报错： ValueError: too many values to unpack (expected 2)
复现RGCL 模型，遇到的问题，该代码里面报错原因是： collate_fn 里的 tokens 是多层嵌套列表 （每个样本是分词后的 token 列表，batch 打包后变成 [[token1,token2...], [xxx]]），但旧版 transformers（3.1.0）在 is_pretokenized=True 时内部解包逻辑不兼容，或者数据里混入了非法空样本、维度错乱。
解决方案
原 collate_fn：
def collate_fn(data):
u, i, r, tokens = zip(*data)
encoding = bert_tokenizer(tokens, return_tensors='pt', padding=True,
truncation=True, is_pretokenized=True)
return torch.Tensor(u), torch.Tensor(i), torch.Tensor(r), \
encoding['input_ids'], encoding['attention_mask']改成下面版本，去掉 is_pretokenized=True，改用字符串拼接 / 直接传原文本，彻底规避解包 bug：
def collate_fn(data):
u, i, r, token_lists = zip(*data)
# 把分词列表还原成字符串，不用is_pretokenized
text_list = [" ".join(toks) for toks in token_lists]
encoding = bert_tokenizer(text_list,
return_tensors='pt',
padding=True,
truncation=True,
max_length=args.review_max_length)
return torch.Tensor(u), torch.Tensor(i), torch.Tensor(r), \
encoding['input_ids'], encoding['attention_mask']infact,在修改代码之前，我还多次尝试改变transformers的版本但是还是报一样的错误，可能是没有改到正好可以用的那个版本。（ps.之前跑自己的模型也遇到过版本问题，修改transformers的版本就重新运行了）
完整报错：
(base) root@autodl-container-6d25459816-01b42b10:~/Work/ReviewGraph-main/BERT# python bert_whitening.py
libgomp: Invalid value for environment variable OMP_NUM_THREADS
2026-07-06 00:23:43 - Load_Data - Start reading data to pandas.
Clean string: 100%|███████████████████████████████████████████████████████████████████████| 963441/963441 [01:36<00:00, 9971.91it/s]
Delete unused words: 100%|██████████████████████████████████████████████████████████████| 963441/963441 [00:08<00:00, 107299.97it/s]
2026-07-06 00:25:55 - Load_Data - Truncate review length to 56 words
Delete unused words: 100%|██████████████████████████████████████████████████████████████| 963441/963441 [00:03<00:00, 255554.45it/s]
check data split: 770753it [00:33, 23069.12it/s]
/root/Work/ReviewGraph-main/BERT/../load_data.py:190: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.
train_data = train_data.append([valid_data.loc[valid_drop_user_data_index],
pre tokenize: 100%|███████████████████████████████████████████████████████████████████████| 973763/973763 [09:04<00:00, 1789.32it/s]
0%| | 0/7608 [00:00<?, ?it/s]
Traceback (most recent call last):
File "bert_whitening.py", line 175, in <module>
save_sentence_feat(args)
File "/root/miniconda3/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "bert_whitening.py", line 149, in save_sentence_feat
for u, i, r, input_ids, mask in tqdm(data_loader):
File "/root/miniconda3/lib/python3.8/site-packages/tqdm/std.py", line 1185, in __iter__
for obj in iterable:
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 530, in __next__
data = self._next_data()
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 570, in _next_data
data = self._dataset_fetcher.fetch(index) # may raise StopIteration
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 52, in fetch
return self.collate_fn(data)
File "bert_whitening.py", line 100, in collate_fn
encoding = bert_tokenizer(tokens, return_tensors='pt', padding=True,
File "/root/miniconda3/lib/python3.8/site-packages/transformers/tokenization_utils_base.py", line 3021, in __call__
encodings = self._call_one(text=text, text_pair=text_pair, **all_kwargs)
File "/root/miniconda3/lib/python3.8/site-packages/transformers/tokenization_utils_base.py", line 3109, in _call_one
return self.batch_encode_plus(
File "/root/miniconda3/lib/python3.8/site-packages/transformers/tokenization_utils_base.py", line 3311, in batch_encode_plus
return self._batch_encode_plus(
File "/root/miniconda3/lib/python3.8/site-packages/transformers/tokenization_utils.py", line 886, in _batch_encode_plus
ids, pair_ids = ids_or_pair_ids
ValueError: too many values to unpack (expected 2)