Text Classification
Transformers
Safetensors
English
HHEMv2Config
custom_code

Finetuning

#8
by dimidd - opened

Thanks for this model!

I'd like to finetune it on my own data and have a question:
In the Vitamin C dataset, the label is an ordinal variable which has 3 levels: "REFUTES", "NOT ENOUGH DATA", and "SUPPORTS".
However, your training code expects the label to be an int:

for i, row in df_train.iterrows():
   train_examples.append(InputExample(texts=[row['source'], row['summary']], label=int(row['label'])))

Could you please elaborate how are these levels converted to ints?

Vectara org

I'm not the author, but I'd imagine you convert REFUTES to 0, SUPPORTS to 1, and then you have to options for NOT ENOUGH DATA: you could either drop it, or treat it as 0.

Right, thank you Amin. I just want to be consistent with the original training data.

forrest-vectara changed discussion status to closed

Sign up or log in to comment