xprof-nightly 2.22.3a20251208__cp311-none-manylinux2014_x86_64.whl

xprof/profile_plugin_loader.py

@@ -0,0 +1,82 @@
+# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Loader for `profile_plugin`."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import argparse
+import logging
+
+from xprof.standalone.tensorboard_shim import base_plugin
+logger = logging.getLogger('tensorboard-plugin-profile')
+
+
+class ProfilePluginLoader(base_plugin.TBLoader):
+  """ProfilePlugin factory."""
+
+  def define_flags(self, parser):
+    group = parser.add_argument_group('profile plugin')
+    try:
+      group.add_argument(
+          '--master_tpu_unsecure_channel',
+          metavar='ADDR',
+          type=str,
+          default='',
+          help="""\
+  IP address of "master tpu", used for getting streaming trace data
+  through tpu profiler analysis grpc. The grpc channel is not secured.\
+  """,
+      )
+    except argparse.ArgumentError:
+      # This same flag is registered by TensorBoard's static profile
+      # plugin, as long as it continues to be bundled. Nothing to do.
+      pass
+
+  def load(self, context):
+    """Returns the plugin, if possible.
+
+    Args:
+      context: The TBContext flags.
+
+    Returns:
+      A ProfilePlugin instance or None if it couldn't be loaded.
+    """
+    # Ensure that we have TensorFlow and the `profiler_client`, which
+    # was added in TensorFlow 1.14.
+    try:
+      # pylint: disable=g-import-not-at-top
+      # pylint: disable=g-direct-tensorflow-import
+      import tensorflow
+      from tensorflow.python.profiler import profiler_client
+      # pylint: enable=g-import-not-at-top
+      # pylint: enable=g-direct-tensorflow-import
+      del tensorflow
+      del profiler_client
+    except ImportError:
+      try:
+        from xprof.convert import _pywrap_profiler_plugin  # pylint: disable=g-import-not-at-top
+        del _pywrap_profiler_plugin
+        logger.info('Loading profiler plugin with limited functionality')
+      except ImportError as err:
+        logger.warning('Unable to load profiler plugin. Import error: %s', err)
+        return None
+
+    # pylint: disable=g-import-not-at-top
+    from xprof import profile_plugin
+    # pylint: enable=g-import-not-at-top
+
+    return profile_plugin.ProfilePlugin(context)

xprof/protobuf/dcn_collective_info_pb2.py

@@ -0,0 +1,44 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/dcn_collective_info.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/dcn_collective_info.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n/plugin/xprof/protobuf/dcn_collective_info.proto\x12\x13tensorflow.profiler\"\xd8\x05\n\x16\x44\x63nCollectiveInfoProto\x12O\n\rtransfer_type\x18\x01 \x01(\x0e\x32\x38.tensorflow.profiler.DcnCollectiveInfoProto.TransferType\x12R\n\x0f\x65ndpoint_groups\x18\x02 \x03(\x0b\x32\x39.tensorflow.profiler.DcnCollectiveInfoProto.EndpointGroup\x12T\n\x11one_to_one_groups\x18\x03 \x03(\x0b\x32\x39.tensorflow.profiler.DcnCollectiveInfoProto.OneToOneGroup\x1a/\n\x08\x45ndpoint\x12\x10\n\x08slice_id\x18\x01 \x01(\x05\x12\x11\n\tdevice_id\x18\x02 \x01(\x05\x1aX\n\rEndpointGroup\x12G\n\tendpoints\x18\x01 \x03(\x0b\x32\x34.tensorflow.profiler.DcnCollectiveInfoProto.Endpoint\x1a\xa0\x01\n\rOneToOneGroup\x12\x44\n\x06source\x18\x01 \x01(\x0b\x32\x34.tensorflow.profiler.DcnCollectiveInfoProto.Endpoint\x12I\n\x0b\x64\x65stination\x18\x02 \x01(\x0b\x32\x34.tensorflow.profiler.DcnCollectiveInfoProto.Endpoint\"\x94\x01\n\x0cTransferType\x12\x19\n\x15UNKNOWN_TRANSFER_TYPE\x10\x00\x12\x0e\n\nALL_TO_ALL\x10\x01\x12\x0e\n\nONE_TO_ONE\x10\x02\x12\x12\n\x0eREDUCE_SCATTER\x10\x03\x12\x0e\n\nALL_GATHER\x10\x04\x12\x0e\n\nALL_REDUCE\x10\x05\x12\x15\n\x11RAGGED_ALL_TO_ALL\x10\x06\x62\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.dcn_collective_info_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_DCNCOLLECTIVEINFOPROTO']._serialized_start=73
+  _globals['_DCNCOLLECTIVEINFOPROTO']._serialized_end=801
+  _globals['_DCNCOLLECTIVEINFOPROTO_ENDPOINT']._serialized_start=350
+  _globals['_DCNCOLLECTIVEINFOPROTO_ENDPOINT']._serialized_end=397
+  _globals['_DCNCOLLECTIVEINFOPROTO_ENDPOINTGROUP']._serialized_start=399
+  _globals['_DCNCOLLECTIVEINFOPROTO_ENDPOINTGROUP']._serialized_end=487
+  _globals['_DCNCOLLECTIVEINFOPROTO_ONETOONEGROUP']._serialized_start=490
+  _globals['_DCNCOLLECTIVEINFOPROTO_ONETOONEGROUP']._serialized_end=650
+  _globals['_DCNCOLLECTIVEINFOPROTO_TRANSFERTYPE']._serialized_start=653
+  _globals['_DCNCOLLECTIVEINFOPROTO_TRANSFERTYPE']._serialized_end=801
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/dcn_slack_analysis_pb2.py

@@ -0,0 +1,42 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/dcn_slack_analysis.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/dcn_slack_analysis.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n.plugin/xprof/protobuf/dcn_slack_analysis.proto\x12\x13tensorflow.profiler\"8\n\nOpInstance\x12\x15\n\rstart_time_ps\x18\x01 \x01(\x04\x12\x13\n\x0b\x64uration_ps\x18\x02 \x01(\x04\"\xf6\x03\n\x08\x44\x63nSlack\x12\x12\n\nrendezvous\x18\x01 \x01(\t\x12\x1a\n\x12send_start_time_us\x18\x02 \x01(\x04\x12\x1d\n\x15recv_done_end_time_us\x18\x03 \x01(\x04\x12\x10\n\x08slack_us\x18\x04 \x01(\x04\x12&\n\x1e\x62ytes_transmitted_over_network\x18\x05 \x01(\x04\x12\x19\n\x11stall_duration_us\x18\x06 \x01(\x04\x12\x14\n\x0crecv_op_name\x18\x07 \x01(\t\x12\x14\n\x0csend_op_name\x18\x08 \x01(\t\x12-\n\x04send\x18\t \x01(\x0b\x32\x1f.tensorflow.profiler.OpInstance\x12\x32\n\tsend_done\x18\n \x01(\x0b\x32\x1f.tensorflow.profiler.OpInstance\x12-\n\x04recv\x18\x0b \x01(\x0b\x32\x1f.tensorflow.profiler.OpInstance\x12\x32\n\trecv_done\x18\x0c \x01(\x0b\x32\x1f.tensorflow.profiler.OpInstance\x12\x15\n\rtransfer_type\x18\r \x01(\t\x12=\n\x14host_graph_execution\x18\x0e \x01(\x0b\x32\x1f.tensorflow.profiler.OpInstance\"\x94\x03\n\x0f\x44\x63nSlackSummary\x12\x12\n\nrendezvous\x18\x01 \x01(\t\x12\x10\n\x08slack_us\x18\x02 \x01(\x04\x12\x13\n\x0boccurrences\x18\x03 \x01(\x04\x12&\n\x1e\x62ytes_transmitted_over_network\x18\x04 \x01(\x04\x12\x19\n\x11stall_duration_us\x18\x05 \x01(\x04\x12\x1c\n\x14observed_duration_us\x18\x06 \x01(\x04\x12\x14\n\x0crecv_op_name\x18\x07 \x01(\t\x12\x14\n\x0csend_op_name\x18\x08 \x01(\t\x12\x18\n\x10send_duration_us\x18\t \x01(\x04\x12\x18\n\x10recv_duration_us\x18\n \x01(\x04\x12\x1d\n\x15send_done_duration_us\x18\x0b \x01(\x04\x12\x1d\n\x15recv_done_duration_us\x18\x0c \x01(\x04\x12\x15\n\rtransfer_type\x18\r \x01(\t\x12\x15\n\rhost_stall_us\x18\x0e \x01(\x03\x12\x19\n\x11host_events_count\x18\x0f \x01(\x04\"\x85\x01\n\x10\x44\x63nSlackAnalysis\x12\x30\n\tdcn_slack\x18\x01 \x03(\x0b\x32\x1d.tensorflow.profiler.DcnSlack\x12?\n\x11\x64\x63n_slack_summary\x18\x02 \x03(\x0b\x32$.tensorflow.profiler.DcnSlackSummaryb\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.dcn_slack_analysis_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_OPINSTANCE']._serialized_start=71
+  _globals['_OPINSTANCE']._serialized_end=127
+  _globals['_DCNSLACK']._serialized_start=130
+  _globals['_DCNSLACK']._serialized_end=632
+  _globals['_DCNSLACKSUMMARY']._serialized_start=635
+  _globals['_DCNSLACKSUMMARY']._serialized_end=1039
+  _globals['_DCNSLACKANALYSIS']._serialized_start=1042
+  _globals['_DCNSLACKANALYSIS']._serialized_end=1175
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/diagnostics_pb2.py

@@ -0,0 +1,36 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/diagnostics.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/diagnostics.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\'plugin/xprof/protobuf/diagnostics.proto\x12\x13tensorflow.profiler\"=\n\x0b\x44iagnostics\x12\x0c\n\x04info\x18\x01 \x03(\t\x12\x10\n\x08warnings\x18\x02 \x03(\t\x12\x0e\n\x06\x65rrors\x18\x03 \x03(\tb\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.diagnostics_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_DIAGNOSTICS']._serialized_start=64
+  _globals['_DIAGNOSTICS']._serialized_end=125
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/event_time_fraction_analyzer_pb2.py

@@ -0,0 +1,42 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/event_time_fraction_analyzer.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/event_time_fraction_analyzer.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n8plugin/xprof/protobuf/event_time_fraction_analyzer.proto\x12\x13tensorflow.profiler\"D\n\x18\x45ventTimeFractionPerChip\x12\n\n\x02id\x18\x01 \x01(\t\x12\x1c\n\x14\x65vent_time_fractions\x18\x02 \x03(\x02\"\x84\x02\n\x1f\x45ventTimeFractionAnalyzerResult\x12s\n\x19\x63hip_event_time_fractions\x18\x01 \x03(\x0b\x32P.tensorflow.profiler.EventTimeFractionAnalyzerResult.ChipEventTimeFractionsEntry\x1al\n\x1b\x43hipEventTimeFractionsEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12<\n\x05value\x18\x02 \x01(\x0b\x32-.tensorflow.profiler.EventTimeFractionPerChip:\x02\x38\x01\x62\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.event_time_fraction_analyzer_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_EVENTTIMEFRACTIONANALYZERRESULT_CHIPEVENTTIMEFRACTIONSENTRY']._loaded_options = None
+  _globals['_EVENTTIMEFRACTIONANALYZERRESULT_CHIPEVENTTIMEFRACTIONSENTRY']._serialized_options = b'8\001'
+  _globals['_EVENTTIMEFRACTIONPERCHIP']._serialized_start=81
+  _globals['_EVENTTIMEFRACTIONPERCHIP']._serialized_end=149
+  _globals['_EVENTTIMEFRACTIONANALYZERRESULT']._serialized_start=152
+  _globals['_EVENTTIMEFRACTIONANALYZERRESULT']._serialized_end=412
+  _globals['_EVENTTIMEFRACTIONANALYZERRESULT_CHIPEVENTTIMEFRACTIONSENTRY']._serialized_start=304
+  _globals['_EVENTTIMEFRACTIONANALYZERRESULT_CHIPEVENTTIMEFRACTIONSENTRY']._serialized_end=412
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/hardware_types_pb2.py

@@ -0,0 +1,40 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/hardware_types.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/hardware_types.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n*plugin/xprof/protobuf/hardware_types.proto\x12\x13tensorflow.profiler\"4\n\x14GPUComputeCapability\x12\r\n\x05major\x18\x01 \x01(\r\x12\r\n\x05minor\x18\x02 \x01(\r\"\xd8\x01\n\x12\x44\x65viceCapabilities\x12\x19\n\x11\x63lock_rate_in_ghz\x18\x01 \x01(\x01\x12\x11\n\tnum_cores\x18\x02 \x01(\r\x12\x1c\n\x14memory_size_in_bytes\x18\x03 \x01(\x04\x12\x18\n\x10memory_bandwidth\x18\x04 \x01(\x04\x12\x45\n\x12\x63ompute_capability\x18\x05 \x01(\x0b\x32).tensorflow.profiler.GPUComputeCapability\x12\x15\n\rdevice_vendor\x18\x06 \x01(\t*D\n\x0cHardwareType\x12\x14\n\x10UNKNOWN_HARDWARE\x10\x00\x12\x0c\n\x08\x43PU_ONLY\x10\x01\x12\x07\n\x03GPU\x10\x02\x12\x07\n\x03TPU\x10\x03\x62\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.hardware_types_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_HARDWARETYPE']._serialized_start=340
+  _globals['_HARDWARETYPE']._serialized_end=408
+  _globals['_GPUCOMPUTECAPABILITY']._serialized_start=67
+  _globals['_GPUCOMPUTECAPABILITY']._serialized_end=119
+  _globals['_DEVICECAPABILITIES']._serialized_start=122
+  _globals['_DEVICECAPABILITIES']._serialized_end=338
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/hlo_stats_pb2.py

@@ -0,0 +1,39 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/hlo_stats.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/hlo_stats.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+from xprof.protobuf import source_info_pb2 as plugin_dot_xprof_dot_protobuf_dot_source__info__pb2
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n%plugin/xprof/protobuf/hlo_stats.proto\x12\x1dtensorflow.profiler.hlo_stats\x1a\'plugin/xprof/protobuf/source_info.proto\"[\n\x10HloStatsDatabase\x12G\n\x10hlo_stats_record\x18\x01 \x03(\x0b\x32-.tensorflow.profiler.hlo_stats.HloStatsRecord\"\x87\x08\n\x0eHloStatsRecord\x12\x0c\n\x04rank\x18\x01 \x01(\x04\x12\x12\n\nprogram_id\x18\x1e \x01(\x04\x12\x14\n\x0chlo_category\x18\x11 \x01(\t\x12\x16\n\x0ehlo_expression\x18\x02 \x01(\t\x12\x12\n\ntf_op_name\x18\x15 \x01(\t\x12\x13\n\x0boccurrences\x18\x03 \x01(\x03\x12\x18\n\x10total_time_in_us\x18\x04 \x01(\x01\x12\x16\n\x0e\x61vg_time_in_us\x18\x05 \x01(\x01\x12\x1d\n\x15total_self_time_in_us\x18\x06 \x01(\x01\x12\x1b\n\x13\x61vg_self_time_in_us\x18\x07 \x01(\x01\x12#\n\x1btotal_self_time_as_fraction\x18\x08 \x01(\x01\x12.\n&cumulative_total_self_time_as_fraction\x18\t \x01(\x01\x12\x1a\n\x12\x64ma_stall_fraction\x18\n \x01(\x01\x12\x1a\n\x12measured_flop_rate\x18\r \x01(\x01\x12\x17\n\x0fmodel_flop_rate\x18\" \x01(\x01\x12\x1a\n\x12measured_memory_bw\x18\x0e \x01(\x01\x12\x0e\n\x06hbm_bw\x18\x16 \x01(\x01\x12\x14\n\x0c\x63mem_read_bw\x18\x17 \x01(\x01\x12\x15\n\rcmem_write_bw\x18\x18 \x01(\x01\x12\x14\n\x0cvmem_read_bw\x18# \x01(\x01\x12\x15\n\rvmem_write_bw\x18$ \x01(\x01\x12\x1d\n\x15operational_intensity\x18\x0f \x01(\x01\x12!\n\x19hbm_operational_intensity\x18\x1a \x01(\x01\x12\'\n\x1f\x63mem_read_operational_intensity\x18\x1b \x01(\x01\x12(\n cmem_write_operational_intensity\x18\x1c \x01(\x01\x12\'\n\x1fvmem_read_operational_intensity\x18% \x01(\x01\x12(\n vmem_write_operational_intensity\x18& \x01(\x01\x12(\n bottleneck_operational_intensity\x18\x1d \x01(\x01\x12\x10\n\x08\x62ound_by\x18\x10 \x01(\t\x12\x19\n\x11rematerialization\x18\x14 \x01(\x08\x12\x1b\n\x13outside_compilation\x18\x19 \x01(\x08\x12\x11\n\tautotuned\x18\x1f \x01(\x08\x12\r\n\x05\x66lops\x18  \x01(\x04\x12\x16\n\x0e\x62ytes_accessed\x18! \x01(\x04\x12\x34\n\x0bsource_info\x18\' \x01(\x0b\x32\x1f.tensorflow.profiler.SourceInfoJ\x04\x08\x0b\x10\x0cJ\x04\x08\x0c\x10\rJ\x04\x08\x12\x10\x13J\x04\x08\x13\x10\x14')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.hlo_stats_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_HLOSTATSDATABASE']._serialized_start=113
+  _globals['_HLOSTATSDATABASE']._serialized_end=204
+  _globals['_HLOSTATSRECORD']._serialized_start=207
+  _globals['_HLOSTATSRECORD']._serialized_end=1238
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/inference_stats_pb2.py

@@ -0,0 +1,86 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/inference_stats.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/inference_stats.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n+plugin/xprof/protobuf/inference_stats.proto\x12\x13tensorflow.profiler\"\xd9\x01\n\x11TensorEventDetail\x12\x1c\n\x14tensor_pattern_index\x18\x01 \x01(\x05\x12\x46\n\x05owner\x18\x02 \x01(\x0e\x32\x37.tensorflow.profiler.TensorEventDetail.TensorEventOwner\x12%\n\x1dlinearize_delinearize_time_ps\x18\x03 \x01(\x04\"7\n\x10TensorEventOwner\x12\x0b\n\x07UNKNOWN\x10\x00\x12\x0b\n\x07REQUEST\x10\x01\x12\t\n\x05\x42\x41TCH\x10\x02\"\xc5\x04\n\rRequestDetail\x12\x16\n\nrequest_id\x18\n \x01(\x03:\x02-1\x12\x1a\n\x0emodel_id_index\x18\x08 \x01(\x05:\x02-1\x12\x18\n\rstart_time_ps\x18\x01 \x01(\x04:\x01\x30\x12\x16\n\x0b\x65nd_time_ps\x18\x02 \x01(\x04:\x01\x30\x12\x19\n\x0e\x64\x65vice_time_ps\x18\x07 \x01(\x04:\x01\x30\x12\"\n\x17write_to_device_time_ps\x18\x05 \x01(\x04:\x01\x30\x12#\n\x18read_from_device_time_ps\x18\x06 \x01(\x04:\x01\x30\x12$\n\x19\x62\x61tching_request_delay_ps\x18\x0c \x01(\x04:\x01\x30\x12\x19\n\x11related_batch_ids\x18\x0b \x03(\x03\x12\x1d\n\x15\x62\x61tching_request_size\x18\r \x01(\x05\x12\x1d\n\x15host_preprocessing_ps\x18\x0e \x01(\x04\x12\x1f\n\x17host_batch_formation_ps\x18\x0f \x01(\x04\x12\x17\n\x0fhost_runtime_ps\x18\x10 \x01(\x04\x12\x1e\n\x16host_postprocessing_ps\x18\x11 \x01(\x04\x12\x44\n\x14tensor_event_details\x18\x12 \x03(\x0b\x32&.tensorflow.profiler.TensorEventDetail\x12\x0f\n\x07host_id\x18\x13 \x01(\x05\x12\x12\n\npercentile\x18\x14 \x01(\x01\x12\x14\n\x0cidle_time_ps\x18\x15 \x01(\x01J\x04\x08\x03\x10\x04J\x04\x08\x04\x10\x05J\x04\x08\t\x10\n\"\xe4\x02\n\x0b\x42\x61tchDetail\x12\x14\n\x08\x62\x61tch_id\x18\x01 \x01(\x03:\x02-1\x12\x18\n\rstart_time_ps\x18\x02 \x01(\x04:\x01\x30\x12\x16\n\x0b\x65nd_time_ps\x18\x03 \x01(\x04:\x01\x30\x12\x19\n\x0e\x62\x61tch_delay_ps\x18\x05 \x01(\x04:\x01\x30\x12\x1b\n\x13related_request_ids\x18\x04 \x03(\x03\x12\x16\n\x0epadding_amount\x18\x06 \x01(\x05\x12 \n\x18\x62\x61tch_size_after_padding\x18\x07 \x01(\x05\x12\x16\n\x0emodel_id_index\x18\x08 \x01(\x05\x12\x43\n\x13tensor_event_detail\x18\t \x01(\x0b\x32&.tensorflow.profiler.TensorEventDetail\x12\x0f\n\x07host_id\x18\n \x01(\x05\x12\x12\n\npercentile\x18\x0b \x01(\x01\x12\x19\n\x0e\x64\x65vice_time_ps\x18\x0c \x01(\x04:\x01\x30\"\xa5\x01\n\x15PerHostInferenceStats\x12;\n\x0frequest_details\x18\x03 \x03(\x0b\x32\".tensorflow.profiler.RequestDetail\x12\x37\n\rbatch_details\x18\x05 \x03(\x0b\x32 .tensorflow.profiler.BatchDetailJ\x04\x08\x01\x10\x02J\x04\x08\x02\x10\x03J\x04\x08\x04\x10\x05J\x04\x08\x06\x10\x07\"\x8d\x03\n\x1eTensorTransferAggregatedResult\x12g\n\x16tensor_pattern_results\x18\x01 \x03(\x0b\x32G.tensorflow.profiler.TensorTransferAggregatedResult.TensorPatternResult\x1a\x81\x02\n\x13TensorPatternResult\x12\x1c\n\x14tensor_pattern_index\x18\x01 \x01(\x05\x12\r\n\x05\x63ount\x18\x02 \x01(\x04\x12\x85\x01\n%linearize_delinearize_percentile_time\x18\x03 \x03(\x0b\x32V.tensorflow.profiler.TensorTransferAggregatedResult.TensorPatternResult.PercentileTime\x1a\x35\n\x0ePercentileTime\x12\x12\n\npercentile\x18\x01 \x01(\x01\x12\x0f\n\x07time_ps\x18\x02 \x01(\x04\"\xf2\x01\n\x1cPerBatchSizeAggregatedResult\x12\x12\n\nbatch_size\x18\x01 \x01(\x05\x12\x45\n\x19\x61ggregated_request_result\x18\x02 \x01(\x0b\x32\".tensorflow.profiler.RequestDetail\x12\x41\n\x17\x61ggregated_batch_result\x18\x03 \x01(\x0b\x32 .tensorflow.profiler.BatchDetail\x12\x1a\n\x12request_throughput\x18\x04 \x01(\x01\x12\x18\n\x10\x62\x61tch_throughput\x18\x05 \x01(\x01\"\xd1\x04\n\x16PerModelInferenceStats\x12;\n\x0frequest_details\x18\x01 \x03(\x0b\x32\".tensorflow.profiler.RequestDetail\x12\x45\n\x19\x61ggregated_request_detail\x18\x08 \x01(\x0b\x32\".tensorflow.profiler.RequestDetail\x12\x1a\n\x12request_throughput\x18\x02 \x01(\x01\x12\"\n\x1arequest_average_latency_us\x18\x03 \x01(\x01\x12\x37\n\rbatch_details\x18\x04 \x03(\x0b\x32 .tensorflow.profiler.BatchDetail\x12\x41\n\x17\x61ggregated_batch_detail\x18\t \x01(\x0b\x32 .tensorflow.profiler.BatchDetail\x12\x18\n\x10\x62\x61tch_throughput\x18\x05 \x01(\x01\x12 \n\x18\x62\x61tch_average_latency_us\x18\x06 \x01(\x01\x12^\n!tensor_transfer_aggregated_result\x18\x07 \x01(\x0b\x32\x33.tensorflow.profiler.TensorTransferAggregatedResult\x12[\n per_batch_size_aggregated_result\x18\n \x03(\x0b\x32\x31.tensorflow.profiler.PerBatchSizeAggregatedResult\"\xa0\x01\n\x12\x42\x61tchingParameters\x12\x19\n\x11num_batch_threads\x18\x01 \x01(\x03\x12\x1c\n\x14\x62\x61tch_timeout_micros\x18\x02 \x01(\x03\x12\x16\n\x0emax_batch_size\x18\x03 \x01(\x03\x12\x1c\n\x14max_enqueued_batches\x18\x04 \x01(\x03\x12\x1b\n\x13\x61llowed_batch_sizes\x18\x05 \x01(\t\"\xdb\x02\n\x0fModelIdDatabase\x12\x0b\n\x03ids\x18\x01 \x03(\t\x12H\n\x0bid_to_index\x18\x02 \x03(\x0b\x32\x33.tensorflow.profiler.ModelIdDatabase.IdToIndexEntry\x12[\n\x15id_to_batching_params\x18\x03 \x03(\x0b\x32<.tensorflow.profiler.ModelIdDatabase.IdToBatchingParamsEntry\x1a\x30\n\x0eIdToIndexEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\x05:\x02\x38\x01\x1a\x62\n\x17IdToBatchingParamsEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\x36\n\x05value\x18\x02 \x01(\x0b\x32\'.tensorflow.profiler.BatchingParameters:\x02\x38\x01\"/\n\x15TensorPatternDatabase\x12\x16\n\x0etensor_pattern\x18\x01 \x03(\t\"\x8c\x05\n\x0eInferenceStats\x12`\n\x18inference_stats_per_host\x18\x03 \x03(\x0b\x32>.tensorflow.profiler.InferenceStats.InferenceStatsPerHostEntry\x12\x62\n\x19inference_stats_per_model\x18\x05 \x03(\x0b\x32?.tensorflow.profiler.InferenceStats.InferenceStatsPerModelEntry\x12\x39\n\x0bmodel_id_db\x18\x04 \x01(\x0b\x32$.tensorflow.profiler.ModelIdDatabase\x12\x45\n\x11tensor_pattern_db\x18\x06 \x01(\x0b\x32*.tensorflow.profiler.TensorPatternDatabase\x12P\n\x17sampled_inference_stats\x18\x07 \x01(\x0b\x32/.tensorflow.profiler.SampledInferenceStatsProto\x1ah\n\x1aInferenceStatsPerHostEntry\x12\x0b\n\x03key\x18\x01 \x01(\x05\x12\x39\n\x05value\x18\x02 \x01(\x0b\x32*.tensorflow.profiler.PerHostInferenceStats:\x02\x38\x01\x1aj\n\x1bInferenceStatsPerModelEntry\x12\x0b\n\x03key\x18\x01 \x01(\x05\x12:\n\x05value\x18\x02 \x01(\x0b\x32+.tensorflow.profiler.PerModelInferenceStats:\x02\x38\x01J\x04\x08\x01\x10\x02J\x04\x08\x02\x10\x03\"\x9d\x01\n\"SampledPerModelInferenceStatsProto\x12<\n\x10sampled_requests\x18\x01 \x03(\x0b\x32\".tensorflow.profiler.RequestDetail\x12\x39\n\x0fsampled_batches\x18\x02 \x03(\x0b\x32 .tensorflow.profiler.BatchDetail\"\x9a\x02\n\x1aSampledInferenceStatsProto\x12}\n!sampled_inference_stats_per_model\x18\x01 \x03(\x0b\x32R.tensorflow.profiler.SampledInferenceStatsProto.SampledInferenceStatsPerModelEntry\x1a}\n\"SampledInferenceStatsPerModelEntry\x12\x0b\n\x03key\x18\x01 \x01(\x05\x12\x46\n\x05value\x18\x02 \x01(\x0b\x32\x37.tensorflow.profiler.SampledPerModelInferenceStatsProto:\x02\x38\x01')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.inference_stats_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_MODELIDDATABASE_IDTOINDEXENTRY']._loaded_options = None
+  _globals['_MODELIDDATABASE_IDTOINDEXENTRY']._serialized_options = b'8\001'
+  _globals['_MODELIDDATABASE_IDTOBATCHINGPARAMSENTRY']._loaded_options = None
+  _globals['_MODELIDDATABASE_IDTOBATCHINGPARAMSENTRY']._serialized_options = b'8\001'
+  _globals['_INFERENCESTATS_INFERENCESTATSPERHOSTENTRY']._loaded_options = None
+  _globals['_INFERENCESTATS_INFERENCESTATSPERHOSTENTRY']._serialized_options = b'8\001'
+  _globals['_INFERENCESTATS_INFERENCESTATSPERMODELENTRY']._loaded_options = None
+  _globals['_INFERENCESTATS_INFERENCESTATSPERMODELENTRY']._serialized_options = b'8\001'
+  _globals['_SAMPLEDINFERENCESTATSPROTO_SAMPLEDINFERENCESTATSPERMODELENTRY']._loaded_options = None
+  _globals['_SAMPLEDINFERENCESTATSPROTO_SAMPLEDINFERENCESTATSPERMODELENTRY']._serialized_options = b'8\001'
+  _globals['_TENSOREVENTDETAIL']._serialized_start=69
+  _globals['_TENSOREVENTDETAIL']._serialized_end=286
+  _globals['_TENSOREVENTDETAIL_TENSOREVENTOWNER']._serialized_start=231
+  _globals['_TENSOREVENTDETAIL_TENSOREVENTOWNER']._serialized_end=286
+  _globals['_REQUESTDETAIL']._serialized_start=289
+  _globals['_REQUESTDETAIL']._serialized_end=870
+  _globals['_BATCHDETAIL']._serialized_start=873
+  _globals['_BATCHDETAIL']._serialized_end=1229
+  _globals['_PERHOSTINFERENCESTATS']._serialized_start=1232
+  _globals['_PERHOSTINFERENCESTATS']._serialized_end=1397
+  _globals['_TENSORTRANSFERAGGREGATEDRESULT']._serialized_start=1400
+  _globals['_TENSORTRANSFERAGGREGATEDRESULT']._serialized_end=1797
+  _globals['_TENSORTRANSFERAGGREGATEDRESULT_TENSORPATTERNRESULT']._serialized_start=1540
+  _globals['_TENSORTRANSFERAGGREGATEDRESULT_TENSORPATTERNRESULT']._serialized_end=1797
+  _globals['_TENSORTRANSFERAGGREGATEDRESULT_TENSORPATTERNRESULT_PERCENTILETIME']._serialized_start=1744
+  _globals['_TENSORTRANSFERAGGREGATEDRESULT_TENSORPATTERNRESULT_PERCENTILETIME']._serialized_end=1797
+  _globals['_PERBATCHSIZEAGGREGATEDRESULT']._serialized_start=1800
+  _globals['_PERBATCHSIZEAGGREGATEDRESULT']._serialized_end=2042
+  _globals['_PERMODELINFERENCESTATS']._serialized_start=2045
+  _globals['_PERMODELINFERENCESTATS']._serialized_end=2638
+  _globals['_BATCHINGPARAMETERS']._serialized_start=2641
+  _globals['_BATCHINGPARAMETERS']._serialized_end=2801
+  _globals['_MODELIDDATABASE']._serialized_start=2804
+  _globals['_MODELIDDATABASE']._serialized_end=3151
+  _globals['_MODELIDDATABASE_IDTOINDEXENTRY']._serialized_start=3003
+  _globals['_MODELIDDATABASE_IDTOINDEXENTRY']._serialized_end=3051
+  _globals['_MODELIDDATABASE_IDTOBATCHINGPARAMSENTRY']._serialized_start=3053
+  _globals['_MODELIDDATABASE_IDTOBATCHINGPARAMSENTRY']._serialized_end=3151
+  _globals['_TENSORPATTERNDATABASE']._serialized_start=3153
+  _globals['_TENSORPATTERNDATABASE']._serialized_end=3200
+  _globals['_INFERENCESTATS']._serialized_start=3203
+  _globals['_INFERENCESTATS']._serialized_end=3855
+  _globals['_INFERENCESTATS_INFERENCESTATSPERHOSTENTRY']._serialized_start=3631
+  _globals['_INFERENCESTATS_INFERENCESTATSPERHOSTENTRY']._serialized_end=3735
+  _globals['_INFERENCESTATS_INFERENCESTATSPERMODELENTRY']._serialized_start=3737
+  _globals['_INFERENCESTATS_INFERENCESTATSPERMODELENTRY']._serialized_end=3843
+  _globals['_SAMPLEDPERMODELINFERENCESTATSPROTO']._serialized_start=3858
+  _globals['_SAMPLEDPERMODELINFERENCESTATSPROTO']._serialized_end=4015
+  _globals['_SAMPLEDINFERENCESTATSPROTO']._serialized_start=4018
+  _globals['_SAMPLEDINFERENCESTATSPROTO']._serialized_end=4300
+  _globals['_SAMPLEDINFERENCESTATSPROTO_SAMPLEDINFERENCESTATSPERMODELENTRY']._serialized_start=4175
+  _globals['_SAMPLEDINFERENCESTATSPROTO_SAMPLEDINFERENCESTATSPERMODELENTRY']._serialized_end=4300
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/input_pipeline_pb2.py

@@ -0,0 +1,52 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/input_pipeline.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/input_pipeline.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+from google.protobuf import any_pb2 as google_dot_protobuf_dot_any__pb2
+from xprof.protobuf import diagnostics_pb2 as plugin_dot_xprof_dot_protobuf_dot_diagnostics__pb2
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n*plugin/xprof/protobuf/input_pipeline.proto\x12\x13tensorflow.profiler\x1a\x19google/protobuf/any.proto\x1a\'plugin/xprof/protobuf/diagnostics.proto\"\x80\x03\n\x12\x42ottleneckAnalysis\x12\x15\n\rinput_percent\x18\x07 \x01(\x01\x12\x16\n\x0eoutput_percent\x18\x08 \x01(\x01\x12\x14\n\x0cidle_percent\x18\t \x01(\x01\x12\x17\n\x0f\x63ompute_percent\x18\n \x01(\x01\x12\x1c\n\x14input_classification\x18\x01 \x01(\t\x12\x17\n\x0finput_statement\x18\x02 \x01(\t\x12$\n\x1ckernel_launch_classification\x18\x03 \x01(\t\x12\x1f\n\x17kernel_launch_statement\x18\x04 \x01(\t\x12 \n\x18\x61ll_other_classification\x18\x05 \x01(\t\x12\x1b\n\x13\x61ll_other_statement\x18\x06 \x01(\t\x12)\n!device_collectives_classification\x18\x0b \x01(\t\x12$\n\x1c\x64\x65vice_collectives_statement\x18\x0c \x01(\t\"\\\n\x0bStepSummary\x12\x0f\n\x07\x61verage\x18\x01 \x01(\x01\x12\x1a\n\x12standard_deviation\x18\x02 \x01(\x01\x12\x0f\n\x07minimum\x18\x03 \x01(\x01\x12\x0f\n\x07maximum\x18\x04 \x01(\x01\"\xe0\x02\n\x15PerGenericStepDetails\x12\x13\n\x0bstep_number\x18\x01 \x01(\x05\x12\x11\n\tstep_name\x18\x0e \x01(\t\x12\x14\n\x0cstep_time_ms\x18\x02 \x01(\x01\x12\x17\n\x0funknown_time_ms\x18\x03 \x01(\x01\x12\x1a\n\x12host_wait_input_ms\x18\x0b \x01(\x01\x12\x19\n\x11host_to_device_ms\x18\x0c \x01(\x01\x12\x11\n\toutput_ms\x18\x05 \x01(\x01\x12\x19\n\x11\x64\x65vice_compute_ms\x18\x06 \x01(\x01\x12\x1b\n\x13\x64\x65vice_to_device_ms\x18\x07 \x01(\x01\x12\x1d\n\x15\x64\x65vice_collectives_ms\x18\r \x01(\x01\x12\x17\n\x0fhost_compute_ms\x18\x08 \x01(\x01\x12\x17\n\x0fhost_prepare_ms\x18\t \x01(\x01\x12\x17\n\x0fhost_compile_ms\x18\n \x01(\x01J\x04\x08\x04\x10\x05\"\xa5\x01\n\x12InputTimeBreakdown\x12\x1d\n\x15\x64\x65manded_file_read_us\x18\x01 \x01(\x01\x12\x1d\n\x15\x61\x64vanced_file_read_us\x18\x02 \x01(\x01\x12\x18\n\x10preprocessing_us\x18\x03 \x01(\x01\x12\x12\n\nenqueue_us\x18\x04 \x01(\x01\x12#\n\x1bunclassified_non_enqueue_us\x18\x05 \x01(\x01\"\xa6\x01\n\x0eInputOpDetails\x12\x0f\n\x07op_name\x18\x01 \x01(\t\x12\r\n\x05\x63ount\x18\x02 \x01(\x04\x12\x12\n\ntime_in_ms\x18\x03 \x01(\x01\x12\x17\n\x0ftime_in_percent\x18\x04 \x01(\x01\x12\x17\n\x0fself_time_in_ms\x18\x05 \x01(\x01\x12\x1c\n\x14self_time_in_percent\x18\x06 \x01(\x01\x12\x10\n\x08\x63\x61tegory\x18\x07 \x01(\t\"\x84\x01\n#InputPipelineAnalysisRecommendation\x12\x0f\n\x07\x64\x65tails\x18\x01 \x03(\t\x12\x31\n\x13\x62ottleneck_analysis\x18\x02 \x01(\x0b\x32\x14.google.protobuf.Any\x12\x19\n\x11summary_next_step\x18\x03 \x01(\t\"\x85\x06\n\x18GenericStepTimeBreakdown\x12\x41\n\x17unknown_time_ms_summary\x18\x01 \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12\x44\n\x1ahost_wait_input_ms_summary\x18\t \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12\x43\n\x19host_to_device_ms_summary\x18\n \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12:\n\x10input_ms_summary\x18\x0b \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12;\n\x11output_ms_summary\x18\x03 \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12\x43\n\x19\x64\x65vice_compute_ms_summary\x18\x04 \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12\x45\n\x1b\x64\x65vice_to_device_ms_summary\x18\x05 \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12G\n\x1d\x64\x65vice_collectives_ms_summary\x18\x0c \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12\x41\n\x17host_compute_ms_summary\x18\x06 \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12\x41\n\x17host_prepare_ms_summary\x18\x07 \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12\x41\n\x17host_compile_ms_summary\x18\x08 \x01(\x0b\x32 .tensorflow.profiler.StepSummaryJ\x04\x08\x02\x10\x03\"\x97\x05\n\x1bInputPipelineAnalysisResult\x12\x0b\n\x03tag\x18\x10 \x01(\x08\x12\x15\n\rhardware_type\x18\t \x01(\t\x12;\n\x11step_time_summary\x18\x02 \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12?\n\x15input_percent_summary\x18\x03 \x01(\x0b\x32 .tensorflow.profiler.StepSummary\x12\x15\n\rinput_percent\x18\x0b \x01(\x01\x12\x16\n\x0eoutput_percent\x18\r \x01(\x01\x12\x14\n\x0cidle_percent\x18\x0e \x01(\x01\x12\x17\n\x0f\x63ompute_percent\x18\x0f \x01(\x01\x12*\n\x0cstep_details\x18\x04 \x03(\x0b\x32\x14.google.protobuf.Any\x12\x45\n\x14input_time_breakdown\x18\x05 \x01(\x0b\x32\'.tensorflow.profiler.InputTimeBreakdown\x12=\n\x10input_op_details\x18\x06 \x03(\x0b\x32#.tensorflow.profiler.InputOpDetails\x12P\n\x0erecommendation\x18\x07 \x01(\x0b\x32\x38.tensorflow.profiler.InputPipelineAnalysisRecommendation\x12\x31\n\x13step_time_breakdown\x18\x08 \x01(\x0b\x32\x14.google.protobuf.Any\x12\x35\n\x0b\x64iagnostics\x18\x0c \x01(\x0b\x32 .tensorflow.profiler.DiagnosticsJ\x04\x08\x01\x10\x02J\x04\x08\n\x10\x0b\x62\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.input_pipeline_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_BOTTLENECKANALYSIS']._serialized_start=136
+  _globals['_BOTTLENECKANALYSIS']._serialized_end=520
+  _globals['_STEPSUMMARY']._serialized_start=522
+  _globals['_STEPSUMMARY']._serialized_end=614
+  _globals['_PERGENERICSTEPDETAILS']._serialized_start=617
+  _globals['_PERGENERICSTEPDETAILS']._serialized_end=969
+  _globals['_INPUTTIMEBREAKDOWN']._serialized_start=972
+  _globals['_INPUTTIMEBREAKDOWN']._serialized_end=1137
+  _globals['_INPUTOPDETAILS']._serialized_start=1140
+  _globals['_INPUTOPDETAILS']._serialized_end=1306
+  _globals['_INPUTPIPELINEANALYSISRECOMMENDATION']._serialized_start=1309
+  _globals['_INPUTPIPELINEANALYSISRECOMMENDATION']._serialized_end=1441
+  _globals['_GENERICSTEPTIMEBREAKDOWN']._serialized_start=1444
+  _globals['_GENERICSTEPTIMEBREAKDOWN']._serialized_end=2217
+  _globals['_INPUTPIPELINEANALYSISRESULT']._serialized_start=2220
+  _globals['_INPUTPIPELINEANALYSISRESULT']._serialized_end=2883
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/kernel_stats_pb2.py

@@ -0,0 +1,38 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/kernel_stats.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/kernel_stats.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n(plugin/xprof/protobuf/kernel_stats.proto\x12\x13tensorflow.profiler\"\xeb\x02\n\x0cKernelReport\x12\x0c\n\x04name\x18\x01 \x01(\t\x12\x1c\n\x14registers_per_thread\x18\x02 \x01(\r\x12\x1a\n\x12static_shmem_bytes\x18\x03 \x01(\r\x12\x1b\n\x13\x64ynamic_shmem_bytes\x18\x04 \x01(\r\x12\x11\n\tblock_dim\x18\x05 \x03(\r\x12\x10\n\x08grid_dim\x18\x06 \x03(\r\x12\x19\n\x11total_duration_ns\x18\x07 \x01(\x04\x12\x17\n\x0fmin_duration_ns\x18\x08 \x01(\x04\x12\x17\n\x0fmax_duration_ns\x18\t \x01(\x04\x12#\n\x1bis_kernel_using_tensor_core\x18\n \x01(\x08\x12\"\n\x1ais_op_tensor_core_eligible\x18\x0b \x01(\x08\x12\x0f\n\x07op_name\x18\x0c \x01(\t\x12\x13\n\x0boccurrences\x18\r \x01(\r\x12\x15\n\roccupancy_pct\x18\x0e \x01(\x02\"C\n\rKernelStatsDb\x12\x32\n\x07reports\x18\x01 \x03(\x0b\x32!.tensorflow.profiler.KernelReportb\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.kernel_stats_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_KERNELREPORT']._serialized_start=66
+  _globals['_KERNELREPORT']._serialized_end=429
+  _globals['_KERNELSTATSDB']._serialized_start=431
+  _globals['_KERNELSTATSDB']._serialized_end=498
+# @@protoc_insertion_point(module_scope)

xprof/protobuf/memory_profile_pb2.py

@@ -0,0 +1,54 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: plugin/xprof/protobuf/memory_profile.proto
+# Protobuf Python Version: 6.31.1
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC,
+    6,
+    31,
+    1,
+    '',
+    'plugin/xprof/protobuf/memory_profile.proto'
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n*plugin/xprof/protobuf/memory_profile.proto\x12\x13tensorflow.profiler\"\xa1\x01\n\x16MemoryAggregationStats\x12\x1c\n\x14stack_reserved_bytes\x18\x01 \x01(\x03\x12\x1c\n\x14heap_allocated_bytes\x18\x02 \x01(\x03\x12\x19\n\x11\x66ree_memory_bytes\x18\x03 \x01(\x03\x12\x15\n\rfragmentation\x18\x04 \x01(\x01\x12\x19\n\x11peak_bytes_in_use\x18\x05 \x01(\x03\"\xfd\x01\n\x16MemoryActivityMetadata\x12<\n\x0fmemory_activity\x18\x01 \x01(\x0e\x32#.tensorflow.profiler.MemoryActivity\x12\x17\n\x0frequested_bytes\x18\x02 \x01(\x03\x12\x18\n\x10\x61llocation_bytes\x18\x03 \x01(\x03\x12\x0f\n\x07\x61\x64\x64ress\x18\x04 \x01(\x04\x12\x12\n\ntf_op_name\x18\x05 \x01(\t\x12\x0f\n\x07step_id\x18\x06 \x01(\x03\x12\x13\n\x0bregion_type\x18\x07 \x01(\t\x12\x11\n\tdata_type\x18\x08 \x01(\t\x12\x14\n\x0ctensor_shape\x18\t \x01(\t\"\xbf\x01\n\x15MemoryProfileSnapshot\x12\x16\n\x0etime_offset_ps\x18\x01 \x01(\x03\x12\x46\n\x11\x61ggregation_stats\x18\x02 \x01(\x0b\x32+.tensorflow.profiler.MemoryAggregationStats\x12\x46\n\x11\x61\x63tivity_metadata\x18\x03 \x01(\x0b\x32+.tensorflow.profiler.MemoryActivityMetadata\"\xaf\x01\n\x14MemoryProfileSummary\x12!\n\x19peak_bytes_usage_lifetime\x18\x01 \x01(\x03\x12?\n\npeak_stats\x18\x02 \x01(\x0b\x32+.tensorflow.profiler.MemoryAggregationStats\x12\x1a\n\x12peak_stats_time_ps\x18\x03 \x01(\x03\x12\x17\n\x0fmemory_capacity\x18\x04 \x01(\x03\"Z\n\x10\x41\x63tiveAllocation\x12\x16\n\x0esnapshot_index\x18\x01 \x01(\x03\x12\x15\n\rspecial_index\x18\x02 \x01(\x03\x12\x17\n\x0fnum_occurrences\x18\x03 \x01(\x03\"\x8a\x03\n\x19PerAllocatorMemoryProfile\x12L\n\x18memory_profile_snapshots\x18\x01 \x03(\x0b\x32*.tensorflow.profiler.MemoryProfileSnapshot\x12\x42\n\x0fprofile_summary\x18\x02 \x01(\x0b\x32).tensorflow.profiler.MemoryProfileSummary\x12\x41\n\x12\x61\x63tive_allocations\x18\x03 \x03(\x0b\x32%.tensorflow.profiler.ActiveAllocation\x12H\n\x13special_allocations\x18\x04 \x03(\x0b\x32+.tensorflow.profiler.MemoryActivityMetadata\x12N\n\x1asampled_timeline_snapshots\x18\x05 \x03(\x0b\x32*.tensorflow.profiler.MemoryProfileSnapshot\"\xa8\x02\n\rMemoryProfile\x12g\n\x1cmemory_profile_per_allocator\x18\x01 \x03(\x0b\x32\x41.tensorflow.profiler.MemoryProfile.MemoryProfilePerAllocatorEntry\x12\x11\n\tnum_hosts\x18\x02 \x01(\x05\x12\x12\n\nmemory_ids\x18\x03 \x03(\t\x12\x0f\n\x07version\x18\x05 \x01(\x05\x1ap\n\x1eMemoryProfilePerAllocatorEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12=\n\x05value\x18\x02 \x01(\x0b\x32..tensorflow.profiler.PerAllocatorMemoryProfile:\x02\x38\x01J\x04\x08\x04\x10\x05*h\n\x0eMemoryActivity\x12\x14\n\x10UNKNOWN_ACTIVITY\x10\x00\x12\x0e\n\nALLOCATION\x10\x01\x12\x10\n\x0c\x44\x45\x41LLOCATION\x10\x02\x12\x0f\n\x0bRESERVATION\x10\x03\x12\r\n\tEXPANSION\x10\x04\x62\x06proto3')
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(DESCRIPTOR, 'plugin.xprof.protobuf.memory_profile_pb2', _globals)
+if not _descriptor._USE_C_DESCRIPTORS:
+  DESCRIPTOR._loaded_options = None
+  _globals['_MEMORYPROFILE_MEMORYPROFILEPERALLOCATORENTRY']._loaded_options = None
+  _globals['_MEMORYPROFILE_MEMORYPROFILEPERALLOCATORENTRY']._serialized_options = b'8\001'
+  _globals['_MEMORYACTIVITY']._serialized_start=1647
+  _globals['_MEMORYACTIVITY']._serialized_end=1751
+  _globals['_MEMORYAGGREGATIONSTATS']._serialized_start=68
+  _globals['_MEMORYAGGREGATIONSTATS']._serialized_end=229
+  _globals['_MEMORYACTIVITYMETADATA']._serialized_start=232
+  _globals['_MEMORYACTIVITYMETADATA']._serialized_end=485
+  _globals['_MEMORYPROFILESNAPSHOT']._serialized_start=488
+  _globals['_MEMORYPROFILESNAPSHOT']._serialized_end=679
+  _globals['_MEMORYPROFILESUMMARY']._serialized_start=682
+  _globals['_MEMORYPROFILESUMMARY']._serialized_end=857
+  _globals['_ACTIVEALLOCATION']._serialized_start=859
+  _globals['_ACTIVEALLOCATION']._serialized_end=949
+  _globals['_PERALLOCATORMEMORYPROFILE']._serialized_start=952
+  _globals['_PERALLOCATORMEMORYPROFILE']._serialized_end=1346
+  _globals['_MEMORYPROFILE']._serialized_start=1349
+  _globals['_MEMORYPROFILE']._serialized_end=1645
+  _globals['_MEMORYPROFILE_MEMORYPROFILEPERALLOCATORENTRY']._serialized_start=1527
+  _globals['_MEMORYPROFILE_MEMORYPROFILEPERALLOCATORENTRY']._serialized_end=1639
+# @@protoc_insertion_point(module_scope)