What is the atomic file?

[mayaKaplansky]

Hi,
lets say I have my own dataset, and convert it to the tables as described. When I want to actually save them as atomic files, do I just change the file type manually?

[hyp1231]

Sure it is.

You need to pay attention to two parts mainly:

1. The first line, which contains multiple columns in field_name:field_type format.
2. The suffixes of files, as you mentioned, indicate the file type.

See https://github.com/RUCAIBox/RecBole/tree/master/tests/test_data/test for examples of atomic file.

RecBole/tests/test_data/test/test.inter

Lines 1 to 4 in 511cda8

	user_id:token item_id:token rating:float timestamp:float label:float
	6 86 3 883603013 1
	38 95 5 892430094 0
	97 194 3 884238860 0

[mayaKaplansky]

Thanks! So I have to create a dataframe in this format, and then write it line by line into a new txt file.
Then change the extension from txt to inter?
I cannot use pickle dump?

[hyp1231]

Thanks! So I have to create a dataframe in this format, and then write it line by line into a new txt file.
Then change the extension from txt to inter?
I cannot use pickle dump?

As long as the text file fit the format mentioned above, it's ok. So write a pandas.DataFrame line by line is one of the acceptable ways.

Pickle dump is not available yet.

[mayaKaplansky]

Thank you! In all atomic file types, the delimiter is a tab? I think I saw in some of the examples a space.

[hyp1231]

The delimiter in atomic files can be arbitrary, which should be the same with config['field_separator'].
By default, config['field_separator'] == '\t'.

Please refer to Atomic File Format for more details.

[mayaKaplansky]

Thank you!