from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("PiotrWarzachowski/polish-text-summarizer")
model = AutoModelForSeq2SeqLM.from_pretrained("PiotrWarzachowski/polish-text-summarizer")

article = "Twój długi artykuł po polsku..."

inputs = tokenizer(article, max_length=512, truncation=True, return_tensors="pt")
outputs = model.generate(**inputs, max_length=128, num_beams=4, no_repeat_ngram_size=3)
summary = tokenizer.decode(outputs[0], skip_special_tokens=True)

print(summary)

Limitations

Max input: 512 tokens (~2000-3000 characters)
Max output: 128 tokens (~500 characters)
Polish diacritics (ą, ę, ł, etc.) may be simplified to ASCII equivalents

Training

Optimizer: Adafactor
Batch size: 1 (with gradient accumulation 8)
Epochs: 3
Learning rate: 1e-4