How to use stefanj0/t5gemma-math-corrector with PEFT:
from peft import PeftModel from transformers import AutoModelForSeq2SeqLM base_model = AutoModelForSeq2SeqLM.from_pretrained("google/t5gemma-s-s-ul2-it") model = PeftModel.from_pretrained(base_model, "stefanj0/t5gemma-math-corrector")
How to use stefanj0/t5gemma-math-corrector with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("stefanj0/t5gemma-math-corrector", device_map="auto")
5a5b209 5505706 5a5b209
1
2
3
4
5
6
7
8
{ "train_runtime": 592.2118, "train_samples_per_second": 25.258, "train_steps_per_second": 1.581, "total_flos": 3090020422385664.0, "train_loss": 0.119470198857598, "epoch": 2.5645549318364074 }
