Distilbert Base Multilingual Cased Sentiments Student

🚀 distilbert-base-multilingual-cased-sentiments-student

該模型是在多語言情感數據集上，使用此腳本從零樣本分類管道中蒸餾而來。實際上，多語言情感數據集當然是有標註的，但為了便於示例，我們會假裝並忽略這些標註。

教師模型：MoritzLaurer/mDeBERTa-v3-base-mnli-xnli 教師假設模板：“The sentiment of this text is {}.” 學生模型：distilbert-base-multilingual-cased

🚀 快速開始

本模型可用於多語言情感分析任務，通過蒸餾技術從教師模型中學習，能在多種語言上進行情感分類。

✨ 主要特性

• 多語言支持：支持英語、阿拉伯語、德語、西班牙語、法語、日語、中文、印尼語、印地語、意大利語、馬來語、葡萄牙語等多種語言。
• 零樣本蒸餾：基於零樣本蒸餾技術，在無標註數據上進行模型訓練。

📦 安裝指南

本模型基於Hugging Face的transformers庫，你可以使用以下命令安裝所需依賴：

pip install transformers datasets torch

💻 使用示例

基礎用法

from transformers import pipeline

distilled_student_sentiment_classifier = pipeline(
    model="lxyuan/distilbert-base-multilingual-cased-sentiments-student", 
    return_all_scores=True
)

# 英文示例
distilled_student_sentiment_classifier ("I love this movie and i would watch it again and again!")
>> [[{'label': 'positive', 'score': 0.9731044769287109},
  {'label': 'neutral', 'score': 0.016910076141357422},
  {'label': 'negative', 'score': 0.009985478594899178}]]

# 馬來語示例
distilled_student_sentiment_classifier("Saya suka filem ini dan saya akan menontonnya lagi dan lagi!")
[[{'label': 'positive', 'score': 0.9760093688964844},
  {'label': 'neutral', 'score': 0.01804516464471817},
  {'label': 'negative', 'score': 0.005945465061813593}]]

# 日語示例
distilled_student_sentiment_classifier("私はこの映畫が大好きで、何度も見ます！")
>> [[{'label': 'positive', 'score': 0.9342429041862488},
  {'label': 'neutral', 'score': 0.040193185210227966},
  {'label': 'negative', 'score': 0.025563929229974747}]]

📚 詳細文檔

訓練過程

筆記本鏈接：點擊查看

訓練超參數

可以使用以下命令復現訓練結果：

python transformers/examples/research_projects/zero-shot-distillation/distill_classifier.py \
--data_file ./multilingual-sentiments/train_unlabeled.txt \
--class_names_file ./multilingual-sentiments/class_names.txt \
--hypothesis_template "The sentiment of this text is {}." \
--teacher_name_or_path MoritzLaurer/mDeBERTa-v3-base-mnli-xnli \
--teacher_batch_size 32 \
--student_name_or_path distilbert-base-multilingual-cased \
--output_dir ./distilbert-base-multilingual-cased-sentiments-student \
--per_device_train_batch_size 16 \
--fp16

如果你在Colab上訓練此模型，為避免內存不足錯誤信息，請進行以下代碼修改：

###### 修改第78行以禁用快速分詞器 
default=False,

###### 更新第313行的數據集映射部分
dataset = dataset.map(tokenizer, input_columns="text", fn_kwargs={"padding": "max_length", "truncation": True, "max_length": 512})

###### 在第213行添加以下行
del model
print(f"Manually deleted Teacher model, free some memory for student model.")

###### 在第337行添加以下行
trainer.push_to_hub()
tokenizer.push_to_hub("distilbert-base-multilingual-cased-sentiments-student")
  

訓練日誌

Training completed. Do not forget to share your model on huggingface.co/models =)

{'train_runtime': 2009.8864, 'train_samples_per_second': 73.0, 'train_steps_per_second': 4.563, 'train_loss': 0.6473459283913797, 'epoch': 1.0}
100%|███████████████████████████████████████| 9171/9171 [33:29<00:00,  4.56it/s]
[INFO|trainer.py:762] 2023-05-06 10:56:18,555 >> The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`,  you can safely ignore this message.
[INFO|trainer.py:3129] 2023-05-06 10:56:18,557 >> ***** Running Evaluation *****
[INFO|trainer.py:3131] 2023-05-06 10:56:18,557 >>   Num examples = 146721
[INFO|trainer.py:3134] 2023-05-06 10:56:18,557 >>   Batch size = 128
100%|███████████████████████████████████████| 1147/1147 [08:59<00:00,  2.13it/s]
05/06/2023 11:05:18 - INFO - __main__ - Agreement of student and teacher predictions: 88.29%
[INFO|trainer.py:2868] 2023-05-06 11:05:18,251 >> Saving model checkpoint to ./distilbert-base-multilingual-cased-sentiments-student
[INFO|configuration_utils.py:457] 2023-05-06 11:05:18,251 >> Configuration saved in ./distilbert-base-multilingual-cased-sentiments-student/config.json
[INFO|modeling_utils.py:1847] 2023-05-06 11:05:18,905 >> Model weights saved in ./distilbert-base-multilingual-cased-sentiments-student/pytorch_model.bin
[INFO|tokenization_utils_base.py:2171] 2023-05-06 11:05:18,905 >> tokenizer config file saved in ./distilbert-base-multilingual-cased-sentiments-student/tokenizer_config.json
[INFO|tokenization_utils_base.py:2178] 2023-05-06 11:05:18,905 >> Special tokens file saved in ./distilbert-base-multilingual-cased-sentiments-student/special_tokens_map.json

框架版本

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 2.11.0
• Tokenizers 0.13.3

📄 許可證

本模型採用Apache-2.0許可證。