---
language: 
  - tw
tags:
- albert
- classification
license: afl-3.0
metrics:
- Accuracy
---

# Traditional Chinese news classification

繁體中文新聞分類任務，使用ckiplab/albert-base-chinese預訓練模型，資料集只有2.6萬筆，做為課程的範例模型。

    from transformers import BertTokenizer, AlbertForSequenceClassification
    model_path = "clhuang/albert-news-classification"
    model = AlbertForSequenceClassification.from_pretrained(model_path)
    tokenizer = BertTokenizer.from_pretrained("bert-base-chinese")
    
    # get category probability
    def get_category_proba( text ):
        max_length = 250
        # prepare our text into tokenized sequence
        inputs = tokenizer([text], padding=True, truncation=True, max_length=max_length, return_tensors="pt")
        # perform inference to our model
        outputs = model(**inputs)
        # get output probabilities by doing softmax
        probs = outputs[0].softmax(1)
    
        # executing argmax function to get the candidate label
        # probs.argmax()
        label_index = probs.argmax(dim=1)[0].tolist() # convert tensor to int
        # label_index = np.argmax(probs.detach(), axis=1)
        
        label = idx2cate[ label_index ]
    
        # Note that result is numpy format and it should be convert to float
        proba = round(float(probs.tolist()[0][label_index]),2)
    
        response = {'label': label, 'proba': proba}
    
        return response
       
    get_category_proba('俄羅斯2月24日入侵烏克蘭至今不到3個月，芬蘭已準備好扭轉奉行了75年的軍事不結盟政策，申請加入北約。芬蘭總理馬林昨天表示，「希望我們下星期能與瑞典一起提出申請」。')