[
https://issues.apache.org/jira/browse/MINIFICPP-358?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Andrew Christianson updated MINIFICPP-358:
------------------------------------------
Description: Add support for interpreting output tensors which represent
labels with scores. A list of labels in index order is passed in with attribute
tf.type == "labels". The top N labels are extracted into attributes
tf.label.<N> and tf.score.<N>. The tf.label.0 attribute represents the
highest-scoring label and can be used, for example, to route FlowFiles based on
model inferences. (was: Add support for interpreting output tensors which
represent labels with scores. A list of labels in index order is passed in with
attribute tf.type == "labels". The top N labels are extracted into attributes
tf.label.<N> and tf.score.<N>. the tf.label.0 attribute represents the
highest-scoring label and can be used, for example, to route FlowFiles based on
model inferences.)
> Implement TFExtractTopLabels
> ----------------------------
>
> Key: MINIFICPP-358
> URL: https://issues.apache.org/jira/browse/MINIFICPP-358
> Project: NiFi MiNiFi C++
> Issue Type: Improvement
> Reporter: Andrew Christianson
>
> Add support for interpreting output tensors which represent labels with
> scores. A list of labels in index order is passed in with attribute tf.type
> == "labels". The top N labels are extracted into attributes tf.label.<N> and
> tf.score.<N>. The tf.label.0 attribute represents the highest-scoring label
> and can be used, for example, to route FlowFiles based on model inferences.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
