a d=ic @s|dZddlZddlZddlmZddlmZddlmZdddZdd d Z dd d Z d dZ e e dejZdS)a"Summary creation methods for the HParams plugin. Typical usage for exporting summaries in a hyperparameters-tuning experiment: 1. Create the experiment (once) by calling experiment_pb() and exporting the resulting summary into a top-level (empty) run. 2. In each training session in the experiment, call session_start_pb() before the session starts, exporting the resulting summary into a uniquely named run for the session, say . 3. Train the model in the session, exporting each metric as a scalar summary in runs of the form /, where can be empty a (in which case the run is just the ) and depends on the metric. The name of such a metric is a (group, tag) pair given by (, tag) where tag is the tag of the scalar summary. When calling experiment_pb in step 1, you'll need to pass all the metric names used in the experiemnt. 4. When the session completes, call session_end_pb() and export the resulting summary into the same session run . N)api_pb2)metadata)plugin_data_pb2cCs8|durt}tj|||||d}ttjtj|dS)aCreates a summary that defines a hyperparameter-tuning experiment. Args: hparam_infos: Array of api_pb2.HParamInfo messages. Describes the hyperparameters used in the experiment. metric_infos: Array of api_pb2.MetricInfo messages. Describes the metrics used in the experiment. See the documentation at the top of this file for how to populate this. user: String. An id for the user running the experiment description: String. A description for the experiment. May contain markdown. time_created_secs: float. The time the experiment is created in seconds since the UNIX epoch. If None uses the current time. Returns: A summary protobuffer containing the experiment definition. N) descriptionusertime_created_secs hparam_infos metric_infos) experiment)timerZ Experiment_summaryrZEXPERIMENT_TAGrHParamsPluginData)r r rrrr rt/home/droni/.local/share/virtualenvs/DPS-5Je3_V2c/lib/python3.9/site-packages/tensorboard/plugins/hparams/summary.py experiment_pb-s rcCs|durt}tj||||d}|D]\}}t|ttfrN||j|_q*t|t rf||j|_ q*t|t r~||j|_ q*t|t tfrt ||j|_ q*td||t|fq*ttjtj|dS)aConstructs a SessionStartInfo protobuffer. Creates a summary that contains a training session metadata information. One such summary per training session should be created. Each should have a different run. Args: hparams: A dictionary with string keys. Describes the hyperparameter values used in the session, mapping each hyperparameter name to its value. Supported value types are `bool`, `int`, `float`, `str`, `list`, `tuple`. The type of value must correspond to the type of hyperparameter (defined in the corresponding api_pb2.HParamInfo member of the Experiment protobuf) as follows: +-----------------+---------------------------------+ |Hyperparameter | Allowed (Python) value types | |type | | +-----------------+---------------------------------+ |DATA_TYPE_BOOL | bool | |DATA_TYPE_FLOAT64| int, float | |DATA_TYPE_STRING | str, tuple, list | +-----------------+---------------------------------+ Tuple and list instances will be converted to their string representation. model_uri: See the comment for the field with the same name of plugin_data_pb2.SessionStartInfo. monitor_url: See the comment for the field with the same name of plugin_data_pb2.SessionStartInfo. group_name: See the comment for the field with the same name of plugin_data_pb2.SessionStartInfo. start_time_secs: float. The time to use as the session start time. Represented as seconds since the UNIX epoch. If None uses the current time. Returns: The summary protobuffer mentioned above. N) model_uri monitor_url group_namestart_time_secsz2hparams[%s]=%s has type: %s which is not supported)session_start_info)r rZSessionStartInfoitems isinstancefloatinthparamsZ number_valuestrZ string_valueboolZ bool_valuelisttuple TypeErrortyper rZSESSION_START_INFO_TAGr)rrrrrrZhp_nameZhp_valrrrsession_start_pbOs8)   r"cCs2|durt}tj||d}ttjtj|dS)axConstructs a SessionEndInfo protobuffer. Creates a summary that contains status information for a completed training session. Should be exported after the training session is completed. One such summary per training session should be created. Each should have a different run. Args: status: A tensorboard.hparams.Status enumeration value denoting the status of the session. end_time_secs: float. The time to use as the session end time. Represented as seconds since the unix epoch. If None uses the current time. Returns: The summary protobuffer mentioned above. N)status end_time_secs)session_end_info)r rZSessionEndInfor rZSESSION_END_INFO_TAGr)r#r$r%rrrsession_end_pbs r&cCsDtjj}t|}|}tjjj|}|j j ||t d|S)zReturns a summary holding the given HParamsPluginData message. Helper function. Args: tag: string. The tag to use. hparams_plugin_data: The HParamsPluginData message to use. )tagrZtensor) tfcompatv1SummaryrZcreate_summary_metadataSerializeToStringZSummaryMetadata FromStringvalueadd_TF_NULL_TENSOR)r'Zhparams_plugin_datasummaryZ tb_metadataZ raw_metadataZ tf_metadatarrrr s  r )rrN)rrrN)N)__doc__r Z tensorflowr(Ztensorboard.plugins.hparamsrrrrr"r&r r!Zmake_tensor_protor-Z NULL_TENSORr,r0rrrrs    # G