Get desktop application:
View/edit binary Protocol Buffers messages
Get metadata about an experiment.
Human-readable description of the data source. This might be an absolute path on disk to a log directory of event files, a relative path, a URL to a cloud storage bucket, an address for a local socket server, etc. There does not need to be any way to "dereference" this address to read the original data directly.
User-facing experiment name. This name is typically set by a user explicitly or automatically set by a training system.
User-facing experiment description.
Time that the experiment was created. This does not necessarily have any relation to times of events within an experiment. A new experiment may contain old data, or data may be freshly written to an experiment after it is created. May be unset if no creation time is known.
List plugins that have data for an experiment.
ID of experiment in which to query data.
List of active plugins: i.e., plugins that have data.
List runs within an experiment.
ID of experiment in which to query data.
List metadata about scalar time series.
ID of experiment in which to query data.
Filter for the plugin that owns the scalars. It is an error if `plugin_filter.plugin_name` is the empty string.
Optional filter for time series. If omitted, all time series match.
Read data from scalar time series.
ID of experiment in which to query data.
Filter for the plugin that owns the scalars. It is an error if `plugin_filter.plugin_name` is the empty string.
Optional filter for time series. If omitted, all time series match.
Required downsampling specification describing how many points to return per time series.
List metadata about tensor time series.
ID of experiment in which to query data.
Filter for the plugin that owns the tensors. It is an error if `plugin_filter.plugin_name` is the empty string.
Optional filter for time series. If omitted, all time series match.
Read data from tensor time series.
ID of experiment in which to query data.
Filter for the plugin that owns the tensors. It is an error if `plugin_filter.plugin_name` is the empty string.
Optional filter for time series. If omitted, all time series match.
Required downsampling specification describing how many points to return per time series.
List metadata about blob sequence time series.
ID of experiment in which to query data.
Filter for the plugin that owns the blob sequences. It is an error if `plugin_filter.plugin_name` is the empty string.
Optional filter for time series. If omitted, all time series match.
Read blob references from blob sequence time series. See `ReadBlob` to read the actual blob data.
ID of experiment in which to query data.
Filter for the plugin that owns the blob sequences. It is an error if `plugin_filter.plugin_name` is the empty string.
Optional filter for time series. If omitted, all time series match.
Required downsampling specification describing how many points to return per time series.
Read data for a specific blob.
The bytes in this chunk. Should be concatenated with any other responses in the stream to recover the full blob contents.
A reference to a blob.
Used in:
Unique identifier for a blob, which may be dereferenced via the `ReadBlob` RPC. Must be suitable for inclusion directly in a URL with no further encoding. Case-sensitive. Required; the empty string is not a valid key.
Optional URL from which the blob may be fetched directly, bypassing the data provider interface.
Used in:
A column-major sequence of blob sequence points. Arrays `step`, `wall_time`, and `value` have the same lengths. The `step` and `wall_time` fields are repeated primitive values, so they will be packed on the wire: <https://developers.google.com/protocol-buffers/docs/encoding#packed>
Used in:
Used in:
Largest step value of any datum in this time series.
Largest wall time of any datum in this time series.
Largest number of blobs in any datum in this time series.
Atemporal summary metadata for this time series.
Used in:
, ,Maximum number of points to return. Must be non-negative. Zero means zero.
Used in:
Used in:
Used in:
Used in:
Used in:
Used in:
Used in:
Plugin name, as listed in the `PluginData.plugin_name` field of a `SummaryMetadata` value.
Used in:
, , , , ,Only match data with exactly this plugin name.
Used in:
Used in:
Used in:
Used in:
Used in:
Used in:
Used in:
User-facing name.
Wall time of earliest recorded event, as floating-point seconds since epoch (same as event file format).
Used in:
Only match runs with exactly one of these names. In particular, if this list is empty, no runs match.
Used in:
, , , , ,Optional filter for runs. If omitted, *all* runs match; this is not equivalent to an empty submessage (which matches no tags).
Optional filter for tags. If omitted, *all* tags match; this is not equivalent to an empty submessage (which matches no tags).
A column-major sequence of scalar points. Arrays `step`, `wall_time`, and `value` have the same lengths. These are repeated primitive values, so they will be packed on the wire: <https://developers.google.com/protocol-buffers/docs/encoding#packed>
Used in:
Used in:
Largest step value of any datum in this time series.
Largest wall time of any datum in this time series.
Atemporal summary metadata for this time series.
Used in:
Only match tags with exactly one of these names. In particular, if this list is empty, no tags match.
A column-major sequence of tensor points. Arrays `step`, `wall_time`, and `value` have the same lengths. The `step` and `wall_time` fields are repeated primitive values, so they will be packed on the wire: <https://developers.google.com/protocol-buffers/docs/encoding#packed>
Used in:
Used in:
Largest step value of any datum in this time series.
Largest wall time of any datum in this time series.
Atemporal summary metadata for this time series.