From ba66edae1d4bbe518a738684b37d4e0228e95fc0 Mon Sep 17 00:00:00 2001
From: Ronald Tse <ronald.tse@ribose.com>
Date: Wed, 7 Oct 2026 19:13:51 +0800
Subject: [PATCH] fix: DS.__len__ (RandomSampler requires it); exercised on
 first r8b launch

---
 train_arabic_r8_hf.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/train_arabic_r8_hf.py b/train_arabic_r8_hf.py
index 299e095..407f4b0 100644
--- a/train_arabic_r8_hf.py
+++ b/train_arabic_r8_hf.py
@@ -173,6 +173,9 @@ def add_silver(tgt: str) -> None:
     tok = AutoTokenizer.from_pretrained(TEACHER)
 
     class DS(Dataset):
+        def __len__(self):
+            return len(units)
+
         def __getitem__(self, i):
             src, tgt = units[i]
             e = tok(src, truncation=True, max_length=WINDOW)