Utility to periodically log active tensors See lamp.data.TensorLogger#start
Attributes
- Companion
- class
- Graph
-
- Supertypes
-
trait Producttrait Mirrorclass Objecttrait Matchableclass Any
- Self type
-
TensorLogger.type
Members list
Type members
Inherited types
The names of the product elements
The names of the product elements
Attributes
- Inherited from:
- Mirror
The name of the type
The name of the type
Attributes
- Inherited from:
- Mirror
Value members
Concrete methods
Log active tensors with a given frequency
Log active tensors with a given frequency
The log string itself consists of a summary statistics of active tensors and a list of detailed information on some number of active tensors. The detailed information shows the stack trace where the tensor was allocated. The tensors eligible for detail are specified by a filter on their lifetime.
This method will spawn a new Thread, which repeatedly logs and sleeps until not canceled.
Value parameters
- detailMaxMs
-
maximum lifetime in milliseconds for a tensor to be eligible for detailed logging
- detailMinMs
-
minimum lifetime in milliseconds for a tensor to be eligible for detailed logging
- detailNum
-
max number of tensors in the detailed section
- filter
-
a predicate taking some information about the tensor and the lifetime of the tensor in milliseconds
- frequency
-
the frequency with which the logging will occur
- logger
-
a side effecting lambda which executes the log
Attributes
- Returns
-
an instance of TensorLogger whose sole purpose is its
cancel()
method which stops the logger