From 8dec5b59767eded07e23a795b6e03581c1c7652a Mon Sep 17 00:00:00 2001
From: Marcel <marcel@familienarchiv>
Date: Mon, 13 Apr 2026 20:58:32 +0200
Subject: [PATCH] fix(training): disable DataLoader workers in subprocess
 training

DataLoader worker subprocesses crash inside Docker due to multiprocessing
fork restrictions. Pass --workers 0 to both ketos train and ketos segtrain
so data loading runs in the main process.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 ocr-service/main.py | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/ocr-service/main.py b/ocr-service/main.py
index 955f7fbd..8e4d2288 100644
--- a/ocr-service/main.py
+++ b/ocr-service/main.py
@@ -366,7 +366,8 @@ async def train_model(
             os.makedirs(checkpoint_dir, exist_ok=True)
 
             cmd = [
-                "ketos", "train",
+                "ketos", "--workers", "0",
+                "train",
                 "-f", "path",
                 "-o", checkpoint_dir,
                 "-q", "fixed",
@@ -456,7 +457,8 @@ async def segtrain_model(
             os.makedirs(checkpoint_dir, exist_ok=True)
 
             cmd = [
-                "ketos", "segtrain",
+                "ketos", "--workers", "0",
+                "segtrain",
                 "-o", checkpoint_dir,
                 "-q", "fixed",
                 "-N", "50",