情感分类模型(Emotion Classification Model)

简介

输⼊⼀段⽂本, 模型可以给出对应的情感分类。分类为二分类，将情感分为积极（positive）和消极（negative）。

实验环境

魔搭社区GPU环境

• 8核 32GB 显存24G
• 预装 ModelScope Library
• 预装镜像 ubuntu22.04-cuda12.1.0-py310-torch2.1.2-tf2.14.0-1.11.0

训练方法

数据集

基于深度学习（LSTM）的情感分析（京东商城数据）jd_sentiment_zh

示例代码

# 设置pip全局镜像
pip config set global.index-url https://mirrors.aliyun.com/pypi/simple/
# 安装ms-swift
git clone https://github.com/modelscope/swift.git
cd swift
pip install -e .[llm]

# It requires 5 minutes of training.
# Experimental environment: A10
# 13GB GPU memory
import os
os.environ['CUDA_VISIBLE_DEVICES'] = '0'
import torch
from swift.llm import (
DatasetName, InferArguments, ModelType, SftArguments,
infer_main, sft_main, app_ui_main, merge_lora_main
)
model_type = ModelType.qwen_1_8b
# 超参数:
# https://github.com/modelscope/swift/blob/main/docs/source/LLM/%E5%91%BD%E4%BB%A
# 4%E8%A1%8C%E5%8F%82%E6%95%B0.md
sft_args = SftArguments(
    model_type=model_type,
    sft_type='lora',
    train_dataset_sample=2000,
    dataset=[DatasetName.jd_sentiment_zh],
    output_dir='output')
result = sft_main(sft_args)
# adapter的⼤⼩: 6.1MiB
best_model_checkpoint = result['best_model_checkpoint']
print(f'best_model_checkpoint: {best_model_checkpoint}')
torch.cuda.empty_cache()
infer_args = InferArguments(
    ckpt_dir=best_model_checkpoint,
    load_dataset_config=True,
    do_sample=False)
result = infer_main(infer_args)

推理效果

positive<|endoftext|>

[LABELS]positive

[PROMPT]Task: Sentiment Classification \ Sentence: 店家太不负责任了，衣服质量太差劲了，和图片上的不一样 \ Category: negative, positive \ Output: [OUTPUT]negative<|endoftext|>

[LABELS]negative

[PROMPT]Task: Sentiment Classification \ Sentence: 送国际友人挺好的，不错不错！ \ Category: negative, positive \ Output: [OUTPUT]positive<|endoftext|>

[LABELS]positive

[PROMPT]Task: Sentiment Classification \ Sentence: 很好,装好一定很漂亮 \ Category: negative, positive \ Output: [OUTPUT]positive<|endoftext|>

[LABELS]positive

[PROMPT]Task: Sentiment Classification \ Sentence: 东西给你退回去了，你要黑我钱！！！ \ Category: negative, positive \ Output: [OUTPUT]

negative<|endoftext|>

[LABELS]negative

[PROMPT]Task: Sentiment Classification \ Sentence: 送货很快，书是正品，买书一直京东是首选！ \ Category: negative, positive \ Output: [OUTPUT]positive<|endoftext|>

[LABELS]positive

[PROMPT]Task: Sentiment Classification \ Sentence: 口感相当的好 都想买第二次了 \ Category: negative, positive \ Output: [OUTPUT]positive<|endoftext|>

[LABELS]positive

[PROMPT]Task: Sentiment Classification \ Sentence: 硅胶味道太重，样子与图片差距太大 \ Category: negative, positive \ Output: [OUTPUT]negative<|endoftext|>

[LABELS]negative

[PROMPT]Task: Sentiment Classification \ Sentence: 很伤心，买了放到三星n4尽然不能用，客服各种推 \ Category: negative, positive \ Output: [OUTPUT]negative<|endoftext|>

[LABELS]negative

[PROMPT]Task: Sentiment Classification \ Sentence: 质量不错，大小合适，应当是正品！但是我买的是黑灰，发来的却是纯黑，懒得换了，给个差评，希望以后改进！ \ Category: negative, positive \ Output: [OUTPUT]positive<|endoftext|> [INFO:swift] saveresultpath: /mnt/workspace/output/qwen-18b/v0-20240126-230928/checkpoint-100/inferresult_20240126-231510.jsonl

[LABELS]negative