| import gradio as gr | |
| from transformers import MBartTokenizer, MBartForConditionalGeneration | |
| model_name = "IlyaGusev/mbart_ru_sum_gazeta" | |
| tokenizer = MBartTokenizer.from_pretrained(model_name) | |
| model = MBartForConditionalGeneration.from_pretrained(model_name) | |
| def summarize(text): | |
| input_ids = tokenizer.batch_encode_plus([text], return_tensors="pt", max_length=1024)["input_ids"].to(model.device) | |
| summary_ids = model.generate(input_ids=input_ids, no_repeat_ngram_size=4) | |
| return tokenizer.decode(summary_ids[0], skip_special_tokens=True) | |
| gr.Interface(fn=summarize, inputs="text", outputs="text", description="Russian Summarizer").launch() |