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 0fb51d4 5a5b209
1
2
3
4
5
6
7
8
{ "train_runtime": 671.6999, "train_samples_per_second": 21.099, "train_steps_per_second": 1.322, "total_flos": 3379504493887488.0, "train_loss": 0.13446486646494082, "epoch": 3.0 }
