org.tensorflow.framework.GPUOptions Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of proto Show documentation
Show all versions of proto Show documentation
Java API for TensorFlow protocol buffers.
// Generated by the protocol buffer compiler. DO NOT EDIT!
// source: tensorflow/core/protobuf/config.proto
package org.tensorflow.framework;
/**
* Protobuf type {@code tensorflow.GPUOptions}
*/
public final class GPUOptions extends
com.google.protobuf.GeneratedMessageV3 implements
// @@protoc_insertion_point(message_implements:tensorflow.GPUOptions)
GPUOptionsOrBuilder {
private static final long serialVersionUID = 0L;
// Use GPUOptions.newBuilder() to construct.
private GPUOptions(com.google.protobuf.GeneratedMessageV3.Builder builder) {
super(builder);
}
private GPUOptions() {
perProcessGpuMemoryFraction_ = 0D;
allowGrowth_ = false;
allocatorType_ = "";
deferredDeletionBytes_ = 0L;
visibleDeviceList_ = "";
pollingActiveDelayUsecs_ = 0;
pollingInactiveDelayMsecs_ = 0;
forceGpuCompatible_ = false;
}
@java.lang.Override
public final com.google.protobuf.UnknownFieldSet
getUnknownFields() {
return this.unknownFields;
}
private GPUOptions(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
this();
if (extensionRegistry == null) {
throw new java.lang.NullPointerException();
}
int mutable_bitField0_ = 0;
com.google.protobuf.UnknownFieldSet.Builder unknownFields =
com.google.protobuf.UnknownFieldSet.newBuilder();
try {
boolean done = false;
while (!done) {
int tag = input.readTag();
switch (tag) {
case 0:
done = true;
break;
default: {
if (!parseUnknownFieldProto3(
input, unknownFields, extensionRegistry, tag)) {
done = true;
}
break;
}
case 9: {
perProcessGpuMemoryFraction_ = input.readDouble();
break;
}
case 18: {
java.lang.String s = input.readStringRequireUtf8();
allocatorType_ = s;
break;
}
case 24: {
deferredDeletionBytes_ = input.readInt64();
break;
}
case 32: {
allowGrowth_ = input.readBool();
break;
}
case 42: {
java.lang.String s = input.readStringRequireUtf8();
visibleDeviceList_ = s;
break;
}
case 48: {
pollingActiveDelayUsecs_ = input.readInt32();
break;
}
case 56: {
pollingInactiveDelayMsecs_ = input.readInt32();
break;
}
case 64: {
forceGpuCompatible_ = input.readBool();
break;
}
case 74: {
org.tensorflow.framework.GPUOptions.Experimental.Builder subBuilder = null;
if (experimental_ != null) {
subBuilder = experimental_.toBuilder();
}
experimental_ = input.readMessage(org.tensorflow.framework.GPUOptions.Experimental.parser(), extensionRegistry);
if (subBuilder != null) {
subBuilder.mergeFrom(experimental_);
experimental_ = subBuilder.buildPartial();
}
break;
}
}
}
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
throw e.setUnfinishedMessage(this);
} catch (java.io.IOException e) {
throw new com.google.protobuf.InvalidProtocolBufferException(
e).setUnfinishedMessage(this);
} finally {
this.unknownFields = unknownFields.build();
makeExtensionsImmutable();
}
}
public static final com.google.protobuf.Descriptors.Descriptor
getDescriptor() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_descriptor;
}
protected com.google.protobuf.GeneratedMessageV3.FieldAccessorTable
internalGetFieldAccessorTable() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_fieldAccessorTable
.ensureFieldAccessorsInitialized(
org.tensorflow.framework.GPUOptions.class, org.tensorflow.framework.GPUOptions.Builder.class);
}
public interface ExperimentalOrBuilder extends
// @@protoc_insertion_point(interface_extends:tensorflow.GPUOptions.Experimental)
com.google.protobuf.MessageOrBuilder {
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
java.util.List
getVirtualDevicesList();
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices getVirtualDevices(int index);
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
int getVirtualDevicesCount();
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
java.util.List
getVirtualDevicesOrBuilderList();
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevicesOrBuilder getVirtualDevicesOrBuilder(
int index);
/**
*
* If true, uses CUDA unified memory for memory allocations. If
* per_process_gpu_memory_fraction option is greater than 1.0, then unified
* memory is used regardless of the value for this field. See comments for
* per_process_gpu_memory_fraction field for more details and requirements
* of the unified memory. This option is useful to oversubscribe memory if
* multiple processes are sharing a single GPU while individually using less
* than 1.0 per process memory fraction.
*
*
* bool use_unified_memory = 2;
*/
boolean getUseUnifiedMemory();
/**
*
* If > 1, the number of device-to-device copy streams to create
* for each GPUDevice. Default value is 0, which is automatically
* converted to 1.
*
*
* int32 num_dev_to_dev_copy_streams = 3;
*/
int getNumDevToDevCopyStreams();
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
java.lang.String getCollectiveRingOrder();
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
com.google.protobuf.ByteString
getCollectiveRingOrderBytes();
/**
*
* If true then extra work is done by GPUDevice and GPUBFCAllocator to
* keep track of when GPU memory is freed and when kernels actually
* complete so that we can know when a nominally free memory chunk
* is really not subject to pending use.
*
*
* bool timestamped_allocator = 5;
*/
boolean getTimestampedAllocator();
/**
*
* Parameters for GPUKernelTracker. By default no kernel tracking is done.
* Note that timestamped_allocator is only effective if some tracking is
* specified.
* If kernel_tracker_max_interval = n > 0, then a tracking event
* is inserted after every n kernels without an event.
*
*
* int32 kernel_tracker_max_interval = 7;
*/
int getKernelTrackerMaxInterval();
/**
*
* If kernel_tracker_max_bytes = n > 0, then a tracking event is
* inserted after every series of kernels allocating a sum of
* memory >= n. If one kernel allocates b * n bytes, then one
* event will be inserted after it, but it will count as b against
* the pending limit.
*
*
* int32 kernel_tracker_max_bytes = 8;
*/
int getKernelTrackerMaxBytes();
/**
*
* If kernel_tracker_max_pending > 0 then no more than this many
* tracking events can be outstanding at a time. An attempt to
* launch an additional kernel will stall until an event
* completes.
*
*
* int32 kernel_tracker_max_pending = 9;
*/
int getKernelTrackerMaxPending();
}
/**
* Protobuf type {@code tensorflow.GPUOptions.Experimental}
*/
public static final class Experimental extends
com.google.protobuf.GeneratedMessageV3 implements
// @@protoc_insertion_point(message_implements:tensorflow.GPUOptions.Experimental)
ExperimentalOrBuilder {
private static final long serialVersionUID = 0L;
// Use Experimental.newBuilder() to construct.
private Experimental(com.google.protobuf.GeneratedMessageV3.Builder builder) {
super(builder);
}
private Experimental() {
virtualDevices_ = java.util.Collections.emptyList();
useUnifiedMemory_ = false;
numDevToDevCopyStreams_ = 0;
collectiveRingOrder_ = "";
timestampedAllocator_ = false;
kernelTrackerMaxInterval_ = 0;
kernelTrackerMaxBytes_ = 0;
kernelTrackerMaxPending_ = 0;
}
@java.lang.Override
public final com.google.protobuf.UnknownFieldSet
getUnknownFields() {
return this.unknownFields;
}
private Experimental(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
this();
if (extensionRegistry == null) {
throw new java.lang.NullPointerException();
}
int mutable_bitField0_ = 0;
com.google.protobuf.UnknownFieldSet.Builder unknownFields =
com.google.protobuf.UnknownFieldSet.newBuilder();
try {
boolean done = false;
while (!done) {
int tag = input.readTag();
switch (tag) {
case 0:
done = true;
break;
default: {
if (!parseUnknownFieldProto3(
input, unknownFields, extensionRegistry, tag)) {
done = true;
}
break;
}
case 10: {
if (!((mutable_bitField0_ & 0x00000001) == 0x00000001)) {
virtualDevices_ = new java.util.ArrayList();
mutable_bitField0_ |= 0x00000001;
}
virtualDevices_.add(
input.readMessage(org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.parser(), extensionRegistry));
break;
}
case 16: {
useUnifiedMemory_ = input.readBool();
break;
}
case 24: {
numDevToDevCopyStreams_ = input.readInt32();
break;
}
case 34: {
java.lang.String s = input.readStringRequireUtf8();
collectiveRingOrder_ = s;
break;
}
case 40: {
timestampedAllocator_ = input.readBool();
break;
}
case 56: {
kernelTrackerMaxInterval_ = input.readInt32();
break;
}
case 64: {
kernelTrackerMaxBytes_ = input.readInt32();
break;
}
case 72: {
kernelTrackerMaxPending_ = input.readInt32();
break;
}
}
}
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
throw e.setUnfinishedMessage(this);
} catch (java.io.IOException e) {
throw new com.google.protobuf.InvalidProtocolBufferException(
e).setUnfinishedMessage(this);
} finally {
if (((mutable_bitField0_ & 0x00000001) == 0x00000001)) {
virtualDevices_ = java.util.Collections.unmodifiableList(virtualDevices_);
}
this.unknownFields = unknownFields.build();
makeExtensionsImmutable();
}
}
public static final com.google.protobuf.Descriptors.Descriptor
getDescriptor() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_descriptor;
}
protected com.google.protobuf.GeneratedMessageV3.FieldAccessorTable
internalGetFieldAccessorTable() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_fieldAccessorTable
.ensureFieldAccessorsInitialized(
org.tensorflow.framework.GPUOptions.Experimental.class, org.tensorflow.framework.GPUOptions.Experimental.Builder.class);
}
public interface VirtualDevicesOrBuilder extends
// @@protoc_insertion_point(interface_extends:tensorflow.GPUOptions.Experimental.VirtualDevices)
com.google.protobuf.MessageOrBuilder {
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
java.util.List getMemoryLimitMbList();
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
int getMemoryLimitMbCount();
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
float getMemoryLimitMb(int index);
}
/**
*
* Configuration for breaking down a visible GPU into multiple "virtual"
* devices.
*
*
* Protobuf type {@code tensorflow.GPUOptions.Experimental.VirtualDevices}
*/
public static final class VirtualDevices extends
com.google.protobuf.GeneratedMessageV3 implements
// @@protoc_insertion_point(message_implements:tensorflow.GPUOptions.Experimental.VirtualDevices)
VirtualDevicesOrBuilder {
private static final long serialVersionUID = 0L;
// Use VirtualDevices.newBuilder() to construct.
private VirtualDevices(com.google.protobuf.GeneratedMessageV3.Builder builder) {
super(builder);
}
private VirtualDevices() {
memoryLimitMb_ = java.util.Collections.emptyList();
}
@java.lang.Override
public final com.google.protobuf.UnknownFieldSet
getUnknownFields() {
return this.unknownFields;
}
private VirtualDevices(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
this();
if (extensionRegistry == null) {
throw new java.lang.NullPointerException();
}
int mutable_bitField0_ = 0;
com.google.protobuf.UnknownFieldSet.Builder unknownFields =
com.google.protobuf.UnknownFieldSet.newBuilder();
try {
boolean done = false;
while (!done) {
int tag = input.readTag();
switch (tag) {
case 0:
done = true;
break;
default: {
if (!parseUnknownFieldProto3(
input, unknownFields, extensionRegistry, tag)) {
done = true;
}
break;
}
case 13: {
if (!((mutable_bitField0_ & 0x00000001) == 0x00000001)) {
memoryLimitMb_ = new java.util.ArrayList();
mutable_bitField0_ |= 0x00000001;
}
memoryLimitMb_.add(input.readFloat());
break;
}
case 10: {
int length = input.readRawVarint32();
int limit = input.pushLimit(length);
if (!((mutable_bitField0_ & 0x00000001) == 0x00000001) && input.getBytesUntilLimit() > 0) {
memoryLimitMb_ = new java.util.ArrayList();
mutable_bitField0_ |= 0x00000001;
}
while (input.getBytesUntilLimit() > 0) {
memoryLimitMb_.add(input.readFloat());
}
input.popLimit(limit);
break;
}
}
}
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
throw e.setUnfinishedMessage(this);
} catch (java.io.IOException e) {
throw new com.google.protobuf.InvalidProtocolBufferException(
e).setUnfinishedMessage(this);
} finally {
if (((mutable_bitField0_ & 0x00000001) == 0x00000001)) {
memoryLimitMb_ = java.util.Collections.unmodifiableList(memoryLimitMb_);
}
this.unknownFields = unknownFields.build();
makeExtensionsImmutable();
}
}
public static final com.google.protobuf.Descriptors.Descriptor
getDescriptor() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_VirtualDevices_descriptor;
}
protected com.google.protobuf.GeneratedMessageV3.FieldAccessorTable
internalGetFieldAccessorTable() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_VirtualDevices_fieldAccessorTable
.ensureFieldAccessorsInitialized(
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.class, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder.class);
}
public static final int MEMORY_LIMIT_MB_FIELD_NUMBER = 1;
private java.util.List memoryLimitMb_;
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public java.util.List
getMemoryLimitMbList() {
return memoryLimitMb_;
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public int getMemoryLimitMbCount() {
return memoryLimitMb_.size();
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public float getMemoryLimitMb(int index) {
return memoryLimitMb_.get(index);
}
private int memoryLimitMbMemoizedSerializedSize = -1;
private byte memoizedIsInitialized = -1;
public final boolean isInitialized() {
byte isInitialized = memoizedIsInitialized;
if (isInitialized == 1) return true;
if (isInitialized == 0) return false;
memoizedIsInitialized = 1;
return true;
}
public void writeTo(com.google.protobuf.CodedOutputStream output)
throws java.io.IOException {
getSerializedSize();
if (getMemoryLimitMbList().size() > 0) {
output.writeUInt32NoTag(10);
output.writeUInt32NoTag(memoryLimitMbMemoizedSerializedSize);
}
for (int i = 0; i < memoryLimitMb_.size(); i++) {
output.writeFloatNoTag(memoryLimitMb_.get(i));
}
unknownFields.writeTo(output);
}
public int getSerializedSize() {
int size = memoizedSize;
if (size != -1) return size;
size = 0;
{
int dataSize = 0;
dataSize = 4 * getMemoryLimitMbList().size();
size += dataSize;
if (!getMemoryLimitMbList().isEmpty()) {
size += 1;
size += com.google.protobuf.CodedOutputStream
.computeInt32SizeNoTag(dataSize);
}
memoryLimitMbMemoizedSerializedSize = dataSize;
}
size += unknownFields.getSerializedSize();
memoizedSize = size;
return size;
}
@java.lang.Override
public boolean equals(final java.lang.Object obj) {
if (obj == this) {
return true;
}
if (!(obj instanceof org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices)) {
return super.equals(obj);
}
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices other = (org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices) obj;
boolean result = true;
result = result && getMemoryLimitMbList()
.equals(other.getMemoryLimitMbList());
result = result && unknownFields.equals(other.unknownFields);
return result;
}
@java.lang.Override
public int hashCode() {
if (memoizedHashCode != 0) {
return memoizedHashCode;
}
int hash = 41;
hash = (19 * hash) + getDescriptor().hashCode();
if (getMemoryLimitMbCount() > 0) {
hash = (37 * hash) + MEMORY_LIMIT_MB_FIELD_NUMBER;
hash = (53 * hash) + getMemoryLimitMbList().hashCode();
}
hash = (29 * hash) + unknownFields.hashCode();
memoizedHashCode = hash;
return hash;
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(
java.nio.ByteBuffer data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(
java.nio.ByteBuffer data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(
com.google.protobuf.ByteString data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(
com.google.protobuf.ByteString data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(byte[] data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(
byte[] data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(java.io.InputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(
java.io.InputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseDelimitedFrom(java.io.InputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseDelimitedWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseDelimitedFrom(
java.io.InputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseDelimitedWithIOException(PARSER, input, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(
com.google.protobuf.CodedInputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parseFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input, extensionRegistry);
}
public Builder newBuilderForType() { return newBuilder(); }
public static Builder newBuilder() {
return DEFAULT_INSTANCE.toBuilder();
}
public static Builder newBuilder(org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices prototype) {
return DEFAULT_INSTANCE.toBuilder().mergeFrom(prototype);
}
public Builder toBuilder() {
return this == DEFAULT_INSTANCE
? new Builder() : new Builder().mergeFrom(this);
}
@java.lang.Override
protected Builder newBuilderForType(
com.google.protobuf.GeneratedMessageV3.BuilderParent parent) {
Builder builder = new Builder(parent);
return builder;
}
/**
*
* Configuration for breaking down a visible GPU into multiple "virtual"
* devices.
*
*
* Protobuf type {@code tensorflow.GPUOptions.Experimental.VirtualDevices}
*/
public static final class Builder extends
com.google.protobuf.GeneratedMessageV3.Builder implements
// @@protoc_insertion_point(builder_implements:tensorflow.GPUOptions.Experimental.VirtualDevices)
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevicesOrBuilder {
public static final com.google.protobuf.Descriptors.Descriptor
getDescriptor() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_VirtualDevices_descriptor;
}
protected com.google.protobuf.GeneratedMessageV3.FieldAccessorTable
internalGetFieldAccessorTable() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_VirtualDevices_fieldAccessorTable
.ensureFieldAccessorsInitialized(
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.class, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder.class);
}
// Construct using org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.newBuilder()
private Builder() {
maybeForceBuilderInitialization();
}
private Builder(
com.google.protobuf.GeneratedMessageV3.BuilderParent parent) {
super(parent);
maybeForceBuilderInitialization();
}
private void maybeForceBuilderInitialization() {
if (com.google.protobuf.GeneratedMessageV3
.alwaysUseFieldBuilders) {
}
}
public Builder clear() {
super.clear();
memoryLimitMb_ = java.util.Collections.emptyList();
bitField0_ = (bitField0_ & ~0x00000001);
return this;
}
public com.google.protobuf.Descriptors.Descriptor
getDescriptorForType() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_VirtualDevices_descriptor;
}
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices getDefaultInstanceForType() {
return org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.getDefaultInstance();
}
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices build() {
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices result = buildPartial();
if (!result.isInitialized()) {
throw newUninitializedMessageException(result);
}
return result;
}
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices buildPartial() {
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices result = new org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices(this);
int from_bitField0_ = bitField0_;
if (((bitField0_ & 0x00000001) == 0x00000001)) {
memoryLimitMb_ = java.util.Collections.unmodifiableList(memoryLimitMb_);
bitField0_ = (bitField0_ & ~0x00000001);
}
result.memoryLimitMb_ = memoryLimitMb_;
onBuilt();
return result;
}
public Builder clone() {
return (Builder) super.clone();
}
public Builder setField(
com.google.protobuf.Descriptors.FieldDescriptor field,
java.lang.Object value) {
return (Builder) super.setField(field, value);
}
public Builder clearField(
com.google.protobuf.Descriptors.FieldDescriptor field) {
return (Builder) super.clearField(field);
}
public Builder clearOneof(
com.google.protobuf.Descriptors.OneofDescriptor oneof) {
return (Builder) super.clearOneof(oneof);
}
public Builder setRepeatedField(
com.google.protobuf.Descriptors.FieldDescriptor field,
int index, java.lang.Object value) {
return (Builder) super.setRepeatedField(field, index, value);
}
public Builder addRepeatedField(
com.google.protobuf.Descriptors.FieldDescriptor field,
java.lang.Object value) {
return (Builder) super.addRepeatedField(field, value);
}
public Builder mergeFrom(com.google.protobuf.Message other) {
if (other instanceof org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices) {
return mergeFrom((org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices)other);
} else {
super.mergeFrom(other);
return this;
}
}
public Builder mergeFrom(org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices other) {
if (other == org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.getDefaultInstance()) return this;
if (!other.memoryLimitMb_.isEmpty()) {
if (memoryLimitMb_.isEmpty()) {
memoryLimitMb_ = other.memoryLimitMb_;
bitField0_ = (bitField0_ & ~0x00000001);
} else {
ensureMemoryLimitMbIsMutable();
memoryLimitMb_.addAll(other.memoryLimitMb_);
}
onChanged();
}
this.mergeUnknownFields(other.unknownFields);
onChanged();
return this;
}
public final boolean isInitialized() {
return true;
}
public Builder mergeFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices parsedMessage = null;
try {
parsedMessage = PARSER.parsePartialFrom(input, extensionRegistry);
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
parsedMessage = (org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices) e.getUnfinishedMessage();
throw e.unwrapIOException();
} finally {
if (parsedMessage != null) {
mergeFrom(parsedMessage);
}
}
return this;
}
private int bitField0_;
private java.util.List memoryLimitMb_ = java.util.Collections.emptyList();
private void ensureMemoryLimitMbIsMutable() {
if (!((bitField0_ & 0x00000001) == 0x00000001)) {
memoryLimitMb_ = new java.util.ArrayList(memoryLimitMb_);
bitField0_ |= 0x00000001;
}
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public java.util.List
getMemoryLimitMbList() {
return java.util.Collections.unmodifiableList(memoryLimitMb_);
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public int getMemoryLimitMbCount() {
return memoryLimitMb_.size();
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public float getMemoryLimitMb(int index) {
return memoryLimitMb_.get(index);
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public Builder setMemoryLimitMb(
int index, float value) {
ensureMemoryLimitMbIsMutable();
memoryLimitMb_.set(index, value);
onChanged();
return this;
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public Builder addMemoryLimitMb(float value) {
ensureMemoryLimitMbIsMutable();
memoryLimitMb_.add(value);
onChanged();
return this;
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public Builder addAllMemoryLimitMb(
java.lang.Iterable values) {
ensureMemoryLimitMbIsMutable();
com.google.protobuf.AbstractMessageLite.Builder.addAll(
values, memoryLimitMb_);
onChanged();
return this;
}
/**
*
* Per "virtual" device memory limit, in MB. The number of elements in
* the list is the number of virtual devices to create on the
* corresponding visible GPU (see "virtual_devices" below).
* If empty, it will create single virtual device taking all available
* memory from the device.
* For the concept of "visible" and "virtual" GPU, see the comments for
* "visible_device_list" above for more information.
*
*
* repeated float memory_limit_mb = 1;
*/
public Builder clearMemoryLimitMb() {
memoryLimitMb_ = java.util.Collections.emptyList();
bitField0_ = (bitField0_ & ~0x00000001);
onChanged();
return this;
}
public final Builder setUnknownFields(
final com.google.protobuf.UnknownFieldSet unknownFields) {
return super.setUnknownFieldsProto3(unknownFields);
}
public final Builder mergeUnknownFields(
final com.google.protobuf.UnknownFieldSet unknownFields) {
return super.mergeUnknownFields(unknownFields);
}
// @@protoc_insertion_point(builder_scope:tensorflow.GPUOptions.Experimental.VirtualDevices)
}
// @@protoc_insertion_point(class_scope:tensorflow.GPUOptions.Experimental.VirtualDevices)
private static final org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices DEFAULT_INSTANCE;
static {
DEFAULT_INSTANCE = new org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices();
}
public static org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices getDefaultInstance() {
return DEFAULT_INSTANCE;
}
private static final com.google.protobuf.Parser
PARSER = new com.google.protobuf.AbstractParser() {
public VirtualDevices parsePartialFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return new VirtualDevices(input, extensionRegistry);
}
};
public static com.google.protobuf.Parser parser() {
return PARSER;
}
@java.lang.Override
public com.google.protobuf.Parser getParserForType() {
return PARSER;
}
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices getDefaultInstanceForType() {
return DEFAULT_INSTANCE;
}
}
private int bitField0_;
public static final int VIRTUAL_DEVICES_FIELD_NUMBER = 1;
private java.util.List virtualDevices_;
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public java.util.List getVirtualDevicesList() {
return virtualDevices_;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public java.util.List
getVirtualDevicesOrBuilderList() {
return virtualDevices_;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public int getVirtualDevicesCount() {
return virtualDevices_.size();
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices getVirtualDevices(int index) {
return virtualDevices_.get(index);
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevicesOrBuilder getVirtualDevicesOrBuilder(
int index) {
return virtualDevices_.get(index);
}
public static final int USE_UNIFIED_MEMORY_FIELD_NUMBER = 2;
private boolean useUnifiedMemory_;
/**
*
* If true, uses CUDA unified memory for memory allocations. If
* per_process_gpu_memory_fraction option is greater than 1.0, then unified
* memory is used regardless of the value for this field. See comments for
* per_process_gpu_memory_fraction field for more details and requirements
* of the unified memory. This option is useful to oversubscribe memory if
* multiple processes are sharing a single GPU while individually using less
* than 1.0 per process memory fraction.
*
*
* bool use_unified_memory = 2;
*/
public boolean getUseUnifiedMemory() {
return useUnifiedMemory_;
}
public static final int NUM_DEV_TO_DEV_COPY_STREAMS_FIELD_NUMBER = 3;
private int numDevToDevCopyStreams_;
/**
*
* If > 1, the number of device-to-device copy streams to create
* for each GPUDevice. Default value is 0, which is automatically
* converted to 1.
*
*
* int32 num_dev_to_dev_copy_streams = 3;
*/
public int getNumDevToDevCopyStreams() {
return numDevToDevCopyStreams_;
}
public static final int COLLECTIVE_RING_ORDER_FIELD_NUMBER = 4;
private volatile java.lang.Object collectiveRingOrder_;
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
public java.lang.String getCollectiveRingOrder() {
java.lang.Object ref = collectiveRingOrder_;
if (ref instanceof java.lang.String) {
return (java.lang.String) ref;
} else {
com.google.protobuf.ByteString bs =
(com.google.protobuf.ByteString) ref;
java.lang.String s = bs.toStringUtf8();
collectiveRingOrder_ = s;
return s;
}
}
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
public com.google.protobuf.ByteString
getCollectiveRingOrderBytes() {
java.lang.Object ref = collectiveRingOrder_;
if (ref instanceof java.lang.String) {
com.google.protobuf.ByteString b =
com.google.protobuf.ByteString.copyFromUtf8(
(java.lang.String) ref);
collectiveRingOrder_ = b;
return b;
} else {
return (com.google.protobuf.ByteString) ref;
}
}
public static final int TIMESTAMPED_ALLOCATOR_FIELD_NUMBER = 5;
private boolean timestampedAllocator_;
/**
*
* If true then extra work is done by GPUDevice and GPUBFCAllocator to
* keep track of when GPU memory is freed and when kernels actually
* complete so that we can know when a nominally free memory chunk
* is really not subject to pending use.
*
*
* bool timestamped_allocator = 5;
*/
public boolean getTimestampedAllocator() {
return timestampedAllocator_;
}
public static final int KERNEL_TRACKER_MAX_INTERVAL_FIELD_NUMBER = 7;
private int kernelTrackerMaxInterval_;
/**
*
* Parameters for GPUKernelTracker. By default no kernel tracking is done.
* Note that timestamped_allocator is only effective if some tracking is
* specified.
* If kernel_tracker_max_interval = n > 0, then a tracking event
* is inserted after every n kernels without an event.
*
*
* int32 kernel_tracker_max_interval = 7;
*/
public int getKernelTrackerMaxInterval() {
return kernelTrackerMaxInterval_;
}
public static final int KERNEL_TRACKER_MAX_BYTES_FIELD_NUMBER = 8;
private int kernelTrackerMaxBytes_;
/**
*
* If kernel_tracker_max_bytes = n > 0, then a tracking event is
* inserted after every series of kernels allocating a sum of
* memory >= n. If one kernel allocates b * n bytes, then one
* event will be inserted after it, but it will count as b against
* the pending limit.
*
*
* int32 kernel_tracker_max_bytes = 8;
*/
public int getKernelTrackerMaxBytes() {
return kernelTrackerMaxBytes_;
}
public static final int KERNEL_TRACKER_MAX_PENDING_FIELD_NUMBER = 9;
private int kernelTrackerMaxPending_;
/**
*
* If kernel_tracker_max_pending > 0 then no more than this many
* tracking events can be outstanding at a time. An attempt to
* launch an additional kernel will stall until an event
* completes.
*
*
* int32 kernel_tracker_max_pending = 9;
*/
public int getKernelTrackerMaxPending() {
return kernelTrackerMaxPending_;
}
private byte memoizedIsInitialized = -1;
public final boolean isInitialized() {
byte isInitialized = memoizedIsInitialized;
if (isInitialized == 1) return true;
if (isInitialized == 0) return false;
memoizedIsInitialized = 1;
return true;
}
public void writeTo(com.google.protobuf.CodedOutputStream output)
throws java.io.IOException {
for (int i = 0; i < virtualDevices_.size(); i++) {
output.writeMessage(1, virtualDevices_.get(i));
}
if (useUnifiedMemory_ != false) {
output.writeBool(2, useUnifiedMemory_);
}
if (numDevToDevCopyStreams_ != 0) {
output.writeInt32(3, numDevToDevCopyStreams_);
}
if (!getCollectiveRingOrderBytes().isEmpty()) {
com.google.protobuf.GeneratedMessageV3.writeString(output, 4, collectiveRingOrder_);
}
if (timestampedAllocator_ != false) {
output.writeBool(5, timestampedAllocator_);
}
if (kernelTrackerMaxInterval_ != 0) {
output.writeInt32(7, kernelTrackerMaxInterval_);
}
if (kernelTrackerMaxBytes_ != 0) {
output.writeInt32(8, kernelTrackerMaxBytes_);
}
if (kernelTrackerMaxPending_ != 0) {
output.writeInt32(9, kernelTrackerMaxPending_);
}
unknownFields.writeTo(output);
}
public int getSerializedSize() {
int size = memoizedSize;
if (size != -1) return size;
size = 0;
for (int i = 0; i < virtualDevices_.size(); i++) {
size += com.google.protobuf.CodedOutputStream
.computeMessageSize(1, virtualDevices_.get(i));
}
if (useUnifiedMemory_ != false) {
size += com.google.protobuf.CodedOutputStream
.computeBoolSize(2, useUnifiedMemory_);
}
if (numDevToDevCopyStreams_ != 0) {
size += com.google.protobuf.CodedOutputStream
.computeInt32Size(3, numDevToDevCopyStreams_);
}
if (!getCollectiveRingOrderBytes().isEmpty()) {
size += com.google.protobuf.GeneratedMessageV3.computeStringSize(4, collectiveRingOrder_);
}
if (timestampedAllocator_ != false) {
size += com.google.protobuf.CodedOutputStream
.computeBoolSize(5, timestampedAllocator_);
}
if (kernelTrackerMaxInterval_ != 0) {
size += com.google.protobuf.CodedOutputStream
.computeInt32Size(7, kernelTrackerMaxInterval_);
}
if (kernelTrackerMaxBytes_ != 0) {
size += com.google.protobuf.CodedOutputStream
.computeInt32Size(8, kernelTrackerMaxBytes_);
}
if (kernelTrackerMaxPending_ != 0) {
size += com.google.protobuf.CodedOutputStream
.computeInt32Size(9, kernelTrackerMaxPending_);
}
size += unknownFields.getSerializedSize();
memoizedSize = size;
return size;
}
@java.lang.Override
public boolean equals(final java.lang.Object obj) {
if (obj == this) {
return true;
}
if (!(obj instanceof org.tensorflow.framework.GPUOptions.Experimental)) {
return super.equals(obj);
}
org.tensorflow.framework.GPUOptions.Experimental other = (org.tensorflow.framework.GPUOptions.Experimental) obj;
boolean result = true;
result = result && getVirtualDevicesList()
.equals(other.getVirtualDevicesList());
result = result && (getUseUnifiedMemory()
== other.getUseUnifiedMemory());
result = result && (getNumDevToDevCopyStreams()
== other.getNumDevToDevCopyStreams());
result = result && getCollectiveRingOrder()
.equals(other.getCollectiveRingOrder());
result = result && (getTimestampedAllocator()
== other.getTimestampedAllocator());
result = result && (getKernelTrackerMaxInterval()
== other.getKernelTrackerMaxInterval());
result = result && (getKernelTrackerMaxBytes()
== other.getKernelTrackerMaxBytes());
result = result && (getKernelTrackerMaxPending()
== other.getKernelTrackerMaxPending());
result = result && unknownFields.equals(other.unknownFields);
return result;
}
@java.lang.Override
public int hashCode() {
if (memoizedHashCode != 0) {
return memoizedHashCode;
}
int hash = 41;
hash = (19 * hash) + getDescriptor().hashCode();
if (getVirtualDevicesCount() > 0) {
hash = (37 * hash) + VIRTUAL_DEVICES_FIELD_NUMBER;
hash = (53 * hash) + getVirtualDevicesList().hashCode();
}
hash = (37 * hash) + USE_UNIFIED_MEMORY_FIELD_NUMBER;
hash = (53 * hash) + com.google.protobuf.Internal.hashBoolean(
getUseUnifiedMemory());
hash = (37 * hash) + NUM_DEV_TO_DEV_COPY_STREAMS_FIELD_NUMBER;
hash = (53 * hash) + getNumDevToDevCopyStreams();
hash = (37 * hash) + COLLECTIVE_RING_ORDER_FIELD_NUMBER;
hash = (53 * hash) + getCollectiveRingOrder().hashCode();
hash = (37 * hash) + TIMESTAMPED_ALLOCATOR_FIELD_NUMBER;
hash = (53 * hash) + com.google.protobuf.Internal.hashBoolean(
getTimestampedAllocator());
hash = (37 * hash) + KERNEL_TRACKER_MAX_INTERVAL_FIELD_NUMBER;
hash = (53 * hash) + getKernelTrackerMaxInterval();
hash = (37 * hash) + KERNEL_TRACKER_MAX_BYTES_FIELD_NUMBER;
hash = (53 * hash) + getKernelTrackerMaxBytes();
hash = (37 * hash) + KERNEL_TRACKER_MAX_PENDING_FIELD_NUMBER;
hash = (53 * hash) + getKernelTrackerMaxPending();
hash = (29 * hash) + unknownFields.hashCode();
memoizedHashCode = hash;
return hash;
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(
java.nio.ByteBuffer data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(
java.nio.ByteBuffer data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(
com.google.protobuf.ByteString data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(
com.google.protobuf.ByteString data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(byte[] data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(
byte[] data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(java.io.InputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(
java.io.InputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseDelimitedFrom(java.io.InputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseDelimitedWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseDelimitedFrom(
java.io.InputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseDelimitedWithIOException(PARSER, input, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(
com.google.protobuf.CodedInputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions.Experimental parseFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input, extensionRegistry);
}
public Builder newBuilderForType() { return newBuilder(); }
public static Builder newBuilder() {
return DEFAULT_INSTANCE.toBuilder();
}
public static Builder newBuilder(org.tensorflow.framework.GPUOptions.Experimental prototype) {
return DEFAULT_INSTANCE.toBuilder().mergeFrom(prototype);
}
public Builder toBuilder() {
return this == DEFAULT_INSTANCE
? new Builder() : new Builder().mergeFrom(this);
}
@java.lang.Override
protected Builder newBuilderForType(
com.google.protobuf.GeneratedMessageV3.BuilderParent parent) {
Builder builder = new Builder(parent);
return builder;
}
/**
* Protobuf type {@code tensorflow.GPUOptions.Experimental}
*/
public static final class Builder extends
com.google.protobuf.GeneratedMessageV3.Builder implements
// @@protoc_insertion_point(builder_implements:tensorflow.GPUOptions.Experimental)
org.tensorflow.framework.GPUOptions.ExperimentalOrBuilder {
public static final com.google.protobuf.Descriptors.Descriptor
getDescriptor() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_descriptor;
}
protected com.google.protobuf.GeneratedMessageV3.FieldAccessorTable
internalGetFieldAccessorTable() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_fieldAccessorTable
.ensureFieldAccessorsInitialized(
org.tensorflow.framework.GPUOptions.Experimental.class, org.tensorflow.framework.GPUOptions.Experimental.Builder.class);
}
// Construct using org.tensorflow.framework.GPUOptions.Experimental.newBuilder()
private Builder() {
maybeForceBuilderInitialization();
}
private Builder(
com.google.protobuf.GeneratedMessageV3.BuilderParent parent) {
super(parent);
maybeForceBuilderInitialization();
}
private void maybeForceBuilderInitialization() {
if (com.google.protobuf.GeneratedMessageV3
.alwaysUseFieldBuilders) {
getVirtualDevicesFieldBuilder();
}
}
public Builder clear() {
super.clear();
if (virtualDevicesBuilder_ == null) {
virtualDevices_ = java.util.Collections.emptyList();
bitField0_ = (bitField0_ & ~0x00000001);
} else {
virtualDevicesBuilder_.clear();
}
useUnifiedMemory_ = false;
numDevToDevCopyStreams_ = 0;
collectiveRingOrder_ = "";
timestampedAllocator_ = false;
kernelTrackerMaxInterval_ = 0;
kernelTrackerMaxBytes_ = 0;
kernelTrackerMaxPending_ = 0;
return this;
}
public com.google.protobuf.Descriptors.Descriptor
getDescriptorForType() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_Experimental_descriptor;
}
public org.tensorflow.framework.GPUOptions.Experimental getDefaultInstanceForType() {
return org.tensorflow.framework.GPUOptions.Experimental.getDefaultInstance();
}
public org.tensorflow.framework.GPUOptions.Experimental build() {
org.tensorflow.framework.GPUOptions.Experimental result = buildPartial();
if (!result.isInitialized()) {
throw newUninitializedMessageException(result);
}
return result;
}
public org.tensorflow.framework.GPUOptions.Experimental buildPartial() {
org.tensorflow.framework.GPUOptions.Experimental result = new org.tensorflow.framework.GPUOptions.Experimental(this);
int from_bitField0_ = bitField0_;
int to_bitField0_ = 0;
if (virtualDevicesBuilder_ == null) {
if (((bitField0_ & 0x00000001) == 0x00000001)) {
virtualDevices_ = java.util.Collections.unmodifiableList(virtualDevices_);
bitField0_ = (bitField0_ & ~0x00000001);
}
result.virtualDevices_ = virtualDevices_;
} else {
result.virtualDevices_ = virtualDevicesBuilder_.build();
}
result.useUnifiedMemory_ = useUnifiedMemory_;
result.numDevToDevCopyStreams_ = numDevToDevCopyStreams_;
result.collectiveRingOrder_ = collectiveRingOrder_;
result.timestampedAllocator_ = timestampedAllocator_;
result.kernelTrackerMaxInterval_ = kernelTrackerMaxInterval_;
result.kernelTrackerMaxBytes_ = kernelTrackerMaxBytes_;
result.kernelTrackerMaxPending_ = kernelTrackerMaxPending_;
result.bitField0_ = to_bitField0_;
onBuilt();
return result;
}
public Builder clone() {
return (Builder) super.clone();
}
public Builder setField(
com.google.protobuf.Descriptors.FieldDescriptor field,
java.lang.Object value) {
return (Builder) super.setField(field, value);
}
public Builder clearField(
com.google.protobuf.Descriptors.FieldDescriptor field) {
return (Builder) super.clearField(field);
}
public Builder clearOneof(
com.google.protobuf.Descriptors.OneofDescriptor oneof) {
return (Builder) super.clearOneof(oneof);
}
public Builder setRepeatedField(
com.google.protobuf.Descriptors.FieldDescriptor field,
int index, java.lang.Object value) {
return (Builder) super.setRepeatedField(field, index, value);
}
public Builder addRepeatedField(
com.google.protobuf.Descriptors.FieldDescriptor field,
java.lang.Object value) {
return (Builder) super.addRepeatedField(field, value);
}
public Builder mergeFrom(com.google.protobuf.Message other) {
if (other instanceof org.tensorflow.framework.GPUOptions.Experimental) {
return mergeFrom((org.tensorflow.framework.GPUOptions.Experimental)other);
} else {
super.mergeFrom(other);
return this;
}
}
public Builder mergeFrom(org.tensorflow.framework.GPUOptions.Experimental other) {
if (other == org.tensorflow.framework.GPUOptions.Experimental.getDefaultInstance()) return this;
if (virtualDevicesBuilder_ == null) {
if (!other.virtualDevices_.isEmpty()) {
if (virtualDevices_.isEmpty()) {
virtualDevices_ = other.virtualDevices_;
bitField0_ = (bitField0_ & ~0x00000001);
} else {
ensureVirtualDevicesIsMutable();
virtualDevices_.addAll(other.virtualDevices_);
}
onChanged();
}
} else {
if (!other.virtualDevices_.isEmpty()) {
if (virtualDevicesBuilder_.isEmpty()) {
virtualDevicesBuilder_.dispose();
virtualDevicesBuilder_ = null;
virtualDevices_ = other.virtualDevices_;
bitField0_ = (bitField0_ & ~0x00000001);
virtualDevicesBuilder_ =
com.google.protobuf.GeneratedMessageV3.alwaysUseFieldBuilders ?
getVirtualDevicesFieldBuilder() : null;
} else {
virtualDevicesBuilder_.addAllMessages(other.virtualDevices_);
}
}
}
if (other.getUseUnifiedMemory() != false) {
setUseUnifiedMemory(other.getUseUnifiedMemory());
}
if (other.getNumDevToDevCopyStreams() != 0) {
setNumDevToDevCopyStreams(other.getNumDevToDevCopyStreams());
}
if (!other.getCollectiveRingOrder().isEmpty()) {
collectiveRingOrder_ = other.collectiveRingOrder_;
onChanged();
}
if (other.getTimestampedAllocator() != false) {
setTimestampedAllocator(other.getTimestampedAllocator());
}
if (other.getKernelTrackerMaxInterval() != 0) {
setKernelTrackerMaxInterval(other.getKernelTrackerMaxInterval());
}
if (other.getKernelTrackerMaxBytes() != 0) {
setKernelTrackerMaxBytes(other.getKernelTrackerMaxBytes());
}
if (other.getKernelTrackerMaxPending() != 0) {
setKernelTrackerMaxPending(other.getKernelTrackerMaxPending());
}
this.mergeUnknownFields(other.unknownFields);
onChanged();
return this;
}
public final boolean isInitialized() {
return true;
}
public Builder mergeFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
org.tensorflow.framework.GPUOptions.Experimental parsedMessage = null;
try {
parsedMessage = PARSER.parsePartialFrom(input, extensionRegistry);
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
parsedMessage = (org.tensorflow.framework.GPUOptions.Experimental) e.getUnfinishedMessage();
throw e.unwrapIOException();
} finally {
if (parsedMessage != null) {
mergeFrom(parsedMessage);
}
}
return this;
}
private int bitField0_;
private java.util.List virtualDevices_ =
java.util.Collections.emptyList();
private void ensureVirtualDevicesIsMutable() {
if (!((bitField0_ & 0x00000001) == 0x00000001)) {
virtualDevices_ = new java.util.ArrayList(virtualDevices_);
bitField0_ |= 0x00000001;
}
}
private com.google.protobuf.RepeatedFieldBuilderV3<
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevicesOrBuilder> virtualDevicesBuilder_;
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public java.util.List getVirtualDevicesList() {
if (virtualDevicesBuilder_ == null) {
return java.util.Collections.unmodifiableList(virtualDevices_);
} else {
return virtualDevicesBuilder_.getMessageList();
}
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public int getVirtualDevicesCount() {
if (virtualDevicesBuilder_ == null) {
return virtualDevices_.size();
} else {
return virtualDevicesBuilder_.getCount();
}
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices getVirtualDevices(int index) {
if (virtualDevicesBuilder_ == null) {
return virtualDevices_.get(index);
} else {
return virtualDevicesBuilder_.getMessage(index);
}
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder setVirtualDevices(
int index, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices value) {
if (virtualDevicesBuilder_ == null) {
if (value == null) {
throw new NullPointerException();
}
ensureVirtualDevicesIsMutable();
virtualDevices_.set(index, value);
onChanged();
} else {
virtualDevicesBuilder_.setMessage(index, value);
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder setVirtualDevices(
int index, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder builderForValue) {
if (virtualDevicesBuilder_ == null) {
ensureVirtualDevicesIsMutable();
virtualDevices_.set(index, builderForValue.build());
onChanged();
} else {
virtualDevicesBuilder_.setMessage(index, builderForValue.build());
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder addVirtualDevices(org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices value) {
if (virtualDevicesBuilder_ == null) {
if (value == null) {
throw new NullPointerException();
}
ensureVirtualDevicesIsMutable();
virtualDevices_.add(value);
onChanged();
} else {
virtualDevicesBuilder_.addMessage(value);
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder addVirtualDevices(
int index, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices value) {
if (virtualDevicesBuilder_ == null) {
if (value == null) {
throw new NullPointerException();
}
ensureVirtualDevicesIsMutable();
virtualDevices_.add(index, value);
onChanged();
} else {
virtualDevicesBuilder_.addMessage(index, value);
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder addVirtualDevices(
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder builderForValue) {
if (virtualDevicesBuilder_ == null) {
ensureVirtualDevicesIsMutable();
virtualDevices_.add(builderForValue.build());
onChanged();
} else {
virtualDevicesBuilder_.addMessage(builderForValue.build());
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder addVirtualDevices(
int index, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder builderForValue) {
if (virtualDevicesBuilder_ == null) {
ensureVirtualDevicesIsMutable();
virtualDevices_.add(index, builderForValue.build());
onChanged();
} else {
virtualDevicesBuilder_.addMessage(index, builderForValue.build());
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder addAllVirtualDevices(
java.lang.Iterable values) {
if (virtualDevicesBuilder_ == null) {
ensureVirtualDevicesIsMutable();
com.google.protobuf.AbstractMessageLite.Builder.addAll(
values, virtualDevices_);
onChanged();
} else {
virtualDevicesBuilder_.addAllMessages(values);
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder clearVirtualDevices() {
if (virtualDevicesBuilder_ == null) {
virtualDevices_ = java.util.Collections.emptyList();
bitField0_ = (bitField0_ & ~0x00000001);
onChanged();
} else {
virtualDevicesBuilder_.clear();
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public Builder removeVirtualDevices(int index) {
if (virtualDevicesBuilder_ == null) {
ensureVirtualDevicesIsMutable();
virtualDevices_.remove(index);
onChanged();
} else {
virtualDevicesBuilder_.remove(index);
}
return this;
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder getVirtualDevicesBuilder(
int index) {
return getVirtualDevicesFieldBuilder().getBuilder(index);
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevicesOrBuilder getVirtualDevicesOrBuilder(
int index) {
if (virtualDevicesBuilder_ == null) {
return virtualDevices_.get(index); } else {
return virtualDevicesBuilder_.getMessageOrBuilder(index);
}
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public java.util.List
getVirtualDevicesOrBuilderList() {
if (virtualDevicesBuilder_ != null) {
return virtualDevicesBuilder_.getMessageOrBuilderList();
} else {
return java.util.Collections.unmodifiableList(virtualDevices_);
}
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder addVirtualDevicesBuilder() {
return getVirtualDevicesFieldBuilder().addBuilder(
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.getDefaultInstance());
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder addVirtualDevicesBuilder(
int index) {
return getVirtualDevicesFieldBuilder().addBuilder(
index, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.getDefaultInstance());
}
/**
*
* The multi virtual device settings. If empty (not set), it will create
* single virtual device on each visible GPU, according to the settings
* in "visible_device_list" above. Otherwise, the number of elements in the
* list must be the same as the number of visible GPUs (after
* "visible_device_list" filtering if it is set), and the string represented
* device names (e.g. /device:GPU:<id>) will refer to the virtual
* devices and have the <id> field assigned sequentially starting from 0,
* according to the order they appear in this list and the "memory_limit"
* list inside each element. For example,
* visible_device_list = "1,0"
* virtual_devices { memory_limit: 1GB memory_limit: 2GB }
* virtual_devices {}
* will create three virtual devices as:
* /device:GPU:0 -> visible GPU 1 with 1GB memory
* /device:GPU:1 -> visible GPU 1 with 2GB memory
* /device:GPU:2 -> visible GPU 0 with all available memory
* NOTE:
* 1. It's invalid to set both this and "per_process_gpu_memory_fraction"
* at the same time.
* 2. Currently this setting is per-process, not per-session. Using
* different settings in different sessions within same process will
* result in undefined behavior.
*
*
* repeated .tensorflow.GPUOptions.Experimental.VirtualDevices virtual_devices = 1;
*/
public java.util.List
getVirtualDevicesBuilderList() {
return getVirtualDevicesFieldBuilder().getBuilderList();
}
private com.google.protobuf.RepeatedFieldBuilderV3<
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevicesOrBuilder>
getVirtualDevicesFieldBuilder() {
if (virtualDevicesBuilder_ == null) {
virtualDevicesBuilder_ = new com.google.protobuf.RepeatedFieldBuilderV3<
org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevices.Builder, org.tensorflow.framework.GPUOptions.Experimental.VirtualDevicesOrBuilder>(
virtualDevices_,
((bitField0_ & 0x00000001) == 0x00000001),
getParentForChildren(),
isClean());
virtualDevices_ = null;
}
return virtualDevicesBuilder_;
}
private boolean useUnifiedMemory_ ;
/**
*
* If true, uses CUDA unified memory for memory allocations. If
* per_process_gpu_memory_fraction option is greater than 1.0, then unified
* memory is used regardless of the value for this field. See comments for
* per_process_gpu_memory_fraction field for more details and requirements
* of the unified memory. This option is useful to oversubscribe memory if
* multiple processes are sharing a single GPU while individually using less
* than 1.0 per process memory fraction.
*
*
* bool use_unified_memory = 2;
*/
public boolean getUseUnifiedMemory() {
return useUnifiedMemory_;
}
/**
*
* If true, uses CUDA unified memory for memory allocations. If
* per_process_gpu_memory_fraction option is greater than 1.0, then unified
* memory is used regardless of the value for this field. See comments for
* per_process_gpu_memory_fraction field for more details and requirements
* of the unified memory. This option is useful to oversubscribe memory if
* multiple processes are sharing a single GPU while individually using less
* than 1.0 per process memory fraction.
*
*
* bool use_unified_memory = 2;
*/
public Builder setUseUnifiedMemory(boolean value) {
useUnifiedMemory_ = value;
onChanged();
return this;
}
/**
*
* If true, uses CUDA unified memory for memory allocations. If
* per_process_gpu_memory_fraction option is greater than 1.0, then unified
* memory is used regardless of the value for this field. See comments for
* per_process_gpu_memory_fraction field for more details and requirements
* of the unified memory. This option is useful to oversubscribe memory if
* multiple processes are sharing a single GPU while individually using less
* than 1.0 per process memory fraction.
*
*
* bool use_unified_memory = 2;
*/
public Builder clearUseUnifiedMemory() {
useUnifiedMemory_ = false;
onChanged();
return this;
}
private int numDevToDevCopyStreams_ ;
/**
*
* If > 1, the number of device-to-device copy streams to create
* for each GPUDevice. Default value is 0, which is automatically
* converted to 1.
*
*
* int32 num_dev_to_dev_copy_streams = 3;
*/
public int getNumDevToDevCopyStreams() {
return numDevToDevCopyStreams_;
}
/**
*
* If > 1, the number of device-to-device copy streams to create
* for each GPUDevice. Default value is 0, which is automatically
* converted to 1.
*
*
* int32 num_dev_to_dev_copy_streams = 3;
*/
public Builder setNumDevToDevCopyStreams(int value) {
numDevToDevCopyStreams_ = value;
onChanged();
return this;
}
/**
*
* If > 1, the number of device-to-device copy streams to create
* for each GPUDevice. Default value is 0, which is automatically
* converted to 1.
*
*
* int32 num_dev_to_dev_copy_streams = 3;
*/
public Builder clearNumDevToDevCopyStreams() {
numDevToDevCopyStreams_ = 0;
onChanged();
return this;
}
private java.lang.Object collectiveRingOrder_ = "";
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
public java.lang.String getCollectiveRingOrder() {
java.lang.Object ref = collectiveRingOrder_;
if (!(ref instanceof java.lang.String)) {
com.google.protobuf.ByteString bs =
(com.google.protobuf.ByteString) ref;
java.lang.String s = bs.toStringUtf8();
collectiveRingOrder_ = s;
return s;
} else {
return (java.lang.String) ref;
}
}
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
public com.google.protobuf.ByteString
getCollectiveRingOrderBytes() {
java.lang.Object ref = collectiveRingOrder_;
if (ref instanceof String) {
com.google.protobuf.ByteString b =
com.google.protobuf.ByteString.copyFromUtf8(
(java.lang.String) ref);
collectiveRingOrder_ = b;
return b;
} else {
return (com.google.protobuf.ByteString) ref;
}
}
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
public Builder setCollectiveRingOrder(
java.lang.String value) {
if (value == null) {
throw new NullPointerException();
}
collectiveRingOrder_ = value;
onChanged();
return this;
}
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
public Builder clearCollectiveRingOrder() {
collectiveRingOrder_ = getDefaultInstance().getCollectiveRingOrder();
onChanged();
return this;
}
/**
*
* If non-empty, defines a good GPU ring order on a single worker based on
* device interconnect. This assumes that all workers have the same GPU
* topology. Specify as a comma-separated string, e.g. "3,2,1,0,7,6,5,4".
* This ring order is used by the RingReducer implementation of
* CollectiveReduce, and serves as an override to automatic ring order
* generation in OrderTaskDeviceMap() during CollectiveParam resolution.
*
*
* string collective_ring_order = 4;
*/
public Builder setCollectiveRingOrderBytes(
com.google.protobuf.ByteString value) {
if (value == null) {
throw new NullPointerException();
}
checkByteStringIsUtf8(value);
collectiveRingOrder_ = value;
onChanged();
return this;
}
private boolean timestampedAllocator_ ;
/**
*
* If true then extra work is done by GPUDevice and GPUBFCAllocator to
* keep track of when GPU memory is freed and when kernels actually
* complete so that we can know when a nominally free memory chunk
* is really not subject to pending use.
*
*
* bool timestamped_allocator = 5;
*/
public boolean getTimestampedAllocator() {
return timestampedAllocator_;
}
/**
*
* If true then extra work is done by GPUDevice and GPUBFCAllocator to
* keep track of when GPU memory is freed and when kernels actually
* complete so that we can know when a nominally free memory chunk
* is really not subject to pending use.
*
*
* bool timestamped_allocator = 5;
*/
public Builder setTimestampedAllocator(boolean value) {
timestampedAllocator_ = value;
onChanged();
return this;
}
/**
*
* If true then extra work is done by GPUDevice and GPUBFCAllocator to
* keep track of when GPU memory is freed and when kernels actually
* complete so that we can know when a nominally free memory chunk
* is really not subject to pending use.
*
*
* bool timestamped_allocator = 5;
*/
public Builder clearTimestampedAllocator() {
timestampedAllocator_ = false;
onChanged();
return this;
}
private int kernelTrackerMaxInterval_ ;
/**
*
* Parameters for GPUKernelTracker. By default no kernel tracking is done.
* Note that timestamped_allocator is only effective if some tracking is
* specified.
* If kernel_tracker_max_interval = n > 0, then a tracking event
* is inserted after every n kernels without an event.
*
*
* int32 kernel_tracker_max_interval = 7;
*/
public int getKernelTrackerMaxInterval() {
return kernelTrackerMaxInterval_;
}
/**
*
* Parameters for GPUKernelTracker. By default no kernel tracking is done.
* Note that timestamped_allocator is only effective if some tracking is
* specified.
* If kernel_tracker_max_interval = n > 0, then a tracking event
* is inserted after every n kernels without an event.
*
*
* int32 kernel_tracker_max_interval = 7;
*/
public Builder setKernelTrackerMaxInterval(int value) {
kernelTrackerMaxInterval_ = value;
onChanged();
return this;
}
/**
*
* Parameters for GPUKernelTracker. By default no kernel tracking is done.
* Note that timestamped_allocator is only effective if some tracking is
* specified.
* If kernel_tracker_max_interval = n > 0, then a tracking event
* is inserted after every n kernels without an event.
*
*
* int32 kernel_tracker_max_interval = 7;
*/
public Builder clearKernelTrackerMaxInterval() {
kernelTrackerMaxInterval_ = 0;
onChanged();
return this;
}
private int kernelTrackerMaxBytes_ ;
/**
*
* If kernel_tracker_max_bytes = n > 0, then a tracking event is
* inserted after every series of kernels allocating a sum of
* memory >= n. If one kernel allocates b * n bytes, then one
* event will be inserted after it, but it will count as b against
* the pending limit.
*
*
* int32 kernel_tracker_max_bytes = 8;
*/
public int getKernelTrackerMaxBytes() {
return kernelTrackerMaxBytes_;
}
/**
*
* If kernel_tracker_max_bytes = n > 0, then a tracking event is
* inserted after every series of kernels allocating a sum of
* memory >= n. If one kernel allocates b * n bytes, then one
* event will be inserted after it, but it will count as b against
* the pending limit.
*
*
* int32 kernel_tracker_max_bytes = 8;
*/
public Builder setKernelTrackerMaxBytes(int value) {
kernelTrackerMaxBytes_ = value;
onChanged();
return this;
}
/**
*
* If kernel_tracker_max_bytes = n > 0, then a tracking event is
* inserted after every series of kernels allocating a sum of
* memory >= n. If one kernel allocates b * n bytes, then one
* event will be inserted after it, but it will count as b against
* the pending limit.
*
*
* int32 kernel_tracker_max_bytes = 8;
*/
public Builder clearKernelTrackerMaxBytes() {
kernelTrackerMaxBytes_ = 0;
onChanged();
return this;
}
private int kernelTrackerMaxPending_ ;
/**
*
* If kernel_tracker_max_pending > 0 then no more than this many
* tracking events can be outstanding at a time. An attempt to
* launch an additional kernel will stall until an event
* completes.
*
*
* int32 kernel_tracker_max_pending = 9;
*/
public int getKernelTrackerMaxPending() {
return kernelTrackerMaxPending_;
}
/**
*
* If kernel_tracker_max_pending > 0 then no more than this many
* tracking events can be outstanding at a time. An attempt to
* launch an additional kernel will stall until an event
* completes.
*
*
* int32 kernel_tracker_max_pending = 9;
*/
public Builder setKernelTrackerMaxPending(int value) {
kernelTrackerMaxPending_ = value;
onChanged();
return this;
}
/**
*
* If kernel_tracker_max_pending > 0 then no more than this many
* tracking events can be outstanding at a time. An attempt to
* launch an additional kernel will stall until an event
* completes.
*
*
* int32 kernel_tracker_max_pending = 9;
*/
public Builder clearKernelTrackerMaxPending() {
kernelTrackerMaxPending_ = 0;
onChanged();
return this;
}
public final Builder setUnknownFields(
final com.google.protobuf.UnknownFieldSet unknownFields) {
return super.setUnknownFieldsProto3(unknownFields);
}
public final Builder mergeUnknownFields(
final com.google.protobuf.UnknownFieldSet unknownFields) {
return super.mergeUnknownFields(unknownFields);
}
// @@protoc_insertion_point(builder_scope:tensorflow.GPUOptions.Experimental)
}
// @@protoc_insertion_point(class_scope:tensorflow.GPUOptions.Experimental)
private static final org.tensorflow.framework.GPUOptions.Experimental DEFAULT_INSTANCE;
static {
DEFAULT_INSTANCE = new org.tensorflow.framework.GPUOptions.Experimental();
}
public static org.tensorflow.framework.GPUOptions.Experimental getDefaultInstance() {
return DEFAULT_INSTANCE;
}
private static final com.google.protobuf.Parser
PARSER = new com.google.protobuf.AbstractParser() {
public Experimental parsePartialFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return new Experimental(input, extensionRegistry);
}
};
public static com.google.protobuf.Parser parser() {
return PARSER;
}
@java.lang.Override
public com.google.protobuf.Parser getParserForType() {
return PARSER;
}
public org.tensorflow.framework.GPUOptions.Experimental getDefaultInstanceForType() {
return DEFAULT_INSTANCE;
}
}
public static final int PER_PROCESS_GPU_MEMORY_FRACTION_FIELD_NUMBER = 1;
private double perProcessGpuMemoryFraction_;
/**
*
* Fraction of the available GPU memory to allocate for each process.
* 1 means to allocate all of the GPU memory, 0.5 means the process
* allocates up to ~50% of the available GPU memory.
* GPU memory is pre-allocated unless the allow_growth option is enabled.
* If greater than 1.0, uses CUDA unified memory to potentially oversubscribe
* the amount of memory available on the GPU device by using host memory as a
* swap space. Accessing memory not available on the device will be
* significantly slower as that would require memory transfer between the host
* and the device. Options to reduce the memory requirement should be
* considered before enabling this option as this may come with a negative
* performance impact. Oversubscription using the unified memory requires
* Pascal class or newer GPUs and it is currently only supported on the Linux
* operating system. See
* https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#um-requirements
* for the detailed requirements.
*
*
* double per_process_gpu_memory_fraction = 1;
*/
public double getPerProcessGpuMemoryFraction() {
return perProcessGpuMemoryFraction_;
}
public static final int ALLOW_GROWTH_FIELD_NUMBER = 4;
private boolean allowGrowth_;
/**
*
* If true, the allocator does not pre-allocate the entire specified
* GPU memory region, instead starting small and growing as needed.
*
*
* bool allow_growth = 4;
*/
public boolean getAllowGrowth() {
return allowGrowth_;
}
public static final int ALLOCATOR_TYPE_FIELD_NUMBER = 2;
private volatile java.lang.Object allocatorType_;
/**
*
* The type of GPU allocation strategy to use.
* Allowed values:
* "": The empty string (default) uses a system-chosen default
* which may change over time.
* "BFC": A "Best-fit with coalescing" algorithm, simplified from a
* version of dlmalloc.
*
*
* string allocator_type = 2;
*/
public java.lang.String getAllocatorType() {
java.lang.Object ref = allocatorType_;
if (ref instanceof java.lang.String) {
return (java.lang.String) ref;
} else {
com.google.protobuf.ByteString bs =
(com.google.protobuf.ByteString) ref;
java.lang.String s = bs.toStringUtf8();
allocatorType_ = s;
return s;
}
}
/**
*
* The type of GPU allocation strategy to use.
* Allowed values:
* "": The empty string (default) uses a system-chosen default
* which may change over time.
* "BFC": A "Best-fit with coalescing" algorithm, simplified from a
* version of dlmalloc.
*
*
* string allocator_type = 2;
*/
public com.google.protobuf.ByteString
getAllocatorTypeBytes() {
java.lang.Object ref = allocatorType_;
if (ref instanceof java.lang.String) {
com.google.protobuf.ByteString b =
com.google.protobuf.ByteString.copyFromUtf8(
(java.lang.String) ref);
allocatorType_ = b;
return b;
} else {
return (com.google.protobuf.ByteString) ref;
}
}
public static final int DEFERRED_DELETION_BYTES_FIELD_NUMBER = 3;
private long deferredDeletionBytes_;
/**
*
* Delay deletion of up to this many bytes to reduce the number of
* interactions with gpu driver code. If 0, the system chooses
* a reasonable default (several MBs).
*
*
* int64 deferred_deletion_bytes = 3;
*/
public long getDeferredDeletionBytes() {
return deferredDeletionBytes_;
}
public static final int VISIBLE_DEVICE_LIST_FIELD_NUMBER = 5;
private volatile java.lang.Object visibleDeviceList_;
/**
*
* A comma-separated list of GPU ids that determines the 'visible'
* to 'virtual' mapping of GPU devices. For example, if TensorFlow
* can see 8 GPU devices in the process, and one wanted to map
* visible GPU devices 5 and 3 as "/device:GPU:0", and "/device:GPU:1",
* then one would specify this field as "5,3". This field is similar in
* spirit to the CUDA_VISIBLE_DEVICES environment variable, except
* it applies to the visible GPU devices in the process.
* NOTE:
* 1. The GPU driver provides the process with the visible GPUs
* in an order which is not guaranteed to have any correlation to
* the *physical* GPU id in the machine. This field is used for
* remapping "visible" to "virtual", which means this operates only
* after the process starts. Users are required to use vendor
* specific mechanisms (e.g., CUDA_VISIBLE_DEVICES) to control the
* physical to visible device mapping prior to invoking TensorFlow.
* 2. In the code, the ids in this list are also called "platform GPU id"s,
* and the 'virtual' ids of GPU devices (i.e. the ids in the device
* name "/device:GPU:<id>") are also called "TF GPU id"s. Please
* refer to third_party/tensorflow/core/common_runtime/gpu/gpu_id.h
* for more information.
*
*
* string visible_device_list = 5;
*/
public java.lang.String getVisibleDeviceList() {
java.lang.Object ref = visibleDeviceList_;
if (ref instanceof java.lang.String) {
return (java.lang.String) ref;
} else {
com.google.protobuf.ByteString bs =
(com.google.protobuf.ByteString) ref;
java.lang.String s = bs.toStringUtf8();
visibleDeviceList_ = s;
return s;
}
}
/**
*
* A comma-separated list of GPU ids that determines the 'visible'
* to 'virtual' mapping of GPU devices. For example, if TensorFlow
* can see 8 GPU devices in the process, and one wanted to map
* visible GPU devices 5 and 3 as "/device:GPU:0", and "/device:GPU:1",
* then one would specify this field as "5,3". This field is similar in
* spirit to the CUDA_VISIBLE_DEVICES environment variable, except
* it applies to the visible GPU devices in the process.
* NOTE:
* 1. The GPU driver provides the process with the visible GPUs
* in an order which is not guaranteed to have any correlation to
* the *physical* GPU id in the machine. This field is used for
* remapping "visible" to "virtual", which means this operates only
* after the process starts. Users are required to use vendor
* specific mechanisms (e.g., CUDA_VISIBLE_DEVICES) to control the
* physical to visible device mapping prior to invoking TensorFlow.
* 2. In the code, the ids in this list are also called "platform GPU id"s,
* and the 'virtual' ids of GPU devices (i.e. the ids in the device
* name "/device:GPU:<id>") are also called "TF GPU id"s. Please
* refer to third_party/tensorflow/core/common_runtime/gpu/gpu_id.h
* for more information.
*
*
* string visible_device_list = 5;
*/
public com.google.protobuf.ByteString
getVisibleDeviceListBytes() {
java.lang.Object ref = visibleDeviceList_;
if (ref instanceof java.lang.String) {
com.google.protobuf.ByteString b =
com.google.protobuf.ByteString.copyFromUtf8(
(java.lang.String) ref);
visibleDeviceList_ = b;
return b;
} else {
return (com.google.protobuf.ByteString) ref;
}
}
public static final int POLLING_ACTIVE_DELAY_USECS_FIELD_NUMBER = 6;
private int pollingActiveDelayUsecs_;
/**
*
* In the event polling loop sleep this many microseconds between
* PollEvents calls, when the queue is not empty. If value is not
* set or set to 0, gets set to a non-zero default.
*
*
* int32 polling_active_delay_usecs = 6;
*/
public int getPollingActiveDelayUsecs() {
return pollingActiveDelayUsecs_;
}
public static final int POLLING_INACTIVE_DELAY_MSECS_FIELD_NUMBER = 7;
private int pollingInactiveDelayMsecs_;
/**
*
* This field is deprecated and ignored.
*
*
* int32 polling_inactive_delay_msecs = 7;
*/
public int getPollingInactiveDelayMsecs() {
return pollingInactiveDelayMsecs_;
}
public static final int FORCE_GPU_COMPATIBLE_FIELD_NUMBER = 8;
private boolean forceGpuCompatible_;
/**
*
* Force all tensors to be gpu_compatible. On a GPU-enabled TensorFlow,
* enabling this option forces all CPU tensors to be allocated with Cuda
* pinned memory. Normally, TensorFlow will infer which tensors should be
* allocated as the pinned memory. But in case where the inference is
* incomplete, this option can significantly speed up the cross-device memory
* copy performance as long as it fits the memory.
* Note that this option is not something that should be
* enabled by default for unknown or very large models, since all Cuda pinned
* memory is unpageable, having too much pinned memory might negatively impact
* the overall host system performance.
*
*
* bool force_gpu_compatible = 8;
*/
public boolean getForceGpuCompatible() {
return forceGpuCompatible_;
}
public static final int EXPERIMENTAL_FIELD_NUMBER = 9;
private org.tensorflow.framework.GPUOptions.Experimental experimental_;
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public boolean hasExperimental() {
return experimental_ != null;
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public org.tensorflow.framework.GPUOptions.Experimental getExperimental() {
return experimental_ == null ? org.tensorflow.framework.GPUOptions.Experimental.getDefaultInstance() : experimental_;
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public org.tensorflow.framework.GPUOptions.ExperimentalOrBuilder getExperimentalOrBuilder() {
return getExperimental();
}
private byte memoizedIsInitialized = -1;
public final boolean isInitialized() {
byte isInitialized = memoizedIsInitialized;
if (isInitialized == 1) return true;
if (isInitialized == 0) return false;
memoizedIsInitialized = 1;
return true;
}
public void writeTo(com.google.protobuf.CodedOutputStream output)
throws java.io.IOException {
if (perProcessGpuMemoryFraction_ != 0D) {
output.writeDouble(1, perProcessGpuMemoryFraction_);
}
if (!getAllocatorTypeBytes().isEmpty()) {
com.google.protobuf.GeneratedMessageV3.writeString(output, 2, allocatorType_);
}
if (deferredDeletionBytes_ != 0L) {
output.writeInt64(3, deferredDeletionBytes_);
}
if (allowGrowth_ != false) {
output.writeBool(4, allowGrowth_);
}
if (!getVisibleDeviceListBytes().isEmpty()) {
com.google.protobuf.GeneratedMessageV3.writeString(output, 5, visibleDeviceList_);
}
if (pollingActiveDelayUsecs_ != 0) {
output.writeInt32(6, pollingActiveDelayUsecs_);
}
if (pollingInactiveDelayMsecs_ != 0) {
output.writeInt32(7, pollingInactiveDelayMsecs_);
}
if (forceGpuCompatible_ != false) {
output.writeBool(8, forceGpuCompatible_);
}
if (experimental_ != null) {
output.writeMessage(9, getExperimental());
}
unknownFields.writeTo(output);
}
public int getSerializedSize() {
int size = memoizedSize;
if (size != -1) return size;
size = 0;
if (perProcessGpuMemoryFraction_ != 0D) {
size += com.google.protobuf.CodedOutputStream
.computeDoubleSize(1, perProcessGpuMemoryFraction_);
}
if (!getAllocatorTypeBytes().isEmpty()) {
size += com.google.protobuf.GeneratedMessageV3.computeStringSize(2, allocatorType_);
}
if (deferredDeletionBytes_ != 0L) {
size += com.google.protobuf.CodedOutputStream
.computeInt64Size(3, deferredDeletionBytes_);
}
if (allowGrowth_ != false) {
size += com.google.protobuf.CodedOutputStream
.computeBoolSize(4, allowGrowth_);
}
if (!getVisibleDeviceListBytes().isEmpty()) {
size += com.google.protobuf.GeneratedMessageV3.computeStringSize(5, visibleDeviceList_);
}
if (pollingActiveDelayUsecs_ != 0) {
size += com.google.protobuf.CodedOutputStream
.computeInt32Size(6, pollingActiveDelayUsecs_);
}
if (pollingInactiveDelayMsecs_ != 0) {
size += com.google.protobuf.CodedOutputStream
.computeInt32Size(7, pollingInactiveDelayMsecs_);
}
if (forceGpuCompatible_ != false) {
size += com.google.protobuf.CodedOutputStream
.computeBoolSize(8, forceGpuCompatible_);
}
if (experimental_ != null) {
size += com.google.protobuf.CodedOutputStream
.computeMessageSize(9, getExperimental());
}
size += unknownFields.getSerializedSize();
memoizedSize = size;
return size;
}
@java.lang.Override
public boolean equals(final java.lang.Object obj) {
if (obj == this) {
return true;
}
if (!(obj instanceof org.tensorflow.framework.GPUOptions)) {
return super.equals(obj);
}
org.tensorflow.framework.GPUOptions other = (org.tensorflow.framework.GPUOptions) obj;
boolean result = true;
result = result && (
java.lang.Double.doubleToLongBits(getPerProcessGpuMemoryFraction())
== java.lang.Double.doubleToLongBits(
other.getPerProcessGpuMemoryFraction()));
result = result && (getAllowGrowth()
== other.getAllowGrowth());
result = result && getAllocatorType()
.equals(other.getAllocatorType());
result = result && (getDeferredDeletionBytes()
== other.getDeferredDeletionBytes());
result = result && getVisibleDeviceList()
.equals(other.getVisibleDeviceList());
result = result && (getPollingActiveDelayUsecs()
== other.getPollingActiveDelayUsecs());
result = result && (getPollingInactiveDelayMsecs()
== other.getPollingInactiveDelayMsecs());
result = result && (getForceGpuCompatible()
== other.getForceGpuCompatible());
result = result && (hasExperimental() == other.hasExperimental());
if (hasExperimental()) {
result = result && getExperimental()
.equals(other.getExperimental());
}
result = result && unknownFields.equals(other.unknownFields);
return result;
}
@java.lang.Override
public int hashCode() {
if (memoizedHashCode != 0) {
return memoizedHashCode;
}
int hash = 41;
hash = (19 * hash) + getDescriptor().hashCode();
hash = (37 * hash) + PER_PROCESS_GPU_MEMORY_FRACTION_FIELD_NUMBER;
hash = (53 * hash) + com.google.protobuf.Internal.hashLong(
java.lang.Double.doubleToLongBits(getPerProcessGpuMemoryFraction()));
hash = (37 * hash) + ALLOW_GROWTH_FIELD_NUMBER;
hash = (53 * hash) + com.google.protobuf.Internal.hashBoolean(
getAllowGrowth());
hash = (37 * hash) + ALLOCATOR_TYPE_FIELD_NUMBER;
hash = (53 * hash) + getAllocatorType().hashCode();
hash = (37 * hash) + DEFERRED_DELETION_BYTES_FIELD_NUMBER;
hash = (53 * hash) + com.google.protobuf.Internal.hashLong(
getDeferredDeletionBytes());
hash = (37 * hash) + VISIBLE_DEVICE_LIST_FIELD_NUMBER;
hash = (53 * hash) + getVisibleDeviceList().hashCode();
hash = (37 * hash) + POLLING_ACTIVE_DELAY_USECS_FIELD_NUMBER;
hash = (53 * hash) + getPollingActiveDelayUsecs();
hash = (37 * hash) + POLLING_INACTIVE_DELAY_MSECS_FIELD_NUMBER;
hash = (53 * hash) + getPollingInactiveDelayMsecs();
hash = (37 * hash) + FORCE_GPU_COMPATIBLE_FIELD_NUMBER;
hash = (53 * hash) + com.google.protobuf.Internal.hashBoolean(
getForceGpuCompatible());
if (hasExperimental()) {
hash = (37 * hash) + EXPERIMENTAL_FIELD_NUMBER;
hash = (53 * hash) + getExperimental().hashCode();
}
hash = (29 * hash) + unknownFields.hashCode();
memoizedHashCode = hash;
return hash;
}
public static org.tensorflow.framework.GPUOptions parseFrom(
java.nio.ByteBuffer data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions parseFrom(
java.nio.ByteBuffer data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions parseFrom(
com.google.protobuf.ByteString data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions parseFrom(
com.google.protobuf.ByteString data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions parseFrom(byte[] data)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data);
}
public static org.tensorflow.framework.GPUOptions parseFrom(
byte[] data,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return PARSER.parseFrom(data, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions parseFrom(java.io.InputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions parseFrom(
java.io.InputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions parseDelimitedFrom(java.io.InputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseDelimitedWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions parseDelimitedFrom(
java.io.InputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseDelimitedWithIOException(PARSER, input, extensionRegistry);
}
public static org.tensorflow.framework.GPUOptions parseFrom(
com.google.protobuf.CodedInputStream input)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input);
}
public static org.tensorflow.framework.GPUOptions parseFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
return com.google.protobuf.GeneratedMessageV3
.parseWithIOException(PARSER, input, extensionRegistry);
}
public Builder newBuilderForType() { return newBuilder(); }
public static Builder newBuilder() {
return DEFAULT_INSTANCE.toBuilder();
}
public static Builder newBuilder(org.tensorflow.framework.GPUOptions prototype) {
return DEFAULT_INSTANCE.toBuilder().mergeFrom(prototype);
}
public Builder toBuilder() {
return this == DEFAULT_INSTANCE
? new Builder() : new Builder().mergeFrom(this);
}
@java.lang.Override
protected Builder newBuilderForType(
com.google.protobuf.GeneratedMessageV3.BuilderParent parent) {
Builder builder = new Builder(parent);
return builder;
}
/**
* Protobuf type {@code tensorflow.GPUOptions}
*/
public static final class Builder extends
com.google.protobuf.GeneratedMessageV3.Builder implements
// @@protoc_insertion_point(builder_implements:tensorflow.GPUOptions)
org.tensorflow.framework.GPUOptionsOrBuilder {
public static final com.google.protobuf.Descriptors.Descriptor
getDescriptor() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_descriptor;
}
protected com.google.protobuf.GeneratedMessageV3.FieldAccessorTable
internalGetFieldAccessorTable() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_fieldAccessorTable
.ensureFieldAccessorsInitialized(
org.tensorflow.framework.GPUOptions.class, org.tensorflow.framework.GPUOptions.Builder.class);
}
// Construct using org.tensorflow.framework.GPUOptions.newBuilder()
private Builder() {
maybeForceBuilderInitialization();
}
private Builder(
com.google.protobuf.GeneratedMessageV3.BuilderParent parent) {
super(parent);
maybeForceBuilderInitialization();
}
private void maybeForceBuilderInitialization() {
if (com.google.protobuf.GeneratedMessageV3
.alwaysUseFieldBuilders) {
}
}
public Builder clear() {
super.clear();
perProcessGpuMemoryFraction_ = 0D;
allowGrowth_ = false;
allocatorType_ = "";
deferredDeletionBytes_ = 0L;
visibleDeviceList_ = "";
pollingActiveDelayUsecs_ = 0;
pollingInactiveDelayMsecs_ = 0;
forceGpuCompatible_ = false;
if (experimentalBuilder_ == null) {
experimental_ = null;
} else {
experimental_ = null;
experimentalBuilder_ = null;
}
return this;
}
public com.google.protobuf.Descriptors.Descriptor
getDescriptorForType() {
return org.tensorflow.framework.ConfigProtos.internal_static_tensorflow_GPUOptions_descriptor;
}
public org.tensorflow.framework.GPUOptions getDefaultInstanceForType() {
return org.tensorflow.framework.GPUOptions.getDefaultInstance();
}
public org.tensorflow.framework.GPUOptions build() {
org.tensorflow.framework.GPUOptions result = buildPartial();
if (!result.isInitialized()) {
throw newUninitializedMessageException(result);
}
return result;
}
public org.tensorflow.framework.GPUOptions buildPartial() {
org.tensorflow.framework.GPUOptions result = new org.tensorflow.framework.GPUOptions(this);
result.perProcessGpuMemoryFraction_ = perProcessGpuMemoryFraction_;
result.allowGrowth_ = allowGrowth_;
result.allocatorType_ = allocatorType_;
result.deferredDeletionBytes_ = deferredDeletionBytes_;
result.visibleDeviceList_ = visibleDeviceList_;
result.pollingActiveDelayUsecs_ = pollingActiveDelayUsecs_;
result.pollingInactiveDelayMsecs_ = pollingInactiveDelayMsecs_;
result.forceGpuCompatible_ = forceGpuCompatible_;
if (experimentalBuilder_ == null) {
result.experimental_ = experimental_;
} else {
result.experimental_ = experimentalBuilder_.build();
}
onBuilt();
return result;
}
public Builder clone() {
return (Builder) super.clone();
}
public Builder setField(
com.google.protobuf.Descriptors.FieldDescriptor field,
java.lang.Object value) {
return (Builder) super.setField(field, value);
}
public Builder clearField(
com.google.protobuf.Descriptors.FieldDescriptor field) {
return (Builder) super.clearField(field);
}
public Builder clearOneof(
com.google.protobuf.Descriptors.OneofDescriptor oneof) {
return (Builder) super.clearOneof(oneof);
}
public Builder setRepeatedField(
com.google.protobuf.Descriptors.FieldDescriptor field,
int index, java.lang.Object value) {
return (Builder) super.setRepeatedField(field, index, value);
}
public Builder addRepeatedField(
com.google.protobuf.Descriptors.FieldDescriptor field,
java.lang.Object value) {
return (Builder) super.addRepeatedField(field, value);
}
public Builder mergeFrom(com.google.protobuf.Message other) {
if (other instanceof org.tensorflow.framework.GPUOptions) {
return mergeFrom((org.tensorflow.framework.GPUOptions)other);
} else {
super.mergeFrom(other);
return this;
}
}
public Builder mergeFrom(org.tensorflow.framework.GPUOptions other) {
if (other == org.tensorflow.framework.GPUOptions.getDefaultInstance()) return this;
if (other.getPerProcessGpuMemoryFraction() != 0D) {
setPerProcessGpuMemoryFraction(other.getPerProcessGpuMemoryFraction());
}
if (other.getAllowGrowth() != false) {
setAllowGrowth(other.getAllowGrowth());
}
if (!other.getAllocatorType().isEmpty()) {
allocatorType_ = other.allocatorType_;
onChanged();
}
if (other.getDeferredDeletionBytes() != 0L) {
setDeferredDeletionBytes(other.getDeferredDeletionBytes());
}
if (!other.getVisibleDeviceList().isEmpty()) {
visibleDeviceList_ = other.visibleDeviceList_;
onChanged();
}
if (other.getPollingActiveDelayUsecs() != 0) {
setPollingActiveDelayUsecs(other.getPollingActiveDelayUsecs());
}
if (other.getPollingInactiveDelayMsecs() != 0) {
setPollingInactiveDelayMsecs(other.getPollingInactiveDelayMsecs());
}
if (other.getForceGpuCompatible() != false) {
setForceGpuCompatible(other.getForceGpuCompatible());
}
if (other.hasExperimental()) {
mergeExperimental(other.getExperimental());
}
this.mergeUnknownFields(other.unknownFields);
onChanged();
return this;
}
public final boolean isInitialized() {
return true;
}
public Builder mergeFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws java.io.IOException {
org.tensorflow.framework.GPUOptions parsedMessage = null;
try {
parsedMessage = PARSER.parsePartialFrom(input, extensionRegistry);
} catch (com.google.protobuf.InvalidProtocolBufferException e) {
parsedMessage = (org.tensorflow.framework.GPUOptions) e.getUnfinishedMessage();
throw e.unwrapIOException();
} finally {
if (parsedMessage != null) {
mergeFrom(parsedMessage);
}
}
return this;
}
private double perProcessGpuMemoryFraction_ ;
/**
*
* Fraction of the available GPU memory to allocate for each process.
* 1 means to allocate all of the GPU memory, 0.5 means the process
* allocates up to ~50% of the available GPU memory.
* GPU memory is pre-allocated unless the allow_growth option is enabled.
* If greater than 1.0, uses CUDA unified memory to potentially oversubscribe
* the amount of memory available on the GPU device by using host memory as a
* swap space. Accessing memory not available on the device will be
* significantly slower as that would require memory transfer between the host
* and the device. Options to reduce the memory requirement should be
* considered before enabling this option as this may come with a negative
* performance impact. Oversubscription using the unified memory requires
* Pascal class or newer GPUs and it is currently only supported on the Linux
* operating system. See
* https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#um-requirements
* for the detailed requirements.
*
*
* double per_process_gpu_memory_fraction = 1;
*/
public double getPerProcessGpuMemoryFraction() {
return perProcessGpuMemoryFraction_;
}
/**
*
* Fraction of the available GPU memory to allocate for each process.
* 1 means to allocate all of the GPU memory, 0.5 means the process
* allocates up to ~50% of the available GPU memory.
* GPU memory is pre-allocated unless the allow_growth option is enabled.
* If greater than 1.0, uses CUDA unified memory to potentially oversubscribe
* the amount of memory available on the GPU device by using host memory as a
* swap space. Accessing memory not available on the device will be
* significantly slower as that would require memory transfer between the host
* and the device. Options to reduce the memory requirement should be
* considered before enabling this option as this may come with a negative
* performance impact. Oversubscription using the unified memory requires
* Pascal class or newer GPUs and it is currently only supported on the Linux
* operating system. See
* https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#um-requirements
* for the detailed requirements.
*
*
* double per_process_gpu_memory_fraction = 1;
*/
public Builder setPerProcessGpuMemoryFraction(double value) {
perProcessGpuMemoryFraction_ = value;
onChanged();
return this;
}
/**
*
* Fraction of the available GPU memory to allocate for each process.
* 1 means to allocate all of the GPU memory, 0.5 means the process
* allocates up to ~50% of the available GPU memory.
* GPU memory is pre-allocated unless the allow_growth option is enabled.
* If greater than 1.0, uses CUDA unified memory to potentially oversubscribe
* the amount of memory available on the GPU device by using host memory as a
* swap space. Accessing memory not available on the device will be
* significantly slower as that would require memory transfer between the host
* and the device. Options to reduce the memory requirement should be
* considered before enabling this option as this may come with a negative
* performance impact. Oversubscription using the unified memory requires
* Pascal class or newer GPUs and it is currently only supported on the Linux
* operating system. See
* https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#um-requirements
* for the detailed requirements.
*
*
* double per_process_gpu_memory_fraction = 1;
*/
public Builder clearPerProcessGpuMemoryFraction() {
perProcessGpuMemoryFraction_ = 0D;
onChanged();
return this;
}
private boolean allowGrowth_ ;
/**
*
* If true, the allocator does not pre-allocate the entire specified
* GPU memory region, instead starting small and growing as needed.
*
*
* bool allow_growth = 4;
*/
public boolean getAllowGrowth() {
return allowGrowth_;
}
/**
*
* If true, the allocator does not pre-allocate the entire specified
* GPU memory region, instead starting small and growing as needed.
*
*
* bool allow_growth = 4;
*/
public Builder setAllowGrowth(boolean value) {
allowGrowth_ = value;
onChanged();
return this;
}
/**
*
* If true, the allocator does not pre-allocate the entire specified
* GPU memory region, instead starting small and growing as needed.
*
*
* bool allow_growth = 4;
*/
public Builder clearAllowGrowth() {
allowGrowth_ = false;
onChanged();
return this;
}
private java.lang.Object allocatorType_ = "";
/**
*
* The type of GPU allocation strategy to use.
* Allowed values:
* "": The empty string (default) uses a system-chosen default
* which may change over time.
* "BFC": A "Best-fit with coalescing" algorithm, simplified from a
* version of dlmalloc.
*
*
* string allocator_type = 2;
*/
public java.lang.String getAllocatorType() {
java.lang.Object ref = allocatorType_;
if (!(ref instanceof java.lang.String)) {
com.google.protobuf.ByteString bs =
(com.google.protobuf.ByteString) ref;
java.lang.String s = bs.toStringUtf8();
allocatorType_ = s;
return s;
} else {
return (java.lang.String) ref;
}
}
/**
*
* The type of GPU allocation strategy to use.
* Allowed values:
* "": The empty string (default) uses a system-chosen default
* which may change over time.
* "BFC": A "Best-fit with coalescing" algorithm, simplified from a
* version of dlmalloc.
*
*
* string allocator_type = 2;
*/
public com.google.protobuf.ByteString
getAllocatorTypeBytes() {
java.lang.Object ref = allocatorType_;
if (ref instanceof String) {
com.google.protobuf.ByteString b =
com.google.protobuf.ByteString.copyFromUtf8(
(java.lang.String) ref);
allocatorType_ = b;
return b;
} else {
return (com.google.protobuf.ByteString) ref;
}
}
/**
*
* The type of GPU allocation strategy to use.
* Allowed values:
* "": The empty string (default) uses a system-chosen default
* which may change over time.
* "BFC": A "Best-fit with coalescing" algorithm, simplified from a
* version of dlmalloc.
*
*
* string allocator_type = 2;
*/
public Builder setAllocatorType(
java.lang.String value) {
if (value == null) {
throw new NullPointerException();
}
allocatorType_ = value;
onChanged();
return this;
}
/**
*
* The type of GPU allocation strategy to use.
* Allowed values:
* "": The empty string (default) uses a system-chosen default
* which may change over time.
* "BFC": A "Best-fit with coalescing" algorithm, simplified from a
* version of dlmalloc.
*
*
* string allocator_type = 2;
*/
public Builder clearAllocatorType() {
allocatorType_ = getDefaultInstance().getAllocatorType();
onChanged();
return this;
}
/**
*
* The type of GPU allocation strategy to use.
* Allowed values:
* "": The empty string (default) uses a system-chosen default
* which may change over time.
* "BFC": A "Best-fit with coalescing" algorithm, simplified from a
* version of dlmalloc.
*
*
* string allocator_type = 2;
*/
public Builder setAllocatorTypeBytes(
com.google.protobuf.ByteString value) {
if (value == null) {
throw new NullPointerException();
}
checkByteStringIsUtf8(value);
allocatorType_ = value;
onChanged();
return this;
}
private long deferredDeletionBytes_ ;
/**
*
* Delay deletion of up to this many bytes to reduce the number of
* interactions with gpu driver code. If 0, the system chooses
* a reasonable default (several MBs).
*
*
* int64 deferred_deletion_bytes = 3;
*/
public long getDeferredDeletionBytes() {
return deferredDeletionBytes_;
}
/**
*
* Delay deletion of up to this many bytes to reduce the number of
* interactions with gpu driver code. If 0, the system chooses
* a reasonable default (several MBs).
*
*
* int64 deferred_deletion_bytes = 3;
*/
public Builder setDeferredDeletionBytes(long value) {
deferredDeletionBytes_ = value;
onChanged();
return this;
}
/**
*
* Delay deletion of up to this many bytes to reduce the number of
* interactions with gpu driver code. If 0, the system chooses
* a reasonable default (several MBs).
*
*
* int64 deferred_deletion_bytes = 3;
*/
public Builder clearDeferredDeletionBytes() {
deferredDeletionBytes_ = 0L;
onChanged();
return this;
}
private java.lang.Object visibleDeviceList_ = "";
/**
*
* A comma-separated list of GPU ids that determines the 'visible'
* to 'virtual' mapping of GPU devices. For example, if TensorFlow
* can see 8 GPU devices in the process, and one wanted to map
* visible GPU devices 5 and 3 as "/device:GPU:0", and "/device:GPU:1",
* then one would specify this field as "5,3". This field is similar in
* spirit to the CUDA_VISIBLE_DEVICES environment variable, except
* it applies to the visible GPU devices in the process.
* NOTE:
* 1. The GPU driver provides the process with the visible GPUs
* in an order which is not guaranteed to have any correlation to
* the *physical* GPU id in the machine. This field is used for
* remapping "visible" to "virtual", which means this operates only
* after the process starts. Users are required to use vendor
* specific mechanisms (e.g., CUDA_VISIBLE_DEVICES) to control the
* physical to visible device mapping prior to invoking TensorFlow.
* 2. In the code, the ids in this list are also called "platform GPU id"s,
* and the 'virtual' ids of GPU devices (i.e. the ids in the device
* name "/device:GPU:<id>") are also called "TF GPU id"s. Please
* refer to third_party/tensorflow/core/common_runtime/gpu/gpu_id.h
* for more information.
*
*
* string visible_device_list = 5;
*/
public java.lang.String getVisibleDeviceList() {
java.lang.Object ref = visibleDeviceList_;
if (!(ref instanceof java.lang.String)) {
com.google.protobuf.ByteString bs =
(com.google.protobuf.ByteString) ref;
java.lang.String s = bs.toStringUtf8();
visibleDeviceList_ = s;
return s;
} else {
return (java.lang.String) ref;
}
}
/**
*
* A comma-separated list of GPU ids that determines the 'visible'
* to 'virtual' mapping of GPU devices. For example, if TensorFlow
* can see 8 GPU devices in the process, and one wanted to map
* visible GPU devices 5 and 3 as "/device:GPU:0", and "/device:GPU:1",
* then one would specify this field as "5,3". This field is similar in
* spirit to the CUDA_VISIBLE_DEVICES environment variable, except
* it applies to the visible GPU devices in the process.
* NOTE:
* 1. The GPU driver provides the process with the visible GPUs
* in an order which is not guaranteed to have any correlation to
* the *physical* GPU id in the machine. This field is used for
* remapping "visible" to "virtual", which means this operates only
* after the process starts. Users are required to use vendor
* specific mechanisms (e.g., CUDA_VISIBLE_DEVICES) to control the
* physical to visible device mapping prior to invoking TensorFlow.
* 2. In the code, the ids in this list are also called "platform GPU id"s,
* and the 'virtual' ids of GPU devices (i.e. the ids in the device
* name "/device:GPU:<id>") are also called "TF GPU id"s. Please
* refer to third_party/tensorflow/core/common_runtime/gpu/gpu_id.h
* for more information.
*
*
* string visible_device_list = 5;
*/
public com.google.protobuf.ByteString
getVisibleDeviceListBytes() {
java.lang.Object ref = visibleDeviceList_;
if (ref instanceof String) {
com.google.protobuf.ByteString b =
com.google.protobuf.ByteString.copyFromUtf8(
(java.lang.String) ref);
visibleDeviceList_ = b;
return b;
} else {
return (com.google.protobuf.ByteString) ref;
}
}
/**
*
* A comma-separated list of GPU ids that determines the 'visible'
* to 'virtual' mapping of GPU devices. For example, if TensorFlow
* can see 8 GPU devices in the process, and one wanted to map
* visible GPU devices 5 and 3 as "/device:GPU:0", and "/device:GPU:1",
* then one would specify this field as "5,3". This field is similar in
* spirit to the CUDA_VISIBLE_DEVICES environment variable, except
* it applies to the visible GPU devices in the process.
* NOTE:
* 1. The GPU driver provides the process with the visible GPUs
* in an order which is not guaranteed to have any correlation to
* the *physical* GPU id in the machine. This field is used for
* remapping "visible" to "virtual", which means this operates only
* after the process starts. Users are required to use vendor
* specific mechanisms (e.g., CUDA_VISIBLE_DEVICES) to control the
* physical to visible device mapping prior to invoking TensorFlow.
* 2. In the code, the ids in this list are also called "platform GPU id"s,
* and the 'virtual' ids of GPU devices (i.e. the ids in the device
* name "/device:GPU:<id>") are also called "TF GPU id"s. Please
* refer to third_party/tensorflow/core/common_runtime/gpu/gpu_id.h
* for more information.
*
*
* string visible_device_list = 5;
*/
public Builder setVisibleDeviceList(
java.lang.String value) {
if (value == null) {
throw new NullPointerException();
}
visibleDeviceList_ = value;
onChanged();
return this;
}
/**
*
* A comma-separated list of GPU ids that determines the 'visible'
* to 'virtual' mapping of GPU devices. For example, if TensorFlow
* can see 8 GPU devices in the process, and one wanted to map
* visible GPU devices 5 and 3 as "/device:GPU:0", and "/device:GPU:1",
* then one would specify this field as "5,3". This field is similar in
* spirit to the CUDA_VISIBLE_DEVICES environment variable, except
* it applies to the visible GPU devices in the process.
* NOTE:
* 1. The GPU driver provides the process with the visible GPUs
* in an order which is not guaranteed to have any correlation to
* the *physical* GPU id in the machine. This field is used for
* remapping "visible" to "virtual", which means this operates only
* after the process starts. Users are required to use vendor
* specific mechanisms (e.g., CUDA_VISIBLE_DEVICES) to control the
* physical to visible device mapping prior to invoking TensorFlow.
* 2. In the code, the ids in this list are also called "platform GPU id"s,
* and the 'virtual' ids of GPU devices (i.e. the ids in the device
* name "/device:GPU:<id>") are also called "TF GPU id"s. Please
* refer to third_party/tensorflow/core/common_runtime/gpu/gpu_id.h
* for more information.
*
*
* string visible_device_list = 5;
*/
public Builder clearVisibleDeviceList() {
visibleDeviceList_ = getDefaultInstance().getVisibleDeviceList();
onChanged();
return this;
}
/**
*
* A comma-separated list of GPU ids that determines the 'visible'
* to 'virtual' mapping of GPU devices. For example, if TensorFlow
* can see 8 GPU devices in the process, and one wanted to map
* visible GPU devices 5 and 3 as "/device:GPU:0", and "/device:GPU:1",
* then one would specify this field as "5,3". This field is similar in
* spirit to the CUDA_VISIBLE_DEVICES environment variable, except
* it applies to the visible GPU devices in the process.
* NOTE:
* 1. The GPU driver provides the process with the visible GPUs
* in an order which is not guaranteed to have any correlation to
* the *physical* GPU id in the machine. This field is used for
* remapping "visible" to "virtual", which means this operates only
* after the process starts. Users are required to use vendor
* specific mechanisms (e.g., CUDA_VISIBLE_DEVICES) to control the
* physical to visible device mapping prior to invoking TensorFlow.
* 2. In the code, the ids in this list are also called "platform GPU id"s,
* and the 'virtual' ids of GPU devices (i.e. the ids in the device
* name "/device:GPU:<id>") are also called "TF GPU id"s. Please
* refer to third_party/tensorflow/core/common_runtime/gpu/gpu_id.h
* for more information.
*
*
* string visible_device_list = 5;
*/
public Builder setVisibleDeviceListBytes(
com.google.protobuf.ByteString value) {
if (value == null) {
throw new NullPointerException();
}
checkByteStringIsUtf8(value);
visibleDeviceList_ = value;
onChanged();
return this;
}
private int pollingActiveDelayUsecs_ ;
/**
*
* In the event polling loop sleep this many microseconds between
* PollEvents calls, when the queue is not empty. If value is not
* set or set to 0, gets set to a non-zero default.
*
*
* int32 polling_active_delay_usecs = 6;
*/
public int getPollingActiveDelayUsecs() {
return pollingActiveDelayUsecs_;
}
/**
*
* In the event polling loop sleep this many microseconds between
* PollEvents calls, when the queue is not empty. If value is not
* set or set to 0, gets set to a non-zero default.
*
*
* int32 polling_active_delay_usecs = 6;
*/
public Builder setPollingActiveDelayUsecs(int value) {
pollingActiveDelayUsecs_ = value;
onChanged();
return this;
}
/**
*
* In the event polling loop sleep this many microseconds between
* PollEvents calls, when the queue is not empty. If value is not
* set or set to 0, gets set to a non-zero default.
*
*
* int32 polling_active_delay_usecs = 6;
*/
public Builder clearPollingActiveDelayUsecs() {
pollingActiveDelayUsecs_ = 0;
onChanged();
return this;
}
private int pollingInactiveDelayMsecs_ ;
/**
*
* This field is deprecated and ignored.
*
*
* int32 polling_inactive_delay_msecs = 7;
*/
public int getPollingInactiveDelayMsecs() {
return pollingInactiveDelayMsecs_;
}
/**
*
* This field is deprecated and ignored.
*
*
* int32 polling_inactive_delay_msecs = 7;
*/
public Builder setPollingInactiveDelayMsecs(int value) {
pollingInactiveDelayMsecs_ = value;
onChanged();
return this;
}
/**
*
* This field is deprecated and ignored.
*
*
* int32 polling_inactive_delay_msecs = 7;
*/
public Builder clearPollingInactiveDelayMsecs() {
pollingInactiveDelayMsecs_ = 0;
onChanged();
return this;
}
private boolean forceGpuCompatible_ ;
/**
*
* Force all tensors to be gpu_compatible. On a GPU-enabled TensorFlow,
* enabling this option forces all CPU tensors to be allocated with Cuda
* pinned memory. Normally, TensorFlow will infer which tensors should be
* allocated as the pinned memory. But in case where the inference is
* incomplete, this option can significantly speed up the cross-device memory
* copy performance as long as it fits the memory.
* Note that this option is not something that should be
* enabled by default for unknown or very large models, since all Cuda pinned
* memory is unpageable, having too much pinned memory might negatively impact
* the overall host system performance.
*
*
* bool force_gpu_compatible = 8;
*/
public boolean getForceGpuCompatible() {
return forceGpuCompatible_;
}
/**
*
* Force all tensors to be gpu_compatible. On a GPU-enabled TensorFlow,
* enabling this option forces all CPU tensors to be allocated with Cuda
* pinned memory. Normally, TensorFlow will infer which tensors should be
* allocated as the pinned memory. But in case where the inference is
* incomplete, this option can significantly speed up the cross-device memory
* copy performance as long as it fits the memory.
* Note that this option is not something that should be
* enabled by default for unknown or very large models, since all Cuda pinned
* memory is unpageable, having too much pinned memory might negatively impact
* the overall host system performance.
*
*
* bool force_gpu_compatible = 8;
*/
public Builder setForceGpuCompatible(boolean value) {
forceGpuCompatible_ = value;
onChanged();
return this;
}
/**
*
* Force all tensors to be gpu_compatible. On a GPU-enabled TensorFlow,
* enabling this option forces all CPU tensors to be allocated with Cuda
* pinned memory. Normally, TensorFlow will infer which tensors should be
* allocated as the pinned memory. But in case where the inference is
* incomplete, this option can significantly speed up the cross-device memory
* copy performance as long as it fits the memory.
* Note that this option is not something that should be
* enabled by default for unknown or very large models, since all Cuda pinned
* memory is unpageable, having too much pinned memory might negatively impact
* the overall host system performance.
*
*
* bool force_gpu_compatible = 8;
*/
public Builder clearForceGpuCompatible() {
forceGpuCompatible_ = false;
onChanged();
return this;
}
private org.tensorflow.framework.GPUOptions.Experimental experimental_ = null;
private com.google.protobuf.SingleFieldBuilderV3<
org.tensorflow.framework.GPUOptions.Experimental, org.tensorflow.framework.GPUOptions.Experimental.Builder, org.tensorflow.framework.GPUOptions.ExperimentalOrBuilder> experimentalBuilder_;
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public boolean hasExperimental() {
return experimentalBuilder_ != null || experimental_ != null;
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public org.tensorflow.framework.GPUOptions.Experimental getExperimental() {
if (experimentalBuilder_ == null) {
return experimental_ == null ? org.tensorflow.framework.GPUOptions.Experimental.getDefaultInstance() : experimental_;
} else {
return experimentalBuilder_.getMessage();
}
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public Builder setExperimental(org.tensorflow.framework.GPUOptions.Experimental value) {
if (experimentalBuilder_ == null) {
if (value == null) {
throw new NullPointerException();
}
experimental_ = value;
onChanged();
} else {
experimentalBuilder_.setMessage(value);
}
return this;
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public Builder setExperimental(
org.tensorflow.framework.GPUOptions.Experimental.Builder builderForValue) {
if (experimentalBuilder_ == null) {
experimental_ = builderForValue.build();
onChanged();
} else {
experimentalBuilder_.setMessage(builderForValue.build());
}
return this;
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public Builder mergeExperimental(org.tensorflow.framework.GPUOptions.Experimental value) {
if (experimentalBuilder_ == null) {
if (experimental_ != null) {
experimental_ =
org.tensorflow.framework.GPUOptions.Experimental.newBuilder(experimental_).mergeFrom(value).buildPartial();
} else {
experimental_ = value;
}
onChanged();
} else {
experimentalBuilder_.mergeFrom(value);
}
return this;
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public Builder clearExperimental() {
if (experimentalBuilder_ == null) {
experimental_ = null;
onChanged();
} else {
experimental_ = null;
experimentalBuilder_ = null;
}
return this;
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public org.tensorflow.framework.GPUOptions.Experimental.Builder getExperimentalBuilder() {
onChanged();
return getExperimentalFieldBuilder().getBuilder();
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
public org.tensorflow.framework.GPUOptions.ExperimentalOrBuilder getExperimentalOrBuilder() {
if (experimentalBuilder_ != null) {
return experimentalBuilder_.getMessageOrBuilder();
} else {
return experimental_ == null ?
org.tensorflow.framework.GPUOptions.Experimental.getDefaultInstance() : experimental_;
}
}
/**
*
* Everything inside experimental is subject to change and is not subject
* to API stability guarantees in
* https://www.tensorflow.org/guide/version_compat.
*
*
* .tensorflow.GPUOptions.Experimental experimental = 9;
*/
private com.google.protobuf.SingleFieldBuilderV3<
org.tensorflow.framework.GPUOptions.Experimental, org.tensorflow.framework.GPUOptions.Experimental.Builder, org.tensorflow.framework.GPUOptions.ExperimentalOrBuilder>
getExperimentalFieldBuilder() {
if (experimentalBuilder_ == null) {
experimentalBuilder_ = new com.google.protobuf.SingleFieldBuilderV3<
org.tensorflow.framework.GPUOptions.Experimental, org.tensorflow.framework.GPUOptions.Experimental.Builder, org.tensorflow.framework.GPUOptions.ExperimentalOrBuilder>(
getExperimental(),
getParentForChildren(),
isClean());
experimental_ = null;
}
return experimentalBuilder_;
}
public final Builder setUnknownFields(
final com.google.protobuf.UnknownFieldSet unknownFields) {
return super.setUnknownFieldsProto3(unknownFields);
}
public final Builder mergeUnknownFields(
final com.google.protobuf.UnknownFieldSet unknownFields) {
return super.mergeUnknownFields(unknownFields);
}
// @@protoc_insertion_point(builder_scope:tensorflow.GPUOptions)
}
// @@protoc_insertion_point(class_scope:tensorflow.GPUOptions)
private static final org.tensorflow.framework.GPUOptions DEFAULT_INSTANCE;
static {
DEFAULT_INSTANCE = new org.tensorflow.framework.GPUOptions();
}
public static org.tensorflow.framework.GPUOptions getDefaultInstance() {
return DEFAULT_INSTANCE;
}
private static final com.google.protobuf.Parser
PARSER = new com.google.protobuf.AbstractParser() {
public GPUOptions parsePartialFrom(
com.google.protobuf.CodedInputStream input,
com.google.protobuf.ExtensionRegistryLite extensionRegistry)
throws com.google.protobuf.InvalidProtocolBufferException {
return new GPUOptions(input, extensionRegistry);
}
};
public static com.google.protobuf.Parser parser() {
return PARSER;
}
@java.lang.Override
public com.google.protobuf.Parser getParserForType() {
return PARSER;
}
public org.tensorflow.framework.GPUOptions getDefaultInstanceForType() {
return DEFAULT_INSTANCE;
}
}