doesn't keep the columns order when write with "tfrecords" format from existence Dataframe

tensorflow / ecosystem

Integration of TensorFlow with other open-source frameworks

Apache License 2.0

1.37k stars 392 forks source link

doesn't keep the columns order when write with "tfrecords" format from existence Dataframe #117

Open u650080 opened 5 years ago

u650080 commented 5 years ago

Hi , I have an existence hdfs file in parquet format. When I write this dataframe in "tfrecords" format and later I read this new file( in "tfrecords" format) , the order of columns are changed and it didn't keep same order as parquet format .

nareshr8 commented 4 years ago

Yeah I also faced the same issue. If you are using the data for training in tensorflow, you can use tf.feature_column to map column names to specific input nodes. But beware of the issue mentioned here

Wasnt aware that this issue was there since 2016. I wont expect it to be solved then