Creates a summary database writer in the current context.
tf.contrib.summary.create_db_writer(
db_uri, experiment_name=None, run_name=None, user_name=None, name=None
)
This can be used to write tensors from the execution graph directly
to a database. Only SQLite is supported right now. This function
will create the schema if it doesn't exist. Entries in the Users,
Experiments, and Runs tables will be created automatically if they
don't already exist.
Args |
|---|
db_uri
|
For example "file:/tmp/foo.sqlite".
|
experiment_name
|
Defaults to YYYY-MM-DD in local time if None.
Empty string means the Run will not be associated with an
Experiment. Can't contain ASCII control characters or <>. Case
sensitive.
|
run_name
|
Defaults to HH:MM:SS in local time if None. Empty string
means a Tag will not be associated with any Run. Can't contain
ASCII control characters or <>. Case sensitive.
|
user_name
|
Defaults to system username if None. Empty means the
Experiment will not be associated with a User. Must be valid as
both a DNS label and Linux username.
|
name
|
Shared name for this SummaryWriter resource stored to default
tf.Graph.
|
View source on GitHub