[CI] Parity: fix upstream job links

.automation_scripts/pytorch-unit-test-scripts/download_testlogs

@@ -332,6 +332,7 @@ def _shorten_unzipped_dirs():
     failing job from the parity summary.
     """
     from pathlib import Path
+    job_ids = []
     for d in sorted(Path(".").glob("unzipped-*")):
         if not d.is_dir():
             continue
@@ -346,11 +347,39 @@ def _shorten_unzipped_dirs():
             job_id_match = re.search(r'_(\d{6,})(?:\.zip)?$', d.name)
             if job_id_match:
                 short_name += f"_{job_id_match.group(1)}"
+                job_ids.append(job_id_match.group(1))
             if not Path(short_name).exists():
                 d.rename(short_name)
                 print(f"  Renamed {d.name} -> {short_name}")
             else:
                 print(f"  WARNING: {short_name} already exists, keeping {d.name}")
+    return job_ids
+
+def _record_shard_run_ids(job_ids, workflow_run_id):
+    """Persist a job_id -> upstream run_id map ("_wf_run_ids.json") next to the
+    shard dirs so summarize_xml_testreports.py can build a correct job URL for
+    each shard.
+
+    The default/distributed/inductor configs resolve to DIFFERENT upstream runs
+    (e.g. mi350 default+inductor come from a trunk push while distributed comes
+    from periodic / a trunk fallback), yet they all unzip into the same rocm_xml
+    folder. The legacy single "_wf_run_id" (last write wins) therefore mislabels
+    shards from the other runs, producing "busted" job URLs that point at a run
+    the job never belonged to. Recording the run id per job id fixes that.
+    """
+    if not job_ids:
+        return
+    mapping = {}
+    if os.path.isfile("_wf_run_ids.json"):
+        try:
+            with open("_wf_run_ids.json") as f:
+                mapping = json.load(f)
+        except Exception:
+            mapping = {}
+    for jid in job_ids:
+        mapping[str(jid)] = str(workflow_run_id)
+    with open("_wf_run_ids.json", "w") as f:
+        json.dump(mapping, f)
 
 def download_xml_files(workflow_run_id, workflow_run_attempts, prefixes=[], allowed_substrings=None):
     # Get from S3 artifacts.
@@ -410,8 +439,12 @@ def download_xml_files(workflow_run_id, workflow_run_attempts, prefixes=[], allo
     for path in artifact_paths:
             unzip(path)
 
-    _shorten_unzipped_dirs()
+    job_ids = _shorten_unzipped_dirs()
+    # Record which upstream run each shard's job id came from so job URLs are
+    # built per-shard (different configs come from different runs).
+    _record_shard_run_ids(job_ids, workflow_run_id)
 
+    # Kept for backward compatibility with older summarize layouts.
     with open("_wf_run_id", "w") as f:
         f.write(str(workflow_run_id))
 
@@ -526,7 +559,7 @@ def parse_args():
     parser.add_argument('--no_rocm', action='store_true')
     parser.add_argument('--no_cuda', action='store_true')
     parser.add_argument('--pr_id', type=int, help='The pull request ID')
-    parser.add_argument('--arch', type=str, choices=['mi200', 'mi300', 'mi350', 'navi31', 'nightly'], default='mi350', help='ROCm GPU architecture (mi200, mi300, mi350, navi31, or nightly, default: mi350)')
+    parser.add_argument('--arch', type=str, choices=['mi200', 'mi300', 'mi350', 'navi31', 'preview'], default='mi350', help='ROCm GPU architecture (mi200, mi300, mi350, navi31, or preview, default: mi350)')
     parser.add_argument('--include_inductor_periodic', action='store_true', help='Also download inductor-periodic benchmark artifacts (into a separate directory, not included in parity CSV)')
     parser.add_argument('--baseline_sha', type=str, help='Baseline commit SHA to compare against. Downloads the same ROCm workflows for this commit into baseline_xml/.')
     return parser.parse_args()
@@ -615,7 +648,7 @@ def main():
     # in, job-name prefixes, shard counts and fallbacks) comes from
     # parity_job_config.json - see PARITY_CONFIG at the top of this file.
     global ROCmWorkflowNames
-    arch = args.arch  # 'mi200', 'mi300', 'mi350', 'navi31', or 'nightly'
+    arch = args.arch  # 'mi200', 'mi300', 'mi350', 'navi31', or 'preview'
     arch_config = PARITY_CONFIG["rocm"][arch]
 
     ROCmWorkflowNames, rocm_job_prefix, arch_fallbacks = parity_config_views(arch_config)

.automation_scripts/pytorch-unit-test-scripts/parity_job_config.json

@@ -38,7 +38,7 @@
       "default": [{ "workflow": "rocm-mi300", "job_prefix": "linux-noble-rocm-py3.12-mi300" }],
       "distributed": [{ "workflow": "periodic-rocm-mi300", "job_prefix": "linux-noble-rocm-py3.12-mi300" }],
       "inductor": [{ "workflow": "inductor-rocm-mi300", "job_prefix": "linux-noble-rocm-py3.12-mi300" }],
-      "shard_counts": { "default": 6, "distributed": 3, "inductor": 2 },
+      "shard_counts": { "default": 8, "distributed": 3, "inductor": 2 },
       "checkrun_regex": "rocm.*mi300.*/ test [(](default|distributed|inductor),"
     },
     "mi200": {
@@ -64,12 +64,12 @@
       "shard_counts": { "default": 2, "distributed": 3, "inductor": 2 },
       "checkrun_regex": "rocm.*navi31.*/ test [(]default,"
     },
-    "nightly": {
-      "default": [{ "workflow": "rocm-nightly", "job_prefix": "linux-noble-rocm-nightly-py3.12-mi350" }],
-      "distributed": [{ "workflow": "rocm-nightly", "job_prefix": "linux-noble-rocm-nightly-py3.12-mi350" }],
-      "inductor": [{ "workflow": "rocm-nightly", "job_prefix": "linux-noble-rocm-nightly-py3.12-mi350" }],
-      "shard_counts": { "default": 6, "distributed": 3, "inductor": 2 },
-      "checkrun_regex": "rocm-nightly.*/ test [(](default|distributed|inductor),"
+    "preview": {
+      "default": [{ "workflow": "rocm-preview", "job_prefix": "linux-noble-rocm-preview-py3.12-mi350" }],
+      "distributed": [{ "workflow": "rocm-preview", "job_prefix": "linux-noble-rocm-preview-py3.12-mi350" }],
+      "inductor": [{ "workflow": "rocm-preview", "job_prefix": "linux-noble-rocm-preview-py3.12-mi350" }],
+      "shard_counts": { "default": 8, "distributed": 3, "inductor": 2 },
+      "checkrun_regex": "rocm-preview.*/ test [(](default|distributed|inductor),"
     }
   }
 }

.automation_scripts/pytorch-unit-test-scripts/summarize_xml_testreports.py

@@ -2,6 +2,7 @@
 
 import argparse
 import csv
+import json
 import os
 import re
 import pandas as pd
@@ -73,11 +74,27 @@ def parse_xml_reports_as_dict(workflow_run_id, workflow_run_attempt, tag, path="
     test_config = ""
     test_cases = {}
 
-    # download_testlogs writes the upstream pytorch CI workflow run id
-    # into "_wf_run_id" alongside the shard dirs. We combine it with each
-    # shard dir's trailing "_<job_id>" to form the URL
-    # https://github.com/pytorch/pytorch/actions/runs/<wf>/job/<job_id>
+    # download_testlogs records the upstream pytorch CI run id for each shard's
+    # job id in "_wf_run_ids.json" (job_id -> run_id) alongside the shard dirs,
+    # so we can build the URL
+    # https://github.com/pytorch/pytorch/actions/runs/<run_id>/job/<job_id>
     # surfaced as the "Job ID" column in the FAILED TESTS table.
+    #
+    # A single run id is NOT enough: the default/distributed/inductor configs
+    # resolve to DIFFERENT upstream runs (e.g. mi350 default+inductor come from
+    # a trunk push while distributed comes from periodic/a trunk fallback), so
+    # pairing every shard's job id with one run id produced links to a run the
+    # job never belonged to (a "busted" URL that 404s). We look each job id up
+    # in the per-job map and fall back to the legacy single "_wf_run_id" file
+    # only when the map has no entry (older downloads / backward compat).
+    run_id_by_job = {}
+    run_ids_file = os.path.join(path, "_wf_run_ids.json")
+    if os.path.isfile(run_ids_file):
+        try:
+            with open(run_ids_file) as f:
+                run_id_by_job = {str(k): str(v) for k, v in json.load(f).items()}
+        except Exception:
+            run_id_by_job = {}
     wf_run_id = ""
     wf_id_file = os.path.join(path, "_wf_run_id")
     if os.path.isfile(wf_id_file):
@@ -96,9 +113,13 @@ def parse_xml_reports_as_dict(workflow_run_id, workflow_run_attempt, tag, path="
                 test_config = TestConfigName.inductor.name
             shard = _extract_shard(dir)
             jid = re.search(r'_(\d+)$', dir)
+            job_id = jid.group(1) if jid else ""
+            # Prefer the run id recorded for this specific job; fall back to the
+            # single legacy run id so we never regress older download layouts.
+            shard_run_id = run_id_by_job.get(job_id, wf_run_id)
             job_url = (
-                f"https://github.com/pytorch/pytorch/actions/runs/{wf_run_id}/job/{jid.group(1)}"
-                if wf_run_id and jid else ""
+                f"https://github.com/pytorch/pytorch/actions/runs/{shard_run_id}/job/{job_id}"
+                if shard_run_id and job_id else ""
             )
             for xml_report in Path(new_dir).glob("**/*.xml"):
                 try:

.github/workflows/parity.yml

@@ -31,7 +31,7 @@ on:
         required: false
         type: string
       arch:
-        description: 'ROCm architectures, comma or space separated. Options: mi350, mi300, mi200, nightly, navi31. Example: "nightly, mi350" or "mi300"'
+        description: 'ROCm architectures, comma or space separated. Options: mi350, mi300, mi200, navi31, preview. Example: "preview, mi350" or "mi300"'
         required: false
         default: 'mi350, mi300, mi200'
         type: string
@@ -72,7 +72,7 @@ on:
         type: boolean
       # summarize_xml_testreports flags
       set1_name:
-        description: 'Label for ROCm columns in output CSV. Examples: rocm, nightly, mi300. Default: rocm'
+        description: 'Label for ROCm columns in output CSV. Examples: rocm, preview, mi300. Default: rocm'
         required: false
         default: 'rocm'
         type: string