waka-gpt

Model Description

waka-gpt is a GPT-based language model designed for Japanese text generation. It was constructed by fine-tuning the original base model, rinna/japanese-gpt2-small (based on base model information). Features and Use Cases (Samples): Generating classical and waka-style texts Generating Japanese poetry and creative writing Conversation/Creative writing assistance

• Developed by: supertakerin2
• Funded by: No one
• Shared by: No one
• Model type: Text-generation
• Language(s) (NLP): Japanese,Classical Japanese
• License: Mit License
• Finetuned from model : rinna/japanese-gpt2-small

Model Sources

The Eight Imperially Compiled Anthologies of Waka Poetry, compiled during the golden age of waka culture

Direct Uses

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("supertakerin2/waka-gpt")
model = AutoModelForCausalLM.from_pretrained("supertakerin2/waka-gpt")

input_text = "古き良き日本の風景を詠む和歌："
inputs = tokenizer(input_text, return_tensors="pt")

outputs = model.generate(**inputs, max_length=100)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Bias, Risks, and Limitations

waka-gpt is a statistical language model trained on large volumes of text, and the following points require attention: Potential for factual errors in output Potential for inclusion of biases Potential to generate inappropriate expressions in specific contexts Proper filtering and operation under supervision are recommended.

Model Card Contact

[email protected]