README.md

XLNet Multi-class Text Classification for Chinese and English

• Available for both multi-label and single-label classification.
• Loss functions used: CrossEntropyLoss & BCEWithLogitsLoss
• Note: Example here is using Chinese pre-trained model. English pre-trained model is commented out.
• Codes are modified from here.

User Guide

1. To evaluate our model, we first split the training dataset into training and testing part.
   train_test_split_v2.ipynb
2. Use the training part to train the model and the testing part to evaluate the model. (Do any modification until you get a satisfying score.)
   • For single-label testing, you may use "Cross Entropy Loss Function".
     product/xlnet_multi_class_chinese_train_product_single_label_ce_oversample.ipynb
   • For multi-label testing, you may use "Binary Cross Entropy with Logits Loss Function".
     product/xlnet_multi_class_chinese_train_product_single_label_bcel.ipynb
3. Use the complete training dataset to train the model, and predict classes of a new dataset (answer unknown).
   xlnet_multi_class_chinese_brand_single_label_bcel.ipynb

Reference

1. Official XLNet

• Paper: https://paperswithcode.com/paper/xlnet-generalized-autoregressive-pretraining
• Code: https://github.com/zihangdai/xlnet

2. Hugging Face Transformers

• Source: https://github.com/huggingface/transformers
• XLNet doc: https://huggingface.co/transformers/v2.11.0/model_doc/xlnet.html

!pip install transformers
!pip install sentencepiece

tokenizer = XLNetTokenizer.from_pretrained('xlnet-base-cased', do_lower_case=True)
model = XLNetModel.from_pretrained('xlnet-base-cased')

3. Multi-label Text Classification (important)

• Source: https://towardsdatascience.com/multi-label-text-classification-with-xlnet-b5f5755302df

4. XLNet Chinese Pre-trained Model (中文預訓練模型)

• 資料來源：https://blog.csdn.net/ciacai/article/details/105008287
• 程式碼：https://github.com/ymcui/Chinese-XLNet

可以直接透過 transformers 使用

!pip install transformers
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("MODEL_NAME")
model = AutoModel.from_pretrained("MODEL_NAME")

模型名	MODEL_NAME
XLNet-mid	hfl/chinese-xlnet-mid
XLNet-base	hfl/chinese-xlnet-base

Prevent Google Colab from Disconnecting

Right click ➡ Inspect ➡ Console

function ConnectButton(){
    console.log("Connect pushed");
    document.querySelector("#top-toolbar > colab-connect-button").shadowRoot.querySelector("#connect").click()
}
setInterval(ConnectButton,60000);