パブリック静的最終クラスConfigProto.Builder
Session configuration parameters. The system picks appropriate values for fields that are not set.Protobuf 型
tensorflow.ConfigProto パブリックメソッド
| ConfigProto.Builder | addAllDeviceFilters (Iterable<String> 値) When any filters are present sessions will ignore all devices which do not match the filters. |
| ConfigProto.Builder | addAllSessionInterOpThreadPool (Iterable<? extends ThreadPoolOptionProto > 値) This option is experimental - it may be replaced with a different mechanism in the future. |
| ConfigProto.Builder | addDeviceFilters (文字列値) When any filters are present sessions will ignore all devices which do not match the filters. |
| ConfigProto.Builder | addDeviceFiltersBytes (com.google.protobuf.ByteString 値) When any filters are present sessions will ignore all devices which do not match the filters. |
| ConfigProto.Builder | addRepeatedField (com.google.protobuf.Descriptors.FieldDescriptor フィールド、オブジェクト値) |
| ConfigProto.Builder | addSessionInterOpThreadPool ( ThreadPoolOptionProto値) This option is experimental - it may be replaced with a different mechanism in the future. |
| ConfigProto.Builder | addSessionInterOpThreadPool (int インデックス、 ThreadPoolOptionProto値) This option is experimental - it may be replaced with a different mechanism in the future. |
| ConfigProto.Builder | addSessionInterOpThreadPool ( ThreadPoolOptionProto.Builder builderForValue) This option is experimental - it may be replaced with a different mechanism in the future. |
| ConfigProto.Builder | addSessionInterOpThreadPool (int インデックス、 ThreadPoolOptionProto.Builder builderForValue) This option is experimental - it may be replaced with a different mechanism in the future. |
| ThreadPoolOptionProto.Builder | addSessionInterOpThreadPoolBuilder (int インデックス) This option is experimental - it may be replaced with a different mechanism in the future. |
| ThreadPoolOptionProto.Builder | addSessionInterOpThreadPoolBuilder () This option is experimental - it may be replaced with a different mechanism in the future. |
| コンフィグプロト | 建てる（） |
| コンフィグプロト | buildPartial () |
| ConfigProto.Builder | クリア（） |
| ConfigProto.Builder | ClearAllowSoftPlacement () Whether soft placement is allowed. |
| ConfigProto.Builder | ClearClusterDef () Optional list of all workers to use in this session. |
| ConfigProto.Builder | |
| ConfigProto.Builder | クリアデバイスフィルター() When any filters are present sessions will ignore all devices which do not match the filters. |
| ConfigProto.Builder | クリア実験的() .tensorflow.ConfigProto.Experimental experimental = 16; |
| ConfigProto.Builder | clearField (com.google.protobuf.Descriptors.FieldDescriptor フィールド) |
| ConfigProto.Builder | ClearGpuOptions () Options that apply to all GPUs. |
| ConfigProto.Builder | クリアグラフオプション() Options that apply to all graphs. |
| ConfigProto.Builder | クリアInterOpParallelismThreads () Nodes that perform blocking operations are enqueued on a pool of inter_op_parallelism_threads available in each process. |
| ConfigProto.Builder | clearIntraOpParallelismThreads () The execution of an individual op (for some op types) can be parallelized on a pool of intra_op_parallelism_threads. |
| ConfigProto.Builder | clearIsolateSessionState () If true, any resources such as Variables used in the session will not be shared with other sessions. |
| ConfigProto.Builder | ClearLogDevicePlacement () Whether device placements should be logged. |
| ConfigProto.Builder | clearOneof (com.google.protobuf.Descriptors.OneofDescriptor oneof) |
| ConfigProto.Builder | clearOperationTimeoutInMs () Global timeout for all blocking operations in this session. |
| ConfigProto.Builder | クリア配置期間() Assignment of Nodes to Devices is recomputed every placement_period steps until the system warms up (at which point the recomputation typically slows down automatically). |
| ConfigProto.Builder | ClearRpcOptions () Options that apply when this session uses the distributed runtime. |
| ConfigProto.Builder | ClearSessionInterOpThreadPool () This option is experimental - it may be replaced with a different mechanism in the future. |
| ConfigProto.Builder | clearShareClusterDevicesInSession () When true, WorkerSessions are created with device attributes from the full cluster. |
| ConfigProto.Builder | ClearUsePerSessionThreads () If true, use a new set of threads for this session rather than the global pool of threads. |
| ConfigProto.Builder | クローン() |
| ブール値 | containsDeviceCount (文字列キー) Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. |
| ブール値 | getAllowSoftPlacement () Whether soft placement is allowed. |
| クラスター定義 | getClusterDef () Optional list of all workers to use in this session. |
| ClusterDef.Builder | getClusterDefBuilder () Optional list of all workers to use in this session. |
| ClusterDefOrBuilder | getClusterDefOrBuilder () Optional list of all workers to use in this session. |
| コンフィグプロト | |
| 最終的な静的 com.google.protobuf.Descriptors.Descriptor | |
| com.google.protobuf.Descriptors.Descriptor | |
| Map<文字列、整数> | getDeviceCount () 代わりに getDeviceCountMap()を使用してください。 |
| 整数 | getDeviceCountCount () Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. |
| Map<文字列、整数> | getDeviceCountMap () Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. |
| 整数 | getDeviceCountOrDefault (文字列キー、int デフォルト値) Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. |
| 整数 | getDeviceCountOrThrow (文字列キー) Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. |
| 弦 | getDeviceFilters (int インデックス) When any filters are present sessions will ignore all devices which do not match the filters. |
| com.google.protobuf.ByteString | getDeviceFiltersBytes (int インデックス) When any filters are present sessions will ignore all devices which do not match the filters. |
| 整数 | getDeviceFiltersCount () When any filters are present sessions will ignore all devices which do not match the filters. |
| com.google.protobuf.ProtocolStringList | getDeviceFiltersList () When any filters are present sessions will ignore all devices which do not match the filters. |
| ConfigProto.Experimental | get実験的() .tensorflow.ConfigProto.Experimental experimental = 16; |
| ConfigProto.Experimental.Builder | getExperimentalBuilder () .tensorflow.ConfigProto.Experimental experimental = 16; |
| ConfigProto.ExperimentalOrBuilder | getExperimentalOrBuilder () .tensorflow.ConfigProto.Experimental experimental = 16; |
| GPUオプション | getGpuOptions () Options that apply to all GPUs. |
| GPUOptions.Builder | getGpuOptionsBuilder () Options that apply to all GPUs. |
| GPUオプションまたはビルダー | getGpuOptionsOrBuilder () Options that apply to all GPUs. |
| グラフオプション | getGraphOptions () Options that apply to all graphs. |
| グラフオプション.ビルダー | getGraphOptionsBuilder () Options that apply to all graphs. |
| グラフオプションまたはビルダー | getGraphOptionsOrBuilder () Options that apply to all graphs. |
| 整数 | getInterOpParallelismThreads () Nodes that perform blocking operations are enqueued on a pool of inter_op_parallelism_threads available in each process. |
| 整数 | getIntraOpParallelismThreads () The execution of an individual op (for some op types) can be parallelized on a pool of intra_op_parallelism_threads. |
| ブール値 | getIsolateSessionState () If true, any resources such as Variables used in the session will not be shared with other sessions. |
| ブール値 | getLogDevicePlacement () Whether device placements should be logged. |
| Map<文字列、整数> | getMutableDeviceCount () 代わりに代替の突然変異アクセサーを使用してください。 |
| 長さ | getOperationTimeoutInMs () Global timeout for all blocking operations in this session. |
| 整数 | getPlacementPeriod () Assignment of Nodes to Devices is recomputed every placement_period steps until the system warms up (at which point the recomputation typically slows down automatically). |
| RPCオプション | getRpcOptions () Options that apply when this session uses the distributed runtime. |
| RPCOptions.Builder | getRpcOptionsBuilder () Options that apply when this session uses the distributed runtime. |
| RPCOptionsOrBuilder | getRpcOptionsOrBuilder () Options that apply when this session uses the distributed runtime. |
| スレッドプールオプションプロト | getSessionInterOpThreadPool (int インデックス) This option is experimental - it may be replaced with a different mechanism in the future. |
| ThreadPoolOptionProto.Builder | getSessionInterOpThreadPoolBuilder (int インデックス) This option is experimental - it may be replaced with a different mechanism in the future. |
| リスト< ThreadPoolOptionProto.Builder > | getSessionInterOpThreadPoolBuilderList () This option is experimental - it may be replaced with a different mechanism in the future. |
| 整数 | getSessionInterOpThreadPoolCount () This option is experimental - it may be replaced with a different mechanism in the future. |
| リスト< ThreadPoolOptionProto > | getSessionInterOpThreadPoolList () This option is experimental - it may be replaced with a different mechanism in the future. |
| ThreadPoolOptionProtoOrBuilder | getSessionInterOpThreadPoolOrBuilder (int インデックス) This option is experimental - it may be replaced with a different mechanism in the future. |
| リスト<? ThreadPoolOptionProtoOrBuilderを拡張 > | getSessionInterOpThreadPoolOrBuilderList () This option is experimental - it may be replaced with a different mechanism in the future. |
| ブール値 | getShareClusterDevicesInSession () When true, WorkerSessions are created with device attributes from the full cluster. |
| ブール値 | getUsePerSessionThreads () If true, use a new set of threads for this session rather than the global pool of threads. |
| ブール値 | hasClusterDef () Optional list of all workers to use in this session. |
| ブール値 | 実験中() .tensorflow.ConfigProto.Experimental experimental = 16; |
| ブール値 | hasGpuOptions () Options that apply to all GPUs. |
| ブール値 | hasGraphOptions () Options that apply to all graphs. |
| ブール値 | hasRpcOptions () Options that apply when this session uses the distributed runtime. |
| 最終ブール値 | |
| ConfigProto.Builder | |
| ConfigProto.Builder | mergeExperimental ( ConfigProto.Experimental値) .tensorflow.ConfigProto.Experimental experimental = 16; |
| ConfigProto.Builder | mergeFrom (com.google.protobuf.Message other) |
| ConfigProto.Builder | mergeFrom (com.google.protobuf.CodedInputStream 入力、com.google.protobuf.ExtensionRegistryLite extensionRegistry) |
| ConfigProto.Builder | |
| ConfigProto.Builder | |
| ConfigProto.Builder | |
| 最終的なConfigProto.Builder | mergeUnknownFields (com.google.protobuf.UnknownFieldSet 不明フィールド) |
| ConfigProto.Builder | putAllDeviceCount (Map<String, Integer> 値) Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. |
| ConfigProto.Builder | putDeviceCount (文字列キー、int 値) Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. |
| ConfigProto.Builder | RemoveDeviceCount (文字列キー) Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. |
| ConfigProto.Builder | RemoveSessionInterOpThreadPool (int インデックス) This option is experimental - it may be replaced with a different mechanism in the future. |
| ConfigProto.Builder | setAllowSoftPlacement (ブール値) Whether soft placement is allowed. |
| ConfigProto.Builder | setClusterDef ( ClusterDef.Builder builderForValue) Optional list of all workers to use in this session. |
| ConfigProto.Builder | |
| ConfigProto.Builder | setDeviceFilters (int インデックス、文字列値) When any filters are present sessions will ignore all devices which do not match the filters. |
| ConfigProto.Builder | setExperimental ( ConfigProto.Experimental値) .tensorflow.ConfigProto.Experimental experimental = 16; |
| ConfigProto.Builder | setExperimental ( ConfigProto.Experimental.Builder builderForValue) .tensorflow.ConfigProto.Experimental experimental = 16; |
| ConfigProto.Builder | setField (com.google.protobuf.Descriptors.FieldDescriptor フィールド、オブジェクト値) |
| ConfigProto.Builder | |
| ConfigProto.Builder | |
| ConfigProto.Builder | |
| ConfigProto.Builder | |
| ConfigProto.Builder | setInterOpParallelismThreads (int 値) Nodes that perform blocking operations are enqueued on a pool of inter_op_parallelism_threads available in each process. |
| ConfigProto.Builder | setIntraOpParallelismThreads (int 値) The execution of an individual op (for some op types) can be parallelized on a pool of intra_op_parallelism_threads. |
| ConfigProto.Builder | setIsolateSessionState (ブール値) If true, any resources such as Variables used in the session will not be shared with other sessions. |
| ConfigProto.Builder | setLogDevicePlacement (ブール値) Whether device placements should be logged. |
| ConfigProto.Builder | setOperationTimeoutInMs (長い値) Global timeout for all blocking operations in this session. |
| ConfigProto.Builder | setPlacementPeriod (int 値) Assignment of Nodes to Devices is recomputed every placement_period steps until the system warms up (at which point the recomputation typically slows down automatically). |
| ConfigProto.Builder | setRepeatedField (com.google.protobuf.Descriptors.FieldDescriptor フィールド、int インデックス、オブジェクト値) |
| ConfigProto.Builder | |
| ConfigProto.Builder | setRpcOptions ( RPCOptions.Builder builderForValue) Options that apply when this session uses the distributed runtime. |
| ConfigProto.Builder | setSessionInterOpThreadPool (int インデックス、 ThreadPoolOptionProto値) This option is experimental - it may be replaced with a different mechanism in the future. |
| ConfigProto.Builder | setSessionInterOpThreadPool (int インデックス、 ThreadPoolOptionProto.Builder builderForValue) This option is experimental - it may be replaced with a different mechanism in the future. |
| ConfigProto.Builder | setShareClusterDevicesInSession (ブール値) When true, WorkerSessions are created with device attributes from the full cluster. |
| 最終的なConfigProto.Builder | setUnknownFields (com.google.protobuf.UnknownFieldSetknownFields) |
| ConfigProto.Builder | setUsePerSessionThreads (ブール値) If true, use a new set of threads for this session rather than the global pool of threads. |
継承されたメソッド
パブリックメソッド
public ConfigProto.Builder addAllDeviceFilters (Iterable<String> 値)
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public ConfigProto.Builder addAllSessionInterOpThreadPool (Iterable<? extends ThreadPoolOptionProto > 値)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder addDeviceFilters (文字列値)
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public ConfigProto.Builder addDeviceFiltersBytes (com.google.protobuf.ByteString 値)
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public ConfigProto.Builder addRepeatedField (com.google.protobuf.Descriptors.FieldDescriptor フィールド、オブジェクト値)
public ConfigProto.Builder addSessionInterOpThreadPool ( ThreadPoolOptionProto値)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder addSessionInterOpThreadPool (int インデックス、 ThreadPoolOptionProto値)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder addSessionInterOpThreadPool ( ThreadPoolOptionProto.Builder builderForValue)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder addSessionInterOpThreadPool (int インデックス、 ThreadPoolOptionProto.Builder builderForValue)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12;public ThreadPoolOptionProto.Builder addSessionInterOpThreadPoolBuilder (int インデックス)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ThreadPoolOptionProto.Builder addSessionInterOpThreadPoolBuilder ()
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder clearAllowSoftPlacement ()
Whether soft placement is allowed. If allow_soft_placement is true, an op will be placed on CPU if 1. there's no GPU implementation for the OP or 2. no GPU devices are known or registered or 3. need to co-locate with reftype input(s) which are from CPU.
bool allow_soft_placement = 7; public ConfigProto.Builder clearClusterDef ()
Optional list of all workers to use in this session.
.tensorflow.ClusterDef cluster_def = 14; public ConfigProto.Builder clearDeviceFilters ()
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public ConfigProto.Builder clearExperimental ()
.tensorflow.ConfigProto.Experimental experimental = 16;
public ConfigProto.Builder clearGpuOptions ()
Options that apply to all GPUs.
.tensorflow.GPUOptions gpu_options = 6; public ConfigProto.Builder clearGraphOptions ()
Options that apply to all graphs.
.tensorflow.GraphOptions graph_options = 10; public ConfigProto.Builder clearInterOpParallelismThreads ()
Nodes that perform blocking operations are enqueued on a pool of inter_op_parallelism_threads available in each process. 0 means the system picks an appropriate number. Negative means all operations are performed in caller's thread. Note that the first Session created in the process sets the number of threads for all future sessions unless use_per_session_threads is true or session_inter_op_thread_pool is configured.
int32 inter_op_parallelism_threads = 5; public ConfigProto.Builder clearIntraOpParallelismThreads ()
The execution of an individual op (for some op types) can be
parallelized on a pool of intra_op_parallelism_threads.
0 means the system picks an appropriate number.
If you create an ordinary session, e.g., from Python or C++,
then there is exactly one intra op thread pool per process.
The first session created determines the number of threads in this pool.
All subsequent sessions reuse/share this one global pool.
There are notable exceptions to the default behavior describe above:
1. There is an environment variable for overriding this thread pool,
named TF_OVERRIDE_GLOBAL_THREADPOOL.
2. When connecting to a server, such as a remote `tf.train.Server`
instance, then this option will be ignored altogether.
int32 intra_op_parallelism_threads = 2; public ConfigProto.Builder clearIsolateSessionState ()
If true, any resources such as Variables used in the session will not be shared with other sessions. However, when clusterspec propagation is enabled, this field is ignored and sessions are always isolated.
bool isolate_session_state = 15; public ConfigProto.Builder clearLogDevicePlacement ()
Whether device placements should be logged.
bool log_device_placement = 8; public ConfigProto.Builder clearOperationTimeoutInMs ()
Global timeout for all blocking operations in this session. If non-zero, and not overridden on a per-operation basis, this value will be used as the deadline for all blocking operations.
int64 operation_timeout_in_ms = 11; public ConfigProto.Builder clearPlacementPeriod ()
Assignment of Nodes to Devices is recomputed every placement_period steps until the system warms up (at which point the recomputation typically slows down automatically).
int32 placement_period = 3; public ConfigProto.Builder clearRpcOptions ()
Options that apply when this session uses the distributed runtime.
.tensorflow.RPCOptions rpc_options = 13; public ConfigProto.Builder clearSessionInterOpThreadPool ()
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder clearShareClusterDevicesInSession ()
When true, WorkerSessions are created with device attributes from the full cluster. This is helpful when a worker wants to partition a graph (for example during a PartitionedCallOp).
bool share_cluster_devices_in_session = 17; public ConfigProto.Builder clearUsePerSessionThreads ()
If true, use a new set of threads for this session rather than the global pool of threads. Only supported by direct sessions. If false, use the global threads created by the first session, or the per-session thread pools configured by session_inter_op_thread_pool. This option is deprecated. The same effect can be achieved by setting session_inter_op_thread_pool to have one element, whose num_threads equals inter_op_parallelism_threads.
bool use_per_session_threads = 9; public boolean containsDeviceCount (文字列キー)
Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. If a particular device type is not found in the map, the system picks an appropriate number.
map<string, int32> device_count = 1; public boolean getAllowSoftPlacement ()
Whether soft placement is allowed. If allow_soft_placement is true, an op will be placed on CPU if 1. there's no GPU implementation for the OP or 2. no GPU devices are known or registered or 3. need to co-locate with reftype input(s) which are from CPU.
bool allow_soft_placement = 7; public ClusterDef getClusterDef ()
Optional list of all workers to use in this session.
.tensorflow.ClusterDef cluster_def = 14; public ClusterDef.Builder getClusterDefBuilder ()
Optional list of all workers to use in this session.
.tensorflow.ClusterDef cluster_def = 14; public ClusterDefOrBuilder getClusterDefOrBuilder ()
Optional list of all workers to use in this session.
.tensorflow.ClusterDef cluster_def = 14; public static Final com.google.protobuf.Descriptors.Descriptor getDescriptor ()
public com.google.protobuf.Descriptors.Descriptor getDescriptorForType ()
public int getDeviceCountCount ()
Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. If a particular device type is not found in the map, the system picks an appropriate number.
map<string, int32> device_count = 1; public Map<String, Integer> getDeviceCountMap ()
Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. If a particular device type is not found in the map, the system picks an appropriate number.
map<string, int32> device_count = 1; public int getDeviceCountOrDefault (文字列キー、int defaultValue)
Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. If a particular device type is not found in the map, the system picks an appropriate number.
map<string, int32> device_count = 1;public int getDeviceCountOrThrow (文字列キー)
Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. If a particular device type is not found in the map, the system picks an appropriate number.
map<string, int32> device_count = 1; public String getDeviceFilters (int インデックス)
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public com.google.protobuf.ByteString getDeviceFiltersBytes (int インデックス)
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public int getDeviceFiltersCount ()
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public com.google.protobuf.ProtocolStringList getDeviceFiltersList ()
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public ConfigProto.Experimental getExperimental ()
.tensorflow.ConfigProto.Experimental experimental = 16;
public ConfigProto.Experimental.Builder getExperimentalBuilder ()
.tensorflow.ConfigProto.Experimental experimental = 16;
public ConfigProto.ExperimentalOrBuilder getExperimentalOrBuilder ()
.tensorflow.ConfigProto.Experimental experimental = 16;
パブリックGPUOptions getGpuOptions ()
Options that apply to all GPUs.
.tensorflow.GPUOptions gpu_options = 6; public GPUOptions.Builder getGpuOptionsBuilder ()
Options that apply to all GPUs.
.tensorflow.GPUOptions gpu_options = 6; public GPUOptionsOrBuilder getGpuOptionsOrBuilder ()
Options that apply to all GPUs.
.tensorflow.GPUOptions gpu_options = 6; public GraphOptions getGraphOptions ()
Options that apply to all graphs.
.tensorflow.GraphOptions graph_options = 10; public GraphOptions.Builder getGraphOptionsBuilder ()
Options that apply to all graphs.
.tensorflow.GraphOptions graph_options = 10; public GraphOptionsOrBuilder getGraphOptionsOrBuilder ()
Options that apply to all graphs.
.tensorflow.GraphOptions graph_options = 10; public int getInterOpParallelismThreads ()
Nodes that perform blocking operations are enqueued on a pool of inter_op_parallelism_threads available in each process. 0 means the system picks an appropriate number. Negative means all operations are performed in caller's thread. Note that the first Session created in the process sets the number of threads for all future sessions unless use_per_session_threads is true or session_inter_op_thread_pool is configured.
int32 inter_op_parallelism_threads = 5; public int getIntraOpParallelismThreads ()
The execution of an individual op (for some op types) can be
parallelized on a pool of intra_op_parallelism_threads.
0 means the system picks an appropriate number.
If you create an ordinary session, e.g., from Python or C++,
then there is exactly one intra op thread pool per process.
The first session created determines the number of threads in this pool.
All subsequent sessions reuse/share this one global pool.
There are notable exceptions to the default behavior describe above:
1. There is an environment variable for overriding this thread pool,
named TF_OVERRIDE_GLOBAL_THREADPOOL.
2. When connecting to a server, such as a remote `tf.train.Server`
instance, then this option will be ignored altogether.
int32 intra_op_parallelism_threads = 2; public boolean getIsolateSessionState ()
If true, any resources such as Variables used in the session will not be shared with other sessions. However, when clusterspec propagation is enabled, this field is ignored and sessions are always isolated.
bool isolate_session_state = 15; public boolean getLogDevicePlacement ()
Whether device placements should be logged.
bool log_device_placement = 8; public Map<String, Integer> getMutableDeviceCount ()
代わりに代替の突然変異アクセサーを使用してください。
public long getOperationTimeoutInMs ()
Global timeout for all blocking operations in this session. If non-zero, and not overridden on a per-operation basis, this value will be used as the deadline for all blocking operations.
int64 operation_timeout_in_ms = 11; public int getPlacementPeriod ()
Assignment of Nodes to Devices is recomputed every placement_period steps until the system warms up (at which point the recomputation typically slows down automatically).
int32 placement_period = 3; public RPCOptions getRpcOptions ()
Options that apply when this session uses the distributed runtime.
.tensorflow.RPCOptions rpc_options = 13; public RPCOptions.Builder getRpcOptionsBuilder ()
Options that apply when this session uses the distributed runtime.
.tensorflow.RPCOptions rpc_options = 13; public RPCOptionsOrBuilder getRpcOptionsOrBuilder ()
Options that apply when this session uses the distributed runtime.
.tensorflow.RPCOptions rpc_options = 13; public ThreadPoolOptionProto getSessionInterOpThreadPool (int インデックス)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12;public ThreadPoolOptionProto.Builder getSessionInterOpThreadPoolBuilder (int インデックス)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public List< ThreadPoolOptionProto.Builder > getSessionInterOpThreadPoolBuilderList ()
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public int getSessionInterOpThreadPoolCount ()
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public List< ThreadPoolOptionProto > getSessionInterOpThreadPoolList ()
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ThreadPoolOptionProtoOrBuilder getSessionInterOpThreadPoolOrBuilder (int インデックス)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; 公開リスト<? extends ThreadPoolOptionProtoOrBuilder > getSessionInterOpThreadPoolOrBuilderList ()
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public boolean getShareClusterDevicesInSession ()
When true, WorkerSessions are created with device attributes from the full cluster. This is helpful when a worker wants to partition a graph (for example during a PartitionedCallOp).
bool share_cluster_devices_in_session = 17; public boolean getUsePerSessionThreads ()
If true, use a new set of threads for this session rather than the global pool of threads. Only supported by direct sessions. If false, use the global threads created by the first session, or the per-session thread pools configured by session_inter_op_thread_pool. This option is deprecated. The same effect can be achieved by setting session_inter_op_thread_pool to have one element, whose num_threads equals inter_op_parallelism_threads.
bool use_per_session_threads = 9; public boolean hasClusterDef ()
Optional list of all workers to use in this session.
.tensorflow.ClusterDef cluster_def = 14; public boolean hasExperimental ()
.tensorflow.ConfigProto.Experimental experimental = 16;
public boolean hasGpuOptions ()
Options that apply to all GPUs.
.tensorflow.GPUOptions gpu_options = 6; public boolean hasGraphOptions ()
Options that apply to all graphs.
.tensorflow.GraphOptions graph_options = 10; public boolean hasRpcOptions ()
Options that apply when this session uses the distributed runtime.
.tensorflow.RPCOptions rpc_options = 13; パブリック最終ブール値isInitialized ()
public ConfigProto.Builder mergeClusterDef ( ClusterDef値)
Optional list of all workers to use in this session.
.tensorflow.ClusterDef cluster_def = 14; public ConfigProto.Builder mergeExperimental ( ConfigProto.Experimental値)
.tensorflow.ConfigProto.Experimental experimental = 16;
public ConfigProto.Builder mergeFrom (com.google.protobuf.CodedInputStream 入力、com.google.protobuf.ExtensionRegistryLite extensionRegistry)
投げる
| IO例外 |
|---|
public ConfigProto.Builder mergeGpuOptions ( GPUOptions値)
Options that apply to all GPUs.
.tensorflow.GPUOptions gpu_options = 6; public ConfigProto.Builder mergeGraphOptions ( GraphOptions値)
Options that apply to all graphs.
.tensorflow.GraphOptions graph_options = 10; public ConfigProto.Builder mergeRpcOptions ( RPCOptions値)
Options that apply when this session uses the distributed runtime.
.tensorflow.RPCOptions rpc_options = 13; public Final ConfigProto.Builder mergeUnknownFields (com.google.protobuf.UnknownFieldSet knownFields)
public ConfigProto.Builder putAllDeviceCount (Map<String, Integer> 値)
Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. If a particular device type is not found in the map, the system picks an appropriate number.
map<string, int32> device_count = 1; public ConfigProto.Builder putDeviceCount (文字列キー、int 値)
Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. If a particular device type is not found in the map, the system picks an appropriate number.
map<string, int32> device_count = 1; public ConfigProto.Builder RemoveDeviceCount (文字列キー)
Map from device type name (e.g., "CPU" or "GPU" ) to maximum number of devices of that type to use. If a particular device type is not found in the map, the system picks an appropriate number.
map<string, int32> device_count = 1; public ConfigProto.Builder RemoveSessionInterOpThreadPool (int インデックス)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder setAllowSoftPlacement (ブール値)
Whether soft placement is allowed. If allow_soft_placement is true, an op will be placed on CPU if 1. there's no GPU implementation for the OP or 2. no GPU devices are known or registered or 3. need to co-locate with reftype input(s) which are from CPU.
bool allow_soft_placement = 7; public ConfigProto.Builder setClusterDef ( ClusterDef.Builder builderForValue)
Optional list of all workers to use in this session.
.tensorflow.ClusterDef cluster_def = 14; public ConfigProto.Builder setClusterDef ( ClusterDef値)
Optional list of all workers to use in this session.
.tensorflow.ClusterDef cluster_def = 14;public ConfigProto.Builder setDeviceFilters (int インデックス、文字列値)
When any filters are present sessions will ignore all devices which do not match the filters. Each filter can be partially specified, e.g. "/job:ps" "/job:worker/replica:3", etc.
repeated string device_filters = 4; public ConfigProto.Builder setExperimental ( ConfigProto.Experimental値)
.tensorflow.ConfigProto.Experimental experimental = 16;
public ConfigProto.Builder setExperimental ( ConfigProto.Experimental.Builder builderForValue)
.tensorflow.ConfigProto.Experimental experimental = 16;
public ConfigProto.Builder setGpuOptions ( GPUOptions.Builder builderForValue)
Options that apply to all GPUs.
.tensorflow.GPUOptions gpu_options = 6; public ConfigProto.Builder setGpuOptions ( GPUOptions値)
Options that apply to all GPUs.
.tensorflow.GPUOptions gpu_options = 6; public ConfigProto.Builder setGraphOptions ( GraphOptions.Builder builderForValue)
Options that apply to all graphs.
.tensorflow.GraphOptions graph_options = 10; public ConfigProto.Builder setGraphOptions ( GraphOptions値)
Options that apply to all graphs.
.tensorflow.GraphOptions graph_options = 10; public ConfigProto.Builder setInterOpParallelismThreads (int 値)
Nodes that perform blocking operations are enqueued on a pool of inter_op_parallelism_threads available in each process. 0 means the system picks an appropriate number. Negative means all operations are performed in caller's thread. Note that the first Session created in the process sets the number of threads for all future sessions unless use_per_session_threads is true or session_inter_op_thread_pool is configured.
int32 inter_op_parallelism_threads = 5; public ConfigProto.Builder setIntraOpParallelismThreads (int 値)
The execution of an individual op (for some op types) can be
parallelized on a pool of intra_op_parallelism_threads.
0 means the system picks an appropriate number.
If you create an ordinary session, e.g., from Python or C++,
then there is exactly one intra op thread pool per process.
The first session created determines the number of threads in this pool.
All subsequent sessions reuse/share this one global pool.
There are notable exceptions to the default behavior describe above:
1. There is an environment variable for overriding this thread pool,
named TF_OVERRIDE_GLOBAL_THREADPOOL.
2. When connecting to a server, such as a remote `tf.train.Server`
instance, then this option will be ignored altogether.
int32 intra_op_parallelism_threads = 2; public ConfigProto.Builder setIsolateSessionState (ブール値)
If true, any resources such as Variables used in the session will not be shared with other sessions. However, when clusterspec propagation is enabled, this field is ignored and sessions are always isolated.
bool isolate_session_state = 15; public ConfigProto.Builder setLogDevicePlacement (ブール値)
Whether device placements should be logged.
bool log_device_placement = 8; public ConfigProto.Builder setOperationTimeoutInMs (長い値)
Global timeout for all blocking operations in this session. If non-zero, and not overridden on a per-operation basis, this value will be used as the deadline for all blocking operations.
int64 operation_timeout_in_ms = 11; public ConfigProto.Builder setPlacementPeriod (int 値)
Assignment of Nodes to Devices is recomputed every placement_period steps until the system warms up (at which point the recomputation typically slows down automatically).
int32 placement_period = 3; public ConfigProto.Builder setRepeatedField (com.google.protobuf.Descriptors.FieldDescriptor フィールド、int インデックス、オブジェクト値)
public ConfigProto.Builder setRpcOptions ( RPCOptions値)
Options that apply when this session uses the distributed runtime.
.tensorflow.RPCOptions rpc_options = 13; public ConfigProto.Builder setRpcOptions ( RPCOptions.Builder builderForValue)
Options that apply when this session uses the distributed runtime.
.tensorflow.RPCOptions rpc_options = 13; public ConfigProto.Builder setSessionInterOpThreadPool (int インデックス、 ThreadPoolOptionProto値)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder setSessionInterOpThreadPool (int インデックス、 ThreadPoolOptionProto.Builder builderForValue)
This option is experimental - it may be replaced with a different mechanism in the future. Configures session thread pools. If this is configured, then RunOptions for a Run call can select the thread pool to use. The intended use is for when some session invocations need to run in a background pool limited to a small number of threads: - For example, a session may be configured to have one large pool (for regular compute) and one small pool (for periodic, low priority work); using the small pool is currently the mechanism for limiting the inter-op parallelism of the low priority work. Note that it does not limit the parallelism of work spawned by a single op kernel implementation. - Using this setting is normally not needed in training, but may help some serving use cases. - It is also generally recommended to set the global_name field of this proto, to avoid creating multiple large pools. It is typically better to run the non-low-priority work, even across sessions, in a single large pool.
repeated .tensorflow.ThreadPoolOptionProto session_inter_op_thread_pool = 12; public ConfigProto.Builder setShareClusterDevicesInSession (ブール値)
When true, WorkerSessions are created with device attributes from the full cluster. This is helpful when a worker wants to partition a graph (for example during a PartitionedCallOp).
bool share_cluster_devices_in_session = 17; public ConfigProto.Builder setUsePerSessionThreads (ブール値)
If true, use a new set of threads for this session rather than the global pool of threads. Only supported by direct sessions. If false, use the global threads created by the first session, or the per-session thread pools configured by session_inter_op_thread_pool. This option is deprecated. The same effect can be achieved by setting session_inter_op_thread_pool to have one element, whose num_threads equals inter_op_parallelism_threads.
bool use_per_session_threads = 9;