Make top candidates view as the default view in user-tools

Signed-off-by: Ahmed Hussein (amahussein) <[email protected]>

Fixes NVIDIA#868

- Use Top-candidate view as default for user-tools qualification
- Revisit the spark_rapids CLI to verify that resetting the filter based
  on user-input does not override the default flag argument
- Fix unit-tests

user_tools/src/spark_rapids_pytools/rapids/qualification.py

@@ -242,14 +242,18 @@ def _process_gpu_cluster_worker_node():
         return gpu_cluster_obj is not None
 
     def _process_offline_cluster_args(self):
-        offline_cluster_opts = self.wrapper_options.get('migrationClustersProps', {})
-        self._process_cpu_cluster_args(offline_cluster_opts)
-        if self.ctxt.get_ctxt('cpuClusterProxy') is None:
-            # if no cpu-cluster is defined, then we are not supposed to run cost calculations
-            enable_savings_flag = False
-        else:
-            # if no gpu-cluster is defined, then we are not supposed to run cost calculations
-            enable_savings_flag = self._process_gpu_cluster_args(offline_cluster_opts)
+        # read the wrapper option defined by the spark_rapids cmd if any.
+        enable_savings_flag = self.wrapper_options.get('savingsCalculations', True)
+        if enable_savings_flag:
+            offline_cluster_opts = self.wrapper_options.get('migrationClustersProps', {})
+            self._process_cpu_cluster_args(offline_cluster_opts)
+            if self.ctxt.get_ctxt('cpuClusterProxy') is None:
+                # if no cpu-cluster is defined, then we are not supposed to run cost calculations
+                enable_savings_flag = False
+            else:
+                # if no gpu-cluster is defined, then we are not supposed to run cost calculations
+                enable_savings_flag = self._process_gpu_cluster_args(offline_cluster_opts)
+
         self._set_savings_calculations_flag(enable_savings_flag)
 
     def _set_savings_calculations_flag(self, enable_flag: bool):

user_tools/src/spark_rapids_pytools/wrappers/databricks_aws_wrapper.py

@@ -37,7 +37,7 @@ def qualification(cpu_cluster: str = None,
                       gpu_cluster: str = None,
                       tools_jar: str = None,
                       credentials_file: str = None,
-                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.SAVINGS),
+                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.get_default()),
                       gpu_cluster_recommendation: str = QualGpuClusterReshapeType.tostring(
                           QualGpuClusterReshapeType.get_default()),
                       estimation_model: str = None,
@@ -81,12 +81,12 @@ def qualification(cpu_cluster: str = None,
                to provide the location of a credential file. The default credentials file exists as
                "~/.databrickscfg" on Unix, Linux, or macOS
         :param filter_apps: filtering criteria of the applications listed in the final STDOUT table
-                is one of the following (ALL, SPEEDUPS, savings, top_candidates).
+                is one of the following (all, speedups, savings, top_candidates).
                 Note that this filter does not affect the CSV report.
-                "ALL" means no filter applied. "SPEEDUPS" lists all the apps that are either
-                'Recommended', or 'Strongly Recommended' based on speedups. "SAVINGS"
+                "all" means no filter applied. "speedups" lists all the apps that are either
+                'Recommended', or 'Strongly Recommended' based on speedups. "savings"
                 lists all the apps that have positive estimated GPU savings except for the apps that
-                are "Not Applicable". "TOP_CANDIDATES" lists all apps that have unsupported operators
+                are "Not Applicable". "top_candidates" lists all apps that have unsupported operators
                 stage duration less than 25% of app duration and speedups greater than 1.3x.
         :param gpu_cluster_recommendation: The type of GPU cluster recommendation to generate.
                It accepts one of the following ("CLUSTER", "JOB" and the default value "MATCH").

user_tools/src/spark_rapids_pytools/wrappers/databricks_azure_wrapper.py

@@ -36,7 +36,7 @@ def qualification(cpu_cluster: str = None,
                       gpu_cluster: str = None,
                       tools_jar: str = None,
                       credentials_file: str = None,
-                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.SAVINGS),
+                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.get_default()),
                       gpu_cluster_recommendation: str = QualGpuClusterReshapeType.tostring(
                           QualGpuClusterReshapeType.get_default()),
                       estimation_model: str = None,
@@ -79,12 +79,12 @@ def qualification(cpu_cluster: str = None,
                to provide the location of a credential file. The default credentials file exists as
                "~/.databrickscfg" on Unix, Linux, or macOS
         :param filter_apps: filtering criteria of the applications listed in the final STDOUT table
-                is one of the following (ALL, SPEEDUPS, SAVINGS, TOP_CANDIDATES).
+                is one of the following (all, speedups, savings, top_candidates).
                 Note that this filter does not affect the CSV report.
-                "ALL" means no filter applied. "SPEEDUPS" lists all the apps that are either
-                'Recommended', or 'Strongly Recommended' based on speedups. "SAVINGS"
+                "all" means no filter applied. "speedups" lists all the apps that are either
+                'Recommended', or 'Strongly Recommended' based on speedups. "savings"
                 lists all the apps that have positive estimated GPU savings except for the apps that
-                are "Not Applicable". "TOP_CANDIDATES" lists all apps that have unsupported operators
+                are "Not Applicable". "top_candidates" lists all apps that have unsupported operators
                 stage duration less than 25% of app duration and speedups greater than 1.3x.
         :param gpu_cluster_recommendation: The type of GPU cluster recommendation to generate.
                It accepts one of the following ("CLUSTER", "JOB" and the default value "MATCH").

user_tools/src/spark_rapids_pytools/wrappers/dataproc_gke_wrapper.py

@@ -33,7 +33,7 @@ def qualification(cpu_cluster: str = None,
                       gpu_cluster: str = None,
                       tools_jar: str = None,
                       credentials_file: str = None,
-                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.SAVINGS),
+                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.get_default()),
                       gpu_cluster_recommendation: str = QualGpuClusterReshapeType.tostring(
                           QualGpuClusterReshapeType.get_default()),
                       estimation_model: str = None,
@@ -76,12 +76,12 @@ def qualification(cpu_cluster: str = None,
                to provide the location of a credential JSON file. The default credentials file exists as
                "$HOME/.config/gcloud/application_default_credentials.json"
         :param filter_apps: filtering criteria of the applications listed in the final STDOUT table
-                is one of the following (ALL, SPEEDUPS, savings, top_candidates).
+                is one of the following (all, speedups, savings, top_candidates).
                 Note that this filter does not affect the CSV report.
-                "ALL" means no filter applied. "SPEEDUPS" lists all the apps that are either
-                'Recommended', or 'Strongly Recommended' based on speedups. "SAVINGS"
+                "all" means no filter applied. "speedups" lists all the apps that are either
+                'Recommended', or 'Strongly Recommended' based on speedups. "savings"
                 lists all the apps that have positive estimated GPU savings except for the apps that
-                are "Not Applicable". "TOP_CANDIDATES" lists all apps that have unsupported operators
+                are "Not Applicable". "top_candidates" lists all apps that have unsupported operators
                 stage duration less than 25% of app duration and speedups greater than 1.3x.
         :param gpu_cluster_recommendation: The type of GPU cluster recommendation to generate.
                It accepts one of the following ("CLUSTER", "JOB" and the default value "MATCH").

user_tools/src/spark_rapids_pytools/wrappers/dataproc_wrapper.py

@@ -36,7 +36,7 @@ def qualification(cpu_cluster: str = None,
                       gpu_cluster: str = None,
                       tools_jar: str = None,
                       credentials_file: str = None,
-                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.SAVINGS),
+                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.get_default()),
                       gpu_cluster_recommendation: str = QualGpuClusterReshapeType.tostring(
                           QualGpuClusterReshapeType.get_default()),
                       estimation_model: str = None,
@@ -78,12 +78,12 @@ def qualification(cpu_cluster: str = None,
                to provide the location of a credential JSON file. The default credentials file exists as
                "$HOME/.config/gcloud/application_default_credentials.json"
         :param filter_apps: filtering criteria of the applications listed in the final STDOUT table
-                is one of the following (ALL, SPEEDUPS, savings, top_candidates).
+                is one of the following (all, speedups, savings, top_candidates).
                 Note that this filter does not affect the CSV report.
-                "ALL" means no filter applied. "SPEEDUPS" lists all the apps that are either
-                'Recommended', or 'Strongly Recommended' based on speedups. "SAVINGS"
+                "all" means no filter applied. "speedups" lists all the apps that are either
+                'Recommended', or 'Strongly Recommended' based on speedups. "savings"
                 lists all the apps that have positive estimated GPU savings except for the apps that
-                are "Not Applicable". "TOP_CANDIDATES" lists all apps that have unsupported operators
+                are "Not Applicable". "top_candidates" lists all apps that have unsupported operators
                 stage duration less than 25% of app duration and speedups greater than 1.3x.
         :param gpu_cluster_recommendation: The type of GPU cluster recommendation to generate.
                It accepts one of the following ("CLUSTER", "JOB" and the default value "MATCH").

user_tools/src/spark_rapids_pytools/wrappers/emr_wrapper.py

@@ -37,7 +37,7 @@ def qualification(cpu_cluster: str = None,
                       remote_folder: str = None,
                       gpu_cluster: str = None,
                       tools_jar: str = None,
-                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.SAVINGS),
+                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.get_default()),
                       gpu_cluster_recommendation: str = QualGpuClusterReshapeType.tostring(
                           QualGpuClusterReshapeType.get_default()),
                       estimation_model: str = None,
@@ -76,12 +76,12 @@ def qualification(cpu_cluster: str = None,
                 or remote S3 url. If missing, the wrapper downloads the latest rapids-4-spark-tools_*.jar
                 from maven repo
         :param filter_apps: filtering criteria of the applications listed in the final STDOUT table
-                is one of the following (ALL, SPEEDUPS, SAVINGS, TOP_CANDIDATES). Default is "SAVINGS".
+                is one of the following (all, speedups, savings, top_candidates).
                 Note that this filter does not affect the CSV report.
-                "ALL" means no filter applied. "SPEEDUPS" lists all the apps that are either
-                'Recommended', or 'Strongly Recommended' based on speedups. "SAVINGS"
+                "all" means no filter applied. "speedups" lists all the apps that are either
+                'Recommended', or 'Strongly Recommended' based on speedups. "savings"
                 lists all the apps that have positive estimated GPU savings except for the apps that
-                are "Not Applicable". "TOP_CANDIDATES" lists all apps that have unsupported operators
+                are "Not Applicable". "top_candidates" lists all apps that have unsupported operators
                 stage duration less than 25% of app duration and speedups greater than 1.3x.
         :param gpu_cluster_recommendation: The type of GPU cluster recommendation to generate.
                It accepts one of the following ("CLUSTER", "JOB" and the default value "MATCH").

user_tools/src/spark_rapids_pytools/wrappers/onprem_wrapper.py

@@ -32,7 +32,7 @@ def qualification(cpu_cluster: str = None,
                       eventlogs: str = None,
                       local_folder: str = None,
                       tools_jar: str = None,
-                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.SPEEDUPS),
+                      filter_apps: str = QualFilterApp.tostring(QualFilterApp.get_default()),
                       target_platform: str = None,
                       gpu_cluster_recommendation: str = QualGpuClusterReshapeType.tostring(
                           QualGpuClusterReshapeType.get_default()),
@@ -55,11 +55,14 @@ def qualification(cpu_cluster: str = None,
                 directory for temporary folders/files. The final output will go into a subdirectory
                 named `qual-${EXEC_ID}` where `exec_id` is an auto-generated unique identifier of the execution.
         :param tools_jar: Path to a bundled jar including RAPIDS tool. The path is a local filesystem path
-        :param filter_apps:  Filtering criteria of the applications listed in the final STDOUT table is one of
-                the following (`ALL`, `SPEEDUPS`, `TOP_CANDIDATES`). "`ALL`" means no filter applied. "`SPEEDUPS`"
-                lists all the apps that are either '_Recommended_', or '_Strongly Recommended_' based on speedups.
-                "`TOP_CANDIDATES`" lists all apps that have unsupported operators stage duration less than 25% of
-                app duration and speedups greater than 1.3x.
+        :param filter_apps: filtering criteria of the applications listed in the final STDOUT table
+                is one of the following (all, speedups, savings, top_candidates).
+                Note that this filter does not affect the CSV report.
+                "all" means no filter applied. "speedups" lists all the apps that are either
+                'Recommended', or 'Strongly Recommended' based on speedups. "savings"
+                lists all the apps that have positive estimated GPU savings except for the apps that
+                are "Not Applicable". "top_candidates" lists all apps that have unsupported operators
+                stage duration less than 25% of app duration and speedups greater than 1.3x.
         :param target_platform: Cost savings and speedup recommendation for comparable cluster in target_platform
                 based on on-premises cluster configuration. Currently only `dataproc` is supported for
                 target_platform.If not provided, the final report will be limited to GPU speedups only
@@ -102,7 +105,7 @@ def qualification(cpu_cluster: str = None,
                 if cpu_cluster is None:
                     raise RuntimeError('OnPrem\'s cluster property file required to calculate'
                                        'savings for ' + target_platform + ' platform')
-                filter_apps: str = QualFilterApp.tostring(QualFilterApp.SAVINGS)
+                # filter_apps: str = QualFilterApp.tostring(QualFilterApp.SAVINGS)
             else:
                 raise RuntimeError(target_platform + ' platform is currently not supported to calculate savings'
                                    ' from OnPrem cluster')

user_tools/src/spark_rapids_tools/cmdli/argprocessor.py

@@ -334,13 +334,17 @@ class QualifyUserArgModel(ToolUserArgModel):
     def init_tool_args(self):
         self.p_args['toolArgs']['platform'] = self.platform
         self.p_args['toolArgs']['savingsCalculations'] = True
-        self.p_args['toolArgs']['filterApps'] = self.filter_apps
         self.p_args['toolArgs']['targetPlatform'] = self.target_platform
         self.p_args['toolArgs']['cpuClusterPrice'] = self.cpu_cluster_price
         self.p_args['toolArgs']['estimatedGpuClusterPrice'] = self.estimated_gpu_cluster_price
         self.p_args['toolArgs']['cpuDiscount'] = self.cpu_discount
         self.p_args['toolArgs']['gpuDiscount'] = self.gpu_discount
         self.p_args['toolArgs']['globalDiscount'] = self.global_discount
+        # check the filter_apps argument
+        if self.filter_apps is None:
+            self.p_args['toolArgs']['filterApps'] = QualFilterApp.get_default()
+        else:
+            self.p_args['toolArgs']['filterApps'] = self.filter_apps
         # check the reshapeType argument
         if self.gpu_cluster_recommendation is None:
             self.p_args['toolArgs']['gpuClusterRecommendation'] = QualGpuClusterReshapeType.get_default()
@@ -406,11 +410,9 @@ def build_tools_args(self) -> dict:
                                                      '"target_platform" argument to generate cost savings')
 
         # check the filter_apps argument
-        if self.p_args['toolArgs']['filterApps'] is None:
-            # set a default filterApps argument to be savings if the cost savings is enabled
-            if self.p_args['toolArgs']['savingsCalculations']:
-                self.p_args['toolArgs']['filterApps'] = QualFilterApp.SAVINGS
-            else:
+        if not self.p_args['toolArgs']['savingsCalculations']:
+            # if savingsCalculations is disabled, we cannot use savings filter
+            if self.p_args['toolArgs']['filterApps'] == QualFilterApp.SAVINGS:
                 self.p_args['toolArgs']['filterApps'] = QualFilterApp.SPEEDUPS
 
         # finally generate the final values

user_tools/src/spark_rapids_tools/enums.py

@@ -118,7 +118,7 @@ class QualFilterApp(EnumeratedType):
 
     @classmethod
     def get_default(cls):
-        return cls.SAVINGS
+        return cls.TOP_CANDIDATES
 
 
 class QualGpuClusterReshapeType(EnumeratedType):

user_tools/tests/spark_rapids_tools_ut/test_tool_argprocessor.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -66,14 +66,14 @@ def validate_args_w_savings_enabled(tool_name: str, t_args: dict):
         if tool_name == 'qualification':
             assert t_args['savingsCalculations']
             # filterApps should be set to savings
-            assert t_args['filterApps'] == QualFilterApp.SAVINGS
+            assert t_args['filterApps'] == QualFilterApp.get_default()
 
     @staticmethod
     def validate_args_w_savings_disabled(tool_name: str, t_args: dict):
         if tool_name == 'qualification':
             assert not t_args['savingsCalculations']
             # filterApps should be set to savings
-            assert t_args['filterApps'] == QualFilterApp.SPEEDUPS
+            assert t_args['filterApps'] != QualFilterApp.SAVINGS
 
     @staticmethod
     def create_tool_args_should_pass(tool_name: str, platform=None, cluster=None, eventlogs=None):