YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

GreenSMKZ Tags — multilingual-e5-base

Модель многоклассовой тематической классификации отзывов и постов о сервисе Green SM KZ (такси). Поддерживает русский язык, казахский кириллический и казахский латинский.

  • Базовая модель: intfloat/multilingual-e5-base
  • Задача: multi-label classification (20 тегов)
  • Пулинг: (CLS + mean_pool) / 2
  • Loss: asymmetric + LabelSmoothing=0.05
  • HEAD tokens: 256 | TAIL tokens: 253
  • Best F1 Micro: 0.8214
  • Best F1 Macro: 0.7880
  • Обучена: 20260707-161838

Теги (20 шт.)

  • Accident
  • App Crashes / Bugs
  • Apps
  • Driver
  • Fast / Efficient Performance
  • Fuel Efficiency / Sustainability
  • Good Navigation
  • Harassment
  • Impolite or Rude Behavior
  • Poor User Experience
  • Professionalism / Courteous Behavior
  • Safe / Smooth Driving
  • Safe Driving
  • Safety
  • Unsafe Driving
  • Unsafe Driving / Vehicle Condition
  • User-Friendly Interface
  • Vehicle Breakdown
  • Vehicle Comfort
  • Vehicles

Инференс

import json, torch
from transformers import AutoTokenizer

with open("tag2id.json") as f: tag2id = json.load(f)
with open("per_tag_thresholds.json") as f: thr_data = json.load(f)
id2tag     = {v: k for k, v in tag2id.items()}
thresholds = thr_data["thresholds"]

tokenizer = AutoTokenizer.from_pretrained("DanielNRU/GreenSMKZ-tags")
text = "query: Отличный сервис, водитель был очень вежлив!"
enc  = tokenizer(text, return_tensors="pt", max_length=512,
                 truncation=True, padding="max_length")
with torch.no_grad():
    logits = model(enc["input_ids"], enc["attention_mask"])
    probs  = torch.sigmoid(logits).squeeze().numpy()
tags = [id2tag[str(i)] for i, p in enumerate(probs)
        if p >= thresholds.get(id2tag[str(i)], 0.3)]
print(tags)
Downloads last month
15
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Space using DanielNRU/GreenSMKZ-tags 1