Upload README.md with huggingface_hub

README.md

@@ -1,65 +1,49 @@
 ---
-library_name: transformers
-base_model: sberbank-ai/ruT5-base
+language: ru
 tags:
-- generated_from_trainer
-metrics:
-- rouge
-model-index:
-- name: RuT5SentimentSummarization
-  results: []
+- summarization
+- t5
+- russian
+- sentiment-analysis
+license: mit
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
+# RuT5-Base for Sentiment Summarization
 
-# RuT5SentimentSummarization
+## Описание
+Модель для генеративной суммаризации отзывов на русском языке. Обучалась на датасете Kinopoisk с использованием дистилляции знаний от LLM.
 
-This model is a fine-tuned version of [sberbank-ai/ruT5-base](https://huggingface.co/sberbank-ai/ruT5-base) on the None dataset.
-It achieves the following results on the evaluation set:
-- Loss: 1.5647
-- Rouge1: 0
-- Rouge2: 0
-- Rougel: 0
+## Задача
+Создание кратких, информативных резюме отзывов с выделением ключевых плюсов и минусов.
 
-## Model description
+## Метрики на тестовой выборке
+(Будут добавлены после завершения обучения)
 
-More information needed
+## Использование
+```python
+from transformers import T5ForConditionalGeneration, T5Tokenizer
 
-## Intended uses & limitations
+model = T5ForConditionalGeneration.from_pretrained("Auttar/RuT5SentimentSummarization")
+tokenizer = T5Tokenizer.from_pretrained("Auttar/RuT5SentimentSummarization")
 
-More information needed
+def summarize(review):
+    input_text = f"summarize: {review}"
+    inputs = tokenizer(input_text, return_tensors="pt", max_length=512, truncation=True)
+    outputs = model.generate(inputs.input_ids, max_length=128, num_beams=4)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
-## Training and evaluation data
+# Пример
+review = "Фильм отличный! Актеры играют великолепно, сюжет захватывает."
+print(summarize(review))
 
-More information needed
+# Обучение
+- Базовая модель: sberbank-ai/ruT5-base
 
-## Training procedure
+- Датасет: Kinopoisk с сгенерированными резюме
 
-### Training hyperparameters
+- Эпохи: 3
 
-The following hyperparameters were used during training:
-- learning_rate: 5e-05
-- train_batch_size: 4
-- eval_batch_size: 4
-- seed: 42
-- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
-- lr_scheduler_type: linear
-- num_epochs: 3
-- mixed_precision_training: Native AMP
+- Learning rate: 5e-5
 
-### Training results
+- Batch size: 4
 
-| Training Loss | Epoch | Step  | Validation Loss | Rouge1 | Rouge2 | Rougel |
-|:-------------:|:-----:|:-----:|:---------------:|:------:|:------:|:------:|
-| 1.9315        | 1.0   | 4090  | 1.6697          | 0      | 0      | 0      |
-| 1.7284        | 2.0   | 8180  | 1.5894          | 0      | 0      | 0      |
-| 1.5832        | 3.0   | 12270 | 1.5647          | 0      | 0      | 0      |
-
-
-### Framework versions
-
-- Transformers 5.0.0
-- Pytorch 2.10.0+cu128
-- Datasets 4.0.0
-- Tokenizers 0.22.2