# 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. # ============================================================================== """Wraps the base_plugin.TBContext to stores additional data shared across API handlers for the HParams plugin backend.""" import collections import os from tensorboard.data import provider from tensorboard.plugins.hparams import api_pb2 from tensorboard.plugins.hparams import json_format_compat from tensorboard.plugins.hparams import metadata from google.protobuf import json_format from tensorboard.plugins.scalar import metadata as scalar_metadata _DISCRETE_DOMAIN_TYPE_TO_DATA_TYPE = { provider.HyperparameterDomainType.DISCRETE_BOOL: api_pb2.DATA_TYPE_BOOL, provider.HyperparameterDomainType.DISCRETE_FLOAT: api_pb2.DATA_TYPE_FLOAT64, provider.HyperparameterDomainType.DISCRETE_STRING: api_pb2.DATA_TYPE_STRING, } class Context: """Wraps the base_plugin.TBContext to stores additional data shared across API handlers for the HParams plugin backend. Before adding fields to this class, carefully consider whether the field truly needs to be accessible to all API handlers or if it can be passed separately to the handler constructor. We want to avoid this class becoming a magic container of variables that have no better place. See http://wiki.c2.com/?MagicContainer """ def __init__(self, tb_context): """Instantiates a context. Args: tb_context: base_plugin.TBContext. The "base" context we extend. """ self._tb_context = tb_context def experiment_from_metadata( self, ctx, experiment_id, include_metrics, hparams_run_to_tag_to_content, data_provider_hparams, hparams_limit=None, ): """Returns the experiment proto defining the experiment. This method first attempts to find a metadata.EXPERIMENT_TAG tag and retrieve the associated proto. If no such tag is found, the method will attempt to build a minimal experiment proto by scanning for all metadata.SESSION_START_INFO_TAG tags (to compute the hparam_infos field of the experiment) and for all scalar tags (to compute the metric_infos field of the experiment). If no metadata.EXPERIMENT_TAG nor metadata.SESSION_START_INFO_TAG tags are found, then will build an experiment proto using the results from DataProvider.list_hyperparameters(). Args: experiment_id: String, from `plugin_util.experiment_id`. include_metrics: Whether to determine metrics_infos and include them in the result. hparams_run_to_tag_to_content: The output from an hparams_metadata() call. A dict `d` such that `d[run][tag]` is a `bytes` value with the summary metadata content for the keyed time series. data_provider_hparams: The output from an hparams_from_data_provider() call, corresponding to DataProvider.list_hyperparameters(). A provider.ListHyperpararametersResult. hparams_limit: Optional number of hyperparameter metadata to include in the result. If unset or zero, all metadata will be included. Returns: The experiment proto. If no data is found for an experiment proto to be built, returns an entirely empty experiment. """ experiment = self._find_experiment_tag( hparams_run_to_tag_to_content, include_metrics ) if experiment: _sort_and_reduce_to_hparams_limit(experiment, hparams_limit) return experiment experiment_from_runs = self._compute_experiment_from_runs( ctx, experiment_id, include_metrics, hparams_run_to_tag_to_content ) if experiment_from_runs: _sort_and_reduce_to_hparams_limit( experiment_from_runs, hparams_limit ) return experiment_from_runs experiment_from_data_provider_hparams = ( self._experiment_from_data_provider_hparams( ctx, experiment_id, include_metrics, data_provider_hparams ) ) return ( experiment_from_data_provider_hparams if experiment_from_data_provider_hparams else api_pb2.Experiment() ) @property def tb_context(self): return self._tb_context def _convert_plugin_metadata(self, data_provider_output): return { run: { tag: time_series.plugin_content for (tag, time_series) in tag_to_time_series.items() } for (run, tag_to_time_series) in data_provider_output.items() } def hparams_metadata(self, ctx, experiment_id, run_tag_filter=None): """Reads summary metadata for all hparams time series. Args: experiment_id: String, from `plugin_util.experiment_id`. run_tag_filter: Optional `data.provider.RunTagFilter`, with the semantics as in `list_tensors`. Returns: A dict `d` such that `d[run][tag]` is a `bytes` value with the summary metadata content for the keyed time series. """ return self._convert_plugin_metadata( self._tb_context.data_provider.list_tensors( ctx, experiment_id=experiment_id, plugin_name=metadata.PLUGIN_NAME, run_tag_filter=run_tag_filter, ) ) def scalars_metadata(self, ctx, experiment_id): """Reads summary metadata for all scalar time series. Args: experiment_id: String, from `plugin_util.experiment_id`. Returns: A dict `d` such that `d[run][tag]` is a `bytes` value with the summary metadata content for the keyed time series. """ return self._convert_plugin_metadata( self._tb_context.data_provider.list_scalars( ctx, experiment_id=experiment_id, plugin_name=scalar_metadata.PLUGIN_NAME, ) ) def read_last_scalars(self, ctx, experiment_id, run_tag_filter): """Reads the most recent values from scalar time series. Args: experiment_id: String. run_tag_filter: Required `data.provider.RunTagFilter`, with the semantics as in `read_last_scalars`. Returns: A dict `d` such that `d[run][tag]` is a `provider.ScalarDatum` value, with keys only for runs and tags that actually had data, which may be a subset of what was requested. """ return self._tb_context.data_provider.read_last_scalars( ctx, experiment_id=experiment_id, plugin_name=scalar_metadata.PLUGIN_NAME, run_tag_filter=run_tag_filter, ) def hparams_from_data_provider(self, ctx, experiment_id, limit): """Calls DataProvider.list_hyperparameters() and returns the result.""" return self._tb_context.data_provider.list_hyperparameters( ctx, experiment_ids=[experiment_id], limit=limit ) def session_groups_from_data_provider( self, ctx, experiment_id, filters, sort, hparams_to_include ): """Calls DataProvider.read_hyperparameters() and returns the result.""" return self._tb_context.data_provider.read_hyperparameters( ctx, experiment_ids=[experiment_id], filters=filters, sort=sort, hparams_to_include=hparams_to_include, ) def _find_experiment_tag( self, hparams_run_to_tag_to_content, include_metrics ): """Finds the experiment associated with the metadata.EXPERIMENT_TAG tag. Returns: The experiment or None if no such experiment is found. """ # We expect only one run to have an `EXPERIMENT_TAG`; look # through all of them and arbitrarily pick the first one. for tags in hparams_run_to_tag_to_content.values(): maybe_content = tags.get(metadata.EXPERIMENT_TAG) if maybe_content is not None: experiment = metadata.parse_experiment_plugin_data( maybe_content ) if not include_metrics: # metric_infos haven't technically been "calculated" in this # case. They have been read directly from the Experiment # proto. # Delete them from the result so that they are not returned # to the client. experiment.ClearField("metric_infos") return experiment return None def _compute_experiment_from_runs( self, ctx, experiment_id, include_metrics, hparams_run_to_tag_to_content ): """Computes a minimal Experiment protocol buffer by scanning the runs. Returns None if there are no hparam infos logged. """ hparam_infos = self._compute_hparam_infos(hparams_run_to_tag_to_content) metric_infos = ( self._compute_metric_infos_from_runs( ctx, experiment_id, hparams_run_to_tag_to_content ) if hparam_infos and include_metrics else [] ) if not hparam_infos and not metric_infos: return None return api_pb2.Experiment( hparam_infos=hparam_infos, metric_infos=metric_infos ) def _compute_hparam_infos(self, hparams_run_to_tag_to_content): """Computes a list of api_pb2.HParamInfo from the current run, tag info. Finds all the SessionStartInfo messages and collects the hparams values appearing in each one. For each hparam attempts to deduce a type that fits all its values. Finally, sets the 'domain' of the resulting HParamInfo to be discrete if the type is string or boolean. Returns: A list of api_pb2.HParamInfo messages. """ # Construct a dict mapping an hparam name to its list of values. hparams = collections.defaultdict(list) for tag_to_content in hparams_run_to_tag_to_content.values(): if metadata.SESSION_START_INFO_TAG not in tag_to_content: continue start_info = metadata.parse_session_start_info_plugin_data( tag_to_content[metadata.SESSION_START_INFO_TAG] ) for (name, value) in start_info.hparams.items(): hparams[name].append(value) # Try to construct an HParamInfo for each hparam from its name and list # of values. result = [] for (name, values) in hparams.items(): hparam_info = self._compute_hparam_info_from_values(name, values) if hparam_info is not None: result.append(hparam_info) return result def _compute_hparam_info_from_values(self, name, values): """Builds an HParamInfo message from the hparam name and list of values. Args: name: string. The hparam name. values: list of google.protobuf.Value messages. The list of values for the hparam. Returns: An api_pb2.HParamInfo message. """ # Figure out the type from the values. # Ignore values whose type is not listed in api_pb2.DataType # If all values have the same type, then that is the type used. # Otherwise, the returned type is DATA_TYPE_STRING. result = api_pb2.HParamInfo(name=name, type=api_pb2.DATA_TYPE_UNSET) for v in values: v_type = _protobuf_value_type(v) if not v_type: continue if result.type == api_pb2.DATA_TYPE_UNSET: result.type = v_type elif result.type != v_type: result.type = api_pb2.DATA_TYPE_STRING if result.type == api_pb2.DATA_TYPE_STRING: # A string result.type does not change, so we can exit the loop. break # If we couldn't figure out a type, then we can't compute the hparam_info. if result.type == api_pb2.DATA_TYPE_UNSET: return None if result.type == api_pb2.DATA_TYPE_STRING: distinct_string_values = set( _protobuf_value_to_string(v) for v in values if _can_be_converted_to_string(v) ) result.domain_discrete.extend(distinct_string_values) result.differs = len(distinct_string_values) > 1 if result.type == api_pb2.DATA_TYPE_BOOL: distinct_bool_values = set(v.bool_value for v in values) result.domain_discrete.extend(distinct_bool_values) result.differs = len(distinct_bool_values) > 1 if result.type == api_pb2.DATA_TYPE_FLOAT64: # Always uses interval domain type for numeric hparam values. distinct_float_values = sorted([v.number_value for v in values]) if distinct_float_values: result.domain_interval.min_value = distinct_float_values[0] result.domain_interval.max_value = distinct_float_values[-1] result.differs = len(set(distinct_float_values)) > 1 return result def _experiment_from_data_provider_hparams( self, ctx, experiment_id, include_metrics, data_provider_hparams, ): """Returns an experiment protobuffer based on data provider hparams. Args: experiment_id: String, from `plugin_util.experiment_id`. include_metrics: Whether to determine metrics_infos and include them in the result. data_provider_hparams: The output from an hparams_from_data_provider() call, corresponding to DataProvider.list_hyperparameters(). A provider.ListHyperparametersResult. Returns: The experiment proto. If there are no hyperparameters in the input, returns None. """ if isinstance(data_provider_hparams, list): # TODO: Support old return value of Collection[provider.Hyperparameters] # until all internal implementations of DataProvider can be # migrated to use new return value of provider.ListHyperparametersResult. hyperparameters = data_provider_hparams session_groups = [] else: # Is instance of provider.ListHyperparametersResult hyperparameters = data_provider_hparams.hyperparameters session_groups = data_provider_hparams.session_groups hparam_infos = [ self._convert_data_provider_hparam(dp_hparam) for dp_hparam in hyperparameters ] metric_infos = ( self.compute_metric_infos_from_data_provider_session_groups( ctx, experiment_id, session_groups ) if include_metrics else [] ) return api_pb2.Experiment( hparam_infos=hparam_infos, metric_infos=metric_infos ) def _convert_data_provider_hparam(self, dp_hparam): """Builds an HParamInfo message from data provider Hyperparameter. Args: dp_hparam: The provider.Hyperparameter returned by the call to provider.DataProvider.list_hyperparameters(). Returns: An HParamInfo to include in the Experiment. """ hparam_info = api_pb2.HParamInfo( name=dp_hparam.hyperparameter_name, display_name=dp_hparam.hyperparameter_display_name, differs=dp_hparam.differs, ) if dp_hparam.domain_type == provider.HyperparameterDomainType.INTERVAL: hparam_info.type = api_pb2.DATA_TYPE_FLOAT64 (dp_hparam_min, dp_hparam_max) = dp_hparam.domain hparam_info.domain_interval.min_value = dp_hparam_min hparam_info.domain_interval.max_value = dp_hparam_max elif dp_hparam.domain_type in _DISCRETE_DOMAIN_TYPE_TO_DATA_TYPE.keys(): hparam_info.type = _DISCRETE_DOMAIN_TYPE_TO_DATA_TYPE.get( dp_hparam.domain_type ) hparam_info.domain_discrete.extend(dp_hparam.domain) return hparam_info def _compute_metric_infos_from_runs( self, ctx, experiment_id, hparams_run_to_tag_to_content ): session_runs = set( run for run, tags in hparams_run_to_tag_to_content.items() if metadata.SESSION_START_INFO_TAG in tags ) return ( api_pb2.MetricInfo(name=api_pb2.MetricName(group=group, tag=tag)) for tag, group in self._compute_metric_names( ctx, experiment_id, session_runs ) ) def compute_metric_infos_from_data_provider_session_groups( self, ctx, experiment_id, session_groups ): session_runs = set( generate_data_provider_session_name(s) for sg in session_groups for s in sg.sessions ) return [ api_pb2.MetricInfo(name=api_pb2.MetricName(group=group, tag=tag)) for tag, group in self._compute_metric_names( ctx, experiment_id, session_runs ) ] def _compute_metric_names(self, ctx, experiment_id, session_runs): """Computes the list of metric names from all the scalar (run, tag) pairs. The return value is a list of (tag, group) pairs representing the metric names. The list is sorted in Python tuple-order (lexicographical). For example, if the scalar (run, tag) pairs are: ("exp/session1", "loss") ("exp/session2", "loss") ("exp/session2/eval", "loss") ("exp/session2/validation", "accuracy") ("exp/no-session", "loss_2"), and the runs corresponding to sessions are "exp/session1", "exp/session2", this method will return [("loss", ""), ("loss", "/eval"), ("accuracy", "/validation")] More precisely, each scalar (run, tag) pair is converted to a (tag, group) metric name, where group is the suffix of run formed by removing the longest prefix which is a session run. If no session run is a prefix of 'run', the pair is skipped. Returns: A python list containing pairs. Each pair is a (tag, group) pair representing a metric name used in some session. """ metric_names_set = set() scalars_run_to_tag_to_content = self.scalars_metadata( ctx, experiment_id ) for run, tags in scalars_run_to_tag_to_content.items(): session = _find_longest_parent_path(session_runs, run) if session is None: continue group = os.path.relpath(run, session) # relpath() returns "." for the 'session' directory, we use an empty # string. if group == ".": group = "" metric_names_set.update((tag, group) for tag in tags) metric_names_list = list(metric_names_set) # Sort metrics for determinism. metric_names_list.sort() return metric_names_list def generate_data_provider_session_name(session): """Generates a name from a HyperparameterSesssionRun. If the HyperparameterSessionRun contains no experiment or run information then the name is set to the original experiment_id. """ if not session.experiment_id and not session.run: return "" elif not session.experiment_id: return session.run elif not session.run: return session.experiment_id else: return f"{session.experiment_id}/{session.run}" def _find_longest_parent_path(path_set, path): """Finds the longest "parent-path" of 'path' in 'path_set'. This function takes and returns "path-like" strings which are strings made of strings separated by os.sep. No file access is performed here, so these strings need not correspond to actual files in some file-system.. This function returns the longest ancestor path For example, for path_set=["/foo/bar", "/foo", "/bar/foo"] and path="/foo/bar/sub_dir", returns "/foo/bar". Args: path_set: set of path-like strings -- e.g. a list of strings separated by os.sep. No actual disk-access is performed here, so these need not correspond to actual files. path: a path-like string. Returns: The element in path_set which is the longest parent directory of 'path'. """ # This could likely be more efficiently implemented with a trie # data-structure, but we don't want to add an extra dependency for that. while path not in path_set: if not path: return None path = os.path.dirname(path) return path def _can_be_converted_to_string(value): if not _protobuf_value_type(value): return False return json_format_compat.is_serializable_value(value) def _protobuf_value_type(value): """Returns the type of the google.protobuf.Value message as an api.DataType. Returns None if the type of 'value' is not one of the types supported in api_pb2.DataType. Args: value: google.protobuf.Value message. """ if value.HasField("number_value"): return api_pb2.DATA_TYPE_FLOAT64 if value.HasField("string_value"): return api_pb2.DATA_TYPE_STRING if value.HasField("bool_value"): return api_pb2.DATA_TYPE_BOOL return None def _protobuf_value_to_string(value): """Returns a string representation of given google.protobuf.Value message. Args: value: google.protobuf.Value message. Assumed to be of type 'number', 'string' or 'bool'. """ value_in_json = json_format.MessageToJson(value) if value.HasField("string_value"): # Remove the quotations. return value_in_json[1:-1] return value_in_json def _sort_and_reduce_to_hparams_limit(experiment, hparams_limit=None): """Sorts and applies limit to the hparams in the given experiment proto. Args: experiment: An api_pb2.Experiment proto, which will be modified in place. hparams_limit: Optional number of hyperparameter metadata to include in the result. If unset or zero, no limit will be applied. Returns: None. `experiment` proto will be modified in place. """ if not hparams_limit: # If limit is unset or zero, returns all hparams. hparams_limit = len(experiment.hparam_infos) # Prioritizes returning HParamInfo protos with `differed` values. # Sorts by `differs` (True first), then by name. limited_hparam_infos = sorted( experiment.hparam_infos, key=lambda hparam_info: (not hparam_info.differs, hparam_info.name), )[:hparams_limit] experiment.ClearField("hparam_infos") experiment.hparam_infos.extend(limited_hparam_infos)