amnghd/persian-poet-generator

Persian Poetry Generator | شاعر فارسی

This model generates classical Persian (Farsi) poetry based on topics or opening verses.

Model Description

• Base model: ai-forever/mGPT-1.3B-persian
• Fine-tuned on: Classical Persian poetry corpus (~120MB, 48 poets)
• Training method: LoRA + 8-bit quantization
• Parameters: 1.3B

Usage

Topic-based generation

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("amnghd/persian-poet-generator")
tokenizer = AutoTokenizer.from_pretrained("amnghd/persian-poet-generator")

# Generate poem about love (عشق)
prompt = "<|topic|>عشق<|startofpoem|>"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=200, temperature=0.8, do_sample=True)
print(tokenizer.decode(outputs[0]))

Verse continuation

prompt = "<|continue|>بشنو این نی چون شکایت می کند<|startofpoem|>"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=200, temperature=0.8, do_sample=True)
print(tokenizer.decode(outputs[0]))

Training Data

Trained on classical Persian poetry from 48 poets including:

• Hafez (حافظ)
• Saadi (سعدی)
• Moulavi/Rumi (مولوی)
• Ferdowsi (فردوسی)
• Khayyam (خیام)
• Attar (عطار)
• And many more...

Total: ~1.8 million verses

GitHub

https://github.com/amnghd/persian-poetry-generator