0001-compatible.patch

From 700ea95f31e1151064a70db5913fd349ddc1ff67 Mon Sep 17 00:00:00 2001
From: root <root@heming-g5.cs.hku.hk>
Date: Thu, 2 Dec 2021 03:10:09 +0000
Subject: [PATCH] compatible

---
 fairseq/data/data_utils.py      | 1 +
 fairseq/data/indexed_dataset.py | 1 +
 2 files changed, 2 insertions(+)

diff --git a/fairseq/data/data_utils.py b/fairseq/data/data_utils.py
index 7914e605..187ef335 100644
--- a/fairseq/data/data_utils.py
+++ b/fairseq/data/data_utils.py
@@ -45,6 +45,7 @@ def collate_tokens(
     pad_to_bsz=None,
 ):
     """Convert a list of 1d tensors into a padded 2d tensor."""
+    pad_to_multiple=1
     size = max(v.size(0) for v in values)
     size = size if pad_to_length is None else max(size, pad_to_length)
     if pad_to_multiple != 1 and size % pad_to_multiple != 0:
diff --git a/fairseq/data/indexed_dataset.py b/fairseq/data/indexed_dataset.py
index d0843926..d783bb8a 100644
--- a/fairseq/data/indexed_dataset.py
+++ b/fairseq/data/indexed_dataset.py
@@ -75,6 +75,7 @@ def make_builder(out_file, impl, vocab_size=None):
 
 
 def make_dataset(path, impl, fix_lua_indexing=False, dictionary=None):
+    impl = "lazy"
     if impl == "raw" and IndexedRawTextDataset.exists(path):
         assert dictionary is not None
         return IndexedRawTextDataset(path, dictionary)
-- 
2.17.1