Updates architext imports. Additional minor changes to the architext script:
The architext script is now configured to use csv files instead of arrow files.
The dataset is now stored in the /dataset directory instead of the cheese root dir.
There is no longer any binary saved directly in the datasets, instead we save images of the generations alongside the csv dataset. The csv entry for each prompt + feedback collection will have the filename of the image.
The script is configured to append to existing datasets rather than replace them each time the script is run.
Updates architext imports. Additional minor changes to the architext script: