Retrain with FLAN-T5-base

[Bachstelze]

We evaluated your and various T5-models on syntax: Evaluating PaLM-FLAN-T5 and previous models on syntax

Gramformer is the the grammar-tuned dot:

The evaluation suggests using FLAN-T5-base

[akesh1235]

@Bachstelze Yeah, Mr. Peter Szemraj trained using flan-t5-large model, size of the model is 3.13 GB as you evaluated sir.
Any suggetion, how can this flan-t5-large model from Mr. Peter Szemraj can be improved further.

[Bachstelze]

@akesh1235 you probably mean flan-t5-large-grammar-synthesis. This model is only trained on the JFLEG dataset. You can add more datasets for the english-centric Flan-T5:

• https://paperswithcode.com/datasets?task=grammatical-error-correction
• https://huggingface.co/datasets?other=grammatical-error-correction
  Also, you can add more languages: Mulitlingual Grammatical Error Correction #32
  Flan-T5 is already pre-trained in plenty languages: English, Spanish, Japanese, Persian, Hindi, French, Chinese, Bengali, Gujarati, German, Telugu, Italian, Arabic, Polish, Tamil, Marathi, Malayalam, Oriya, Panjabi, Portuguese, Urdu, Galician, Hebrew, Korean, Catalan, Thai, Dutch, Indonesian, Vietnamese, Bulgarian, Filipino, Central Khmer, Lao, Turkish, Russian, Croatian, Swedish, Yoruba, Kurdish, Burmese, Malay, Czech, Finnish, Somali, Tagalog, Swahili, Sinhala, Kannada, Zhuang, Igbo, Xhosa, Romanian, Haitian, Estonian, Slovak, Lithuanian, Greek, Nepali, Assamese, Norwegian

[ILG2021]

I have fine tune a flan-t5 model on my own small dataset about 1500 rows. But i find it not works. I only want to correct the error in my own dataset. others just to return the original text. it is not run as i expect.