Skip to content

Instantly share code, notes, and snippets.

@hamletbatista

hamletbatista/t5_config_yaml.py

Last active September 9, 2021 13:48
Show Gist options
  • Save hamletbatista/1f9a86d4364f641f776808bff03f557b to your computer and use it in GitHub Desktop.
Save hamletbatista/1f9a86d4364f641f776808bff03f557b to your computer and use it in GitHub Desktop.
Download ZIP
Raw
t5_config_yaml.py
yaml_config="""
input_features:
-
name: Original_Title
type: text
level: word
encoder: t5
reduce_output: null
-
name: Keyword
type: text
level: word
tied_weights: Original_Title
encoder: t5
reduce_output: null
output_features:
-
name: Optimized_Title
type: sequence
level: word
decoder: generator
"""
with open("config.yaml", "w") as f:
f.write(yaml_config)
@Suvakanta8
Copy link

Hii while I am trying to replace hootsuite_titles.csv dataset with my website dataset I am getting a value error like this:
ValueError: A Concatenate layer requires inputs with matching shapes except for the concat axis. Got inputs shapes: [(128, 32, 512), (128, 224, 512)]
please reply.......

@marcusbianco
Copy link

marcusbianco commented Sep 8, 2021 via email

@Suvakanta8
Copy link

Thanks @marcusbianco can you please explain me how to prepare a dataset properly for training and the optimized score column in hootsuit dataset what is that ? Is that necessary for training?

@Suvakanta8
Copy link

@marcusbianco While I am using this data set https://docs.google.com/spreadsheets/d/e/2PACX-1vQeg6komuBq6U4Rn_SWFNXgGxXIL4fhQcdJp2A8p0B2pmiucvDG5SUpXEGKooJPaTPsN0ivIaCvqIad/pub?gid=0&single=true&output=csv
its running fine but just check this dataset also https://docs.google.com/spreadsheets/d/e/2PACX-1vSbe_3C_JV06TQvL-TR741o7goEycu0oGwO3kfJleawuDQEQiu2ZC2lpgeHhRQayJ4JoZe6PvbdCvbF/pub?gid=0&single=true&output=csv , while I am trying to run this one its throwing that value error... please guide me.. Kindly check and tell me the differences..

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment