Fix late initialization field (#165)

Issue #, if available:
When a distributed/parallelized pytorch workload is run, AlgorithmSpecification.MetricDefinitions does not late initialize which makes the controller re queue for a long time. 

Description of changes:
Edited generator.yaml to remove late_initialize and to ignore comparisons(ACK trainingjob does not support update operations so this wont effect the user). 

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

Author: ananth102

apis/v1alpha1/ack-generate-metadata.yaml

@@ -1,13 +1,13 @@
 ack_generate_info:
-  build_date: "2022-08-12T20:32:53Z"
+  build_date: "2022-08-17T21:49:11Z"
   build_hash: fe61d04673fd4d9848d5f726b01e0689a16d3733
   go_version: go1.17.1
   version: v0.19.3-1-gfe61d04
 api_directory_checksum: 8b3c128d2037d5227679cccb57cd4d78af6aed1b
 api_version: v1alpha1
 aws_sdk_go_version: v1.42.28
 generator_config_info:
-  file_checksum: 92038dd37f8302115ea2733a22973f551e60c78e
+  file_checksum: b97e2519e420215d25350e3df27f956c482e828e
   original_file_name: generator.yaml
 last_modification:
   reason: API generation

apis/v1alpha1/generator.yaml

@@ -171,8 +171,8 @@ resources:
           operation: DescribeTrainingJob
           path: FailureReason 
       AlgorithmSpecification.MetricDefinitions:
-        late_initialize:
-          min_backoff_seconds: 5
+        compare:
+          is_ignored: true
       EnableInterContainerTrafficEncryption:
         late_initialize:
           min_backoff_seconds: 5

generator.yaml

@@ -171,8 +171,8 @@ resources:
           operation: DescribeTrainingJob
           path: FailureReason 
       AlgorithmSpecification.MetricDefinitions:
-        late_initialize:
-          min_backoff_seconds: 5
+        compare:
+          is_ignored: true
       EnableInterContainerTrafficEncryption:
         late_initialize:
           min_backoff_seconds: 5