CoolFace
Modelpublic

nickoo004/karakalpak-gpt2-v3

sourceHugging Facemitupdated 6mo agoView on Hugging Face
0likes18downloads
Model Card

Kaa Karakalpak GPT-2 Model

GPT-2 model for the Karakalpak language.

Model Details

  • —Architecture: GPT-2
  • —Parameters: 97.0M
  • —Layers: 8
  • —Embedding: 768
  • —Vocab Size: 52000
  • —Context Length: 512 tokens

Training

  • —Dataset:
  • —Training Time: 11:00:02
  • —Final Loss: 4.1364
  • —Hardware: Tesla T4 GPU

Usage

python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("nickoo004/karakalpak-gpt2-v3")
tokenizer = AutoTokenizer.from_pretrained("nickoo004/karakalpak-gpt2-v3")

prompt = "Qaraqalpaqstan"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=50, do_sample=True)
print(tokenizer.decode(outputs[0]))

License

MIT License