From 0cfd89980e7b7b17cbf59e1d952cfe52b685933e Mon Sep 17 00:00:00 2001
From: Codex <codex@local>
Date: Mon, 29 Jun 2026 21:21:09 +0300
Subject: [PATCH] Clarify transient training progress upload errors

---
 tools/windows_training_agent.py | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/tools/windows_training_agent.py b/tools/windows_training_agent.py
index 16147be..6290c88 100644
--- a/tools/windows_training_agent.py
+++ b/tools/windows_training_agent.py
@@ -237,10 +237,16 @@ def safe_report_progress(
     message: str,
     log_path: Path,
 ) -> None:
-    try:
-        report_progress(args, job_id, status, phase, progress_percent, message)
-    except Exception as exc:  # noqa: BLE001 - keep the local training process alive.
-        log(log_path, f"Progress report failed: {exc}")
+    last_error: Exception | None = None
+    for attempt in range(1, 4):
+        try:
+            report_progress(args, job_id, status, phase, progress_percent, message)
+            return
+        except Exception as exc:  # noqa: BLE001 - keep the local training process alive.
+            last_error = exc
+            if attempt < 3:
+                time.sleep(attempt * 2)
+    log(log_path, f"Temporary progress upload error; training continues: {last_error}")
 
 
 def api_json(args: argparse.Namespace, path: str, payload: dict[str, Any], timeout: int = 30) -> dict[str, Any]: