zcluster_spec_or_resolverrr!r"Zmake_master_device_defaultr$r cluster_defZis_server_def_changedZcurrent_server_defir2Z local_portjob_defr8 server_defr(Zmaster_job_nameZmaster_task_idtask_idr>Z master_devicecurrent_devicerrrrLs8                  rcCs||r|t|dS|S)N) startswithlen)sprefixrrrrsr)Nr)rrNTN)__doc__r*abslrZ.tensorflow.core.protobuf.tensorflow_server_pb2rtensorflow.pythonrtensorflow.python.distributer-tensorflow.python.distribute.cluster_resolverrtensorflow.python.eagerrtensorflow.python.frameworkrtensorflow.python.platformr tensorflow.python.trainingr tensorflow.python.utilr tensorflow.python.util.tf_exportr rr)rrrrrrrs2            & !