Text Classification
Transformers
Russian
russian
rubert
ruroberta
multi-label-classification
s7-airlines
Eval Results (legacy)
Instructions to use DanielNRU/S7-tags-ruroberta-20260810 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use DanielNRU/S7-tags-ruroberta-20260810 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="DanielNRU/S7-tags-ruroberta-20260810")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("DanielNRU/S7-tags-ruroberta-20260810", device_map="auto") - Notebooks
- Google Colab
- Kaggle
S7 Airlines Tags — ruRoBERTa-large
Модель multi-label тегирования для темы S7 Airlines.
- Базовая модель: ai-forever/ruRoBERTa-large
- Задача: Multi-label классификация
- Датасет: S7_enriched.jsonl
- Теги (65): Fast Track, Gate 7, HR, Nordwind, S7 Cargo, S7 Engineering, S7 Technics, S7 Training, SMS информирование, Seats...
- Best Micro F1: 0.8878 (epoch 15)
- Пулинг: (CLS + mean_pool) / 2
- Обучена: 2026-08-10
Инференс
import json, torch
from transformers import AutoTokenizer
with open("config.json") as f:
cfg = json.load(f)
all_tags = cfg["all_tags"]
tokenizer = AutoTokenizer.from_pretrained("DanielNRU/S7-tags-ruroberta-20260810")
# model = RuRoBERTaTagClassifier("ai-forever/ruRoBERTa-large", num_tags=len(all_tags))
# model.load_state_dict(torch.load("best_pytorch_model.bin", map_location="cpu"))
model.eval()
text = "Потеряли багаж, служба поддержки не отвечает"
enc = tokenizer(
text,
max_length=512,
truncation=True,
padding="max_length",
return_tensors="pt",
)
with torch.no_grad():
logits = model(enc["input_ids"], enc["attention_mask"])
probs = torch.sigmoid(logits).squeeze(0).tolist()
result = {
all_tags[i]: round(probs[i], 3)
for i in range(len(all_tags))
if probs[i] >= 0.3
}
print(result)
- Downloads last month
- 16
Space using DanielNRU/S7-tags-ruroberta-20260810 1
Evaluation results
- Micro F1self-reported0.888