Closed xsthunder closed 1 year ago
According to https://platform.openai.com/docs/api-reference/fine-tunes
, it seems curie:ft-uoft-ml-group-2022-09-26-01-07-30
was a finetuned curie model by the authors. I tried replacing curie
with finetuned curie model, but it didn't work. All the truthful, informative
in function gpt_eval
in file /content/automatic_prompt_engineer/experiments/evaluation/TruthfulQA/fine_tuned_gpt_eval.py
were zeroes, which means no valid evaluation was made.
Not sure what to do next. I wonder if there were any deatils about the finetuned curie model in the paper.
I fine-tuned a curie
similar to curie:ft-uoft-ml-group-2022-09-26-01-07-30
using openai api fine_tunes.create -t finetuned_truth.jsonl -m curie --n_epochs 5 --batch_size 21 --learning_rate_multiplier 0.1 --no_packing
in https://github.com/sylinrl/TruthfulQA
. It costs $11.07 for fine-tuning. Hope it will work.
It worked.
Hi! Thanks for open source and readable code! However, when I run
python experiments/run_truthful_qa.py
, its raise does not exist error. Should I use "curie" instead of "curie:ft-uoft-ml-group-2022-09-26-01-07-30"? thanks in advance.possible source of error
error