🌾 KlonSuphap-LM (แต่งกลอนแปด ด้วย GPT-2)

Visit Demo Space -> Kongfha/KlonSuphap-Generator <br> Visit GitHub Repository -> Kongfha/KlonSuphap-LM <br> Visit Blog (Thai Language) -> 🌾 KlonSuphap-LM แต่งกลอนแปด ด้วย GPT-2

KlonSuphap-LM or GPT-2 for Thai poems (Klon-Paed Poem). I use GPT-2 base Thai as a pre-trained model for fine-tuning exclusively on Thai Klon-Paed Poem (กลอนแปด) retrieved from Thai Literature Corpora (TLC) dataset.

Prior to my recent poem-generation model, PhraAphaiManee-LM, although the model can perform a depiction of Thai Klon-Paed Poems, it still does not adhere to the rules of Thai Klon-Paed (ฉันทลักษณ์) in its generated output. To overcome this challenge I developed techniques that make the model to be more adhere to rules are as follows.

1. Fine-Tuning dataset preprocessing.<br>   As I have a limited quantity of Thai Klon-Paed Poem or about 65770 lines (บาท), thus to succeed in the objective of making the model to be more adhere to rules, I developed a technique called "Rhyme Tagging". <br>   "Rhyme Tagging" performs tag insertion before and after words that are expected to rhyme with the other words based on Klon-Paed Rules. <br> <u>Example</u><br>

     พอได้ยินเสียงระฆังข้างหลัง<s2>เขา</s2><br>เห็นผู้<es2>เฒ่า</es2>ออกจากชะวาก<s2>ผา</s2><br>สรรพางค์ร่างกายแก่ช<es2>รา</es2><br>แต่ผิว<es2>หน้า</es2>นั้นละม้ายคล้ายทา<s3>รก</s3>  

   With "Rhyme Tagging", the potential loss of rhyme information due to an overwhelming flood of non-rhyme-related data can be mitigated. This approach aids the self-attention mechanism in extracting a greater amount of rhyme-related information, ensuring its preservation and relevance throughout the processing.

2. Applying Attention-Mask while fine-tuning.<br>   Apart from performing a common fine-tuning process using the preprocessed dataset, I did fine-tune the model by applying Attention-Mask to non-rhyme-related words to the dataset as following visualization.<br> <u>Visualized Example</u><br>

     ------------------------------<s2>เขา</s2><br>-----<es2>เฒ่า</es2>--------------------<s2>ผา</s2><br>---------------------------<es2>รา</es2><br>------<es2>หน้า</es2>-----------------------<s3>รก</s3>  

   By applying Attention-Mask while fine-tuning, the model can prioritize the extraction of information from both the rhyme-tags and their surrounding words without dropping positional information. This enhances the model's performance in subsequent stages of fine-tuning as if the model were constructing lookup table for rhyme-related words.

3. Performing Reinforcement Learning<br>   After the stage of Supervised Fine-Tuning, I perform Reinforcement Learning to the model using voidful/TextRL by defining Klon-Paed Grader as a PPO Environment.<br>   I perform Reinforcement Learning by randomly pick initial 2-5 syllables from the validation set as text inputs in an observation list, then I force the model to generate only 1 line (บาท) which has only 1 rhyme pair.<br>   TextRL will repeatedly feed text inputs from the observation list to the model and calculate the reward using my Klon-Paed Grader, then update the model's weights based on rewards it recieved.

Cherry-Picked Examples From Demo (Top-P 0.8 Temp 0.8)

  ปัญญาประดิษฐ์องค์ทรงสุรดี<br>เห็นสุดมีบังคมก้มเกศา<br>ต่างยิ้มละลูกยับลงตรงบันลา<br>ถึงว่ารุ่งรางสว่างกลางนวัง  

  ขอขอบคุณบุญกุศลจิต<br>เป็นเพื่อนคิดจะเป็นคู่เคหา<br>ต่างคนกับเหล่านางสร้อยตา<br>ต้องมาก็จะมาไปว่าไร  

  ทรานส์ฟอร์เมอร์มีเซลฟ์แอตเทนชัน<br>ขึ้นบรรลักษณ์ก็เหลือบเขียนฉงน<br>ที่จับต้อนแต่เรือนเพื่อนเหมือนอย่างวน<br>จะต้องชวนมาช่วยให้เชยชม  

Example use

from transformers import pipeline
from transformers import AutoTokenizer, AutoModelForCausalLM

model_name = "Kongfha/KlonSuphap-LM"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

generate = pipeline("text-generation",
                    model=model,
                    tokenizer=tokenizer)

input_sentence = "มิตรแท้"
generated_text = generate(input_sentence,
                          max_length=160,
                          top_p=0.85,
                          temperature=1)
# generation parameters can be varied 

print(f"Input: {input_sentence}")
print(f"Output:\n {generated_text[0]['generated_text']}")