Can you share some experience about how to train a model like this,thanks!

#1
by seasnake - opened

Can you share some experience about how to train a model like this,thanks!

Sure, you can look at this notebook, it is a simple seq2seq model fine tuning.

https://colab.research.google.com/drive/13qTF7Z_Lf-FO-7_j49NTUQWffi7m0_8K?usp=sharing

Whether to train with flux?

Sure, you can look at this notebook, it is a simple seq2seq model fine tuning.

https://colab.research.google.com/drive/13qTF7Z_Lf-FO-7_j49NTUQWffi7m0_8K?usp=sharing

step 5 of this script is brroken ---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
/tmp/ipython-input-4-1280422743.py in <cell line: 0>()
3 epochs = 5
4 model_name = f"{model_checkpoint}-prompt-enhancer"
----> 5 args = Seq2SeqTrainingArguments(
6 model_name,
7 evaluation_strategy="epoch",

TypeError: Seq2SeqTrainingArguments.init() got an unexpected keyword argument 'evaluation_strategy'

Probably needs old version of transformers

pip install transformers[torch] makes it run

Sign up or log in to comment