Change step: 0 to step: 1 to resolve error

EdjeElectronics · web-flow · commit f0cb960a1dbf · 2019-09-14T14:16:26.000-06:00
This resolves the "ValueError(): First step cannot be zero" error that is encountered when using this pipeline config file with newer versions of TensorFlow.
diff --git a/training/faster_rcnn_inception_v2_pets.config b/training/faster_rcnn_inception_v2_pets.config
@@ -89,7 +89,7 @@ train_config: {
         manual_step_learning_rate {
           initial_learning_rate: 0.0002
           schedule {
-            step: 0
+            step: 1
             learning_rate: .0002
           }
           schedule {

Original file line number	Diff line number	Diff line change
`@@ -89,7 +89,7 @@ train_config: {`
`89`	`89`	`manual_step_learning_rate {`
`90`	`90`	`initial_learning_rate: 0.0002`
`91`	`91`	`schedule {`
`92`		`- step: 0`
	`92`	`+ step: 1`
`93`	`93`	`learning_rate: .0002`
`94`	`94`	`}`
`95`	`95`	`schedule {`