From b1375177fcc89b81495889c71dcffd9eceffa7a8 Mon Sep 17 00:00:00 2001
From: Yih-Dar <2521628+ydshieh@users.noreply.github.com>
Date: Tue, 6 May 2025 15:10:29 +0200
Subject: [PATCH] add job links to new model failure report (#37973)

* update for job link

* stye

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
---
 utils/notification_service.py      | 22 ++++++++++++++++++++++
 utils/process_bad_commit_report.py |  8 ++++++++
 2 files changed, 30 insertions(+)

diff --git a/utils/notification_service.py b/utils/notification_service.py
index ea622d6091..399c792c9d 100644
--- a/utils/notification_service.py
+++ b/utils/notification_service.py
@@ -1260,6 +1260,28 @@ if __name__ == "__main__":
                 token=os.environ.get("TRANSFORMERS_CI_RESULTS_UPLOAD_TOKEN", None),
             )
 
+        # Let's create a file contain job --> job link
+        model_job_links = {}
+        sorted_dict = sorted(model_results.items(), key=lambda t: t[0])
+        for job, job_result in sorted_dict:
+            model_name = job
+            if model_name.startswith("models_"):
+                model_name = model_name[len("models_") :]
+            model_job_links[model_name] = job_result["job_link"]
+
+        with open(f"ci_results_{job_name}/model_job_links.json", "w", encoding="UTF-8") as fp:
+            json.dump(model_job_links, fp, indent=4, ensure_ascii=False)
+
+        # upload results to Hub dataset (only for the scheduled daily CI run on `main`)
+        if is_scheduled_ci_run:
+            api.upload_file(
+                path_or_fileobj=f"ci_results_{job_name}/model_job_links.json",
+                path_in_repo=f"{datetime.datetime.today().strftime('%Y-%m-%d')}/ci_results_{job_name}/model_job_links.json",
+                repo_id="hf-internal-testing/transformers_daily_ci",
+                repo_type="dataset",
+                token=os.environ.get("TRANSFORMERS_CI_RESULTS_UPLOAD_TOKEN", None),
+            )
+
     # Must have the same keys as in `additional_results`.
     # The values are used as the file names where to save the corresponding CI job results.
     test_to_result_name = {
diff --git a/utils/process_bad_commit_report.py b/utils/process_bad_commit_report.py
index 19812ff21f..bba03b4bd1 100644
--- a/utils/process_bad_commit_report.py
+++ b/utils/process_bad_commit_report.py
@@ -27,6 +27,9 @@ if __name__ == "__main__":
     with open("new_model_failures_with_bad_commit.json") as fp:
         data = json.load(fp)
 
+    with open("ci_results_run_models_gpu/model_job_links.json") as fp:
+        model_job_links = json.load(fp)
+
     # TODO: extend
     team_members = [
         "ydshieh",
@@ -62,7 +65,12 @@ if __name__ == "__main__":
     for author, _data in new_data_full.items():
         for model, model_result in _data.items():
             for device, failed_tests in model_result.items():
+                # prepare job_link and add it to each entry of new failed test information.
+                # need to change from `single-gpu` to `single` and same for `multi-gpu` to match `job_link`.
+                job_link = model_job_links[model][device.replace("-gpu", "")]
                 failed_tests = [x for x in failed_tests if x["author"] == author or x["merged_by"] == author]
+                for x in failed_tests:
+                    x.update({"job_link": job_link})
                 model_result[device] = failed_tests
             _data[model] = {k: v for k, v in model_result.items() if len(v) > 0}
         new_data_full[author] = {k: v for k, v in _data.items() if len(v) > 0}