Display deltas for cudf.pandas test summary (#16864)

This PR displays delta's for CPU and GPU usage metrics that are extracted from `cudf.pandas` pytests.

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - Jake Awe (https://github.com/AyodeAwe)

URL: #16864

.github/workflows/pr.yaml

@@ -50,6 +50,7 @@ jobs:
       test_java: ${{ steps.changed-files.outputs.java_any_changed == 'true' }}
       test_notebooks: ${{ steps.changed-files.outputs.notebooks_any_changed == 'true' }}
       test_python: ${{ steps.changed-files.outputs.python_any_changed == 'true' }}
+      test_cudf_pandas: ${{ steps.changed-files.outputs.cudf_pandas_any_changed == 'true' }}
     steps:
       - name: Get PR info
         id: get-pr-info
@@ -82,6 +83,7 @@ jobs:
               - '!java/**'
               - '!notebooks/**'
               - '!python/**'
+              - '!ci/cudf_pandas_scripts/**'
             java:
               - '**'
               - '!CONTRIBUTING.md'
@@ -90,11 +92,13 @@ jobs:
               - '!img/**'
               - '!notebooks/**'
               - '!python/**'
+              - '!ci/cudf_pandas_scripts/**'
             notebooks:
               - '**'
               - '!CONTRIBUTING.md'
               - '!README.md'
               - '!java/**'
+              - '!ci/cudf_pandas_scripts/**'
             python:
               - '**'
               - '!CONTRIBUTING.md'
@@ -103,6 +107,16 @@ jobs:
               - '!img/**'
               - '!java/**'
               - '!notebooks/**'
+              - '!ci/cudf_pandas_scripts/**'
+            cudf_pandas:
+              - '**'
+              - 'ci/cudf_pandas_scripts/**'
+              - '!CONTRIBUTING.md'
+              - '!README.md'
+              - '!docs/**'
+              - '!img/**'
+              - '!java/**'
+              - '!notebooks/**'
   checks:
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/[email protected]
@@ -289,7 +303,7 @@ jobs:
     needs: [wheel-build-cudf, changed-files]
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/[email protected]
-    if: needs.changed-files.outputs.test_python == 'true'
+    if: needs.changed-files.outputs.test_python == 'true' || needs.changed-files.outputs.test_cudf_pandas == 'true'
     with:
       # This selects "ARCH=amd64 + the latest supported Python + CUDA".
       matrix_filter: map(select(.ARCH == "amd64")) | group_by(.CUDA_VER|split(".")|map(tonumber)|.[0]) | map(max_by([(.PY_VER|split(".")|map(tonumber)), (.CUDA_VER|split(".")|map(tonumber))]))
@@ -300,7 +314,7 @@ jobs:
     needs: [wheel-build-cudf, changed-files]
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/[email protected]
-    if: needs.changed-files.outputs.test_python == 'true'
+    if: needs.changed-files.outputs.test_python == 'true' || needs.changed-files.outputs.test_cudf_pandas == 'true'
     with:
       # This selects "ARCH=amd64 + the latest supported Python + CUDA".
       matrix_filter: map(select(.ARCH == "amd64")) | group_by(.CUDA_VER|split(".")|map(tonumber)|.[0]) | map(max_by([(.PY_VER|split(".")|map(tonumber)), (.CUDA_VER|split(".")|map(tonumber))]))

ci/cudf_pandas_scripts/pandas-tests/job-summary.py

@@ -67,20 +67,33 @@ def emoji_failed(x):
 # convert pr_results to a pandas DataFrame and then a markdown table
 pr_df = pd.DataFrame.from_dict(pr_results, orient="index").sort_index()
 main_df = pd.DataFrame.from_dict(main_results, orient="index").sort_index()
-diff_df = pr_df - main_df
-total_usage = pr_df['_slow_function_call'] + pr_df['_fast_function_call']
-pr_df['CPU Usage'] = ((pr_df['_slow_function_call']/total_usage)*100.0).round(1)
-pr_df['GPU Usage'] = ((pr_df['_fast_function_call']/total_usage)*100.0).round(1)
+total_usage = main_df["_slow_function_call"] + main_df["_fast_function_call"]
+main_df["CPU Usage"] = ((main_df["_slow_function_call"] / total_usage) * 100.0).round(1)
+main_df["GPU Usage"] = ((main_df["_fast_function_call"] / total_usage) * 100.0).round(1)
+
+total_usage = pr_df["_slow_function_call"] + pr_df["_fast_function_call"]
+pr_df["CPU Usage"] = ((pr_df["_slow_function_call"] / total_usage) * 100.0).round(1)
+pr_df["GPU Usage"] = ((pr_df["_fast_function_call"] / total_usage) * 100.0).round(1)
+
+cpu_usage_mean = pr_df["CPU Usage"].mean().round(2)
+gpu_usage_mean = pr_df["GPU Usage"].mean().round(2)
+
+gpu_usage_rate_change = abs(pr_df["GPU Usage"].mean() - main_df["GPU Usage"].mean())
+pr_df["CPU Usage"] = pr_df["CPU Usage"].fillna(0)
+pr_df["GPU Usage"] = pr_df["GPU Usage"].fillna(0)
+main_df["CPU Usage"] = main_df["CPU Usage"].fillna(0)
+main_df["GPU Usage"] = main_df["GPU Usage"].fillna(0)
 
-cpu_usage_mean = pr_df['CPU Usage'].mean().round(2)
-gpu_usage_mean = pr_df['GPU Usage'].mean().round(2)
+diff_df = pr_df - main_df
+diff_df["CPU Usage"] = diff_df["CPU Usage"].round(1).fillna(0)
+diff_df["GPU Usage"] = diff_df["GPU Usage"].round(1).fillna(0)
 
-# Add '%' suffix to 'CPU Usage' and 'GPU Usage' columns
-pr_df['CPU Usage'] = pr_df['CPU Usage'].fillna(0).astype(str) + '%'
-pr_df['GPU Usage'] = pr_df['GPU Usage'].fillna(0).astype(str) + '%'
+# Add '%' suffix to "CPU Usage" and "GPU Usage" columns
+pr_df["CPU Usage"] = pr_df["CPU Usage"].astype(str) + "%"
+pr_df["GPU Usage"] = pr_df["GPU Usage"].astype(str) + "%"
 
-pr_df = pr_df[["total", "passed", "failed", "skipped", 'CPU Usage', 'GPU Usage']]
-diff_df = diff_df[["total", "passed", "failed", "skipped"]]
+pr_df = pr_df[["total", "passed", "failed", "skipped", "CPU Usage", "GPU Usage"]]
+diff_df = diff_df[["total", "passed", "failed", "skipped", "CPU Usage", "GPU Usage"]]
 diff_df.columns = diff_df.columns + "_diff"
 diff_df["passed_diff"] = diff_df["passed_diff"].map(emoji_passed)
 diff_df["failed_diff"] = diff_df["failed_diff"].map(emoji_failed)
@@ -99,13 +112,36 @@ def emoji_failed(x):
         "passed_diff": "Passed delta",
         "failed_diff": "Failed delta",
         "skipped_diff": "Skipped delta",
+        "CPU Usage_diff": "CPU Usage delta",
+        "GPU Usage_diff": "GPU Usage delta",
     }
 )
-df = df.sort_values(by=["Failed tests", "Skipped tests"], ascending=False)
-
+df = df.sort_values(by=["CPU Usage delta", "Total tests"], ascending=False)
+df["CPU Usage delta"] = df["CPU Usage delta"].map(emoji_failed)
+df["GPU Usage delta"] = df["GPU Usage delta"].map(emoji_passed)
+df = df[
+    [
+        "Total tests",
+        "CPU Usage delta",
+        "GPU Usage delta",
+        "Passed tests",
+        "Failed tests",
+        "Skipped tests",
+        "CPU Usage",
+        "GPU Usage",
+        "Total delta",
+        "Passed delta",
+        "Failed delta",
+        "Skipped delta",
+    ]
+]
 print(comment)
 print()
-print(f"Average CPU and GPU usage for the tests: {cpu_usage_mean}% and {gpu_usage_mean}%")
+print(
+    f"Average GPU usage: {gpu_usage_mean}% {'an increase' if gpu_usage_rate_change > 0 else 'a decrease'} by {gpu_usage_rate_change}%"
+)
+print()
+print(f"Average CPU usage: {cpu_usage_mean}%")
 print()
 print("Here are the results of running the Pandas tests against this PR:")
 print()