在加载数据时报错:column names don't match

#3
by DavinciEvans - opened

在使用 load_dataset 的时候,在加载到第477237条记录时报错:

Failed to read file '/home/ubuntu/smoltalk-chinese/data/math23k_zh.parquet' with error <class 'datasets.table.CastError'>: Couldn't cast
conversations: list<item: struct<content: string, role: string>>
  child 0, item: struct<content: string, role: string>
      child 0, content: string
      child 1, role: string
source: string
answer: string
text_len: int64
-- schema metadata --
pandas: '{"index_columns": [], "column_indexes": [], "columns": [{"name":' + 543
to
{'conversations': [{'content': Value(dtype='string', id=None), 'role': Value(dtype='string', id=None)}], 'system_prompt_key': Value(dtype='string', id=None), 'magpie_model': Value(dtype='string', id=None), 'difficulty': Value(dtype='string', id=None), 'score': Value(dtype='int64', id=None), 'classify': Value(dtype='string', id=None), 'n_turn': Value(dtype='float64', id=None), 'text_len': Value(dtype='int64', id=None)}
because column names don't match

似乎是列名没有对齐,看了一下 hf 上的 数据浏览页,也是到了这个位置就无法加载了。麻烦检查一下

opencsg org

已修改

yuyijiong changed discussion status to closed

Sign up or log in to comment