MindSpore QuickStart——LSTM算法实践心得体会

加粗样式MindSpore QuickStart——LSTM算法实践

安装0.3.0-alpha

# 在矩阵池中更新mindspore 
pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/0.3.0-alpha/MindSpore/gpu/ubuntu_x86/cuda-10.1/mindspore_gpu-0.3.0-cp37-cp37m-linux_x86_64.whl

下载Glove数据集

Unzip the aclImdb_v1 dataset to any path you want and the folder structure should be as follows:
.
├── train # train dataset
└── test # infer dataset
Unzip the glove.6B.zip to any path you want and the folder structure should be as follows:
.
├── glove.6B.100d.txt
├── glove.6B.200d.txt
├── glove.6B.300d.txt # we will use this one later.
└── glove.6B.50d.txt
Adding a new line at the beginning of the file which named glove.6B.300d.txt. It means reading a total of 400,000 words, each represented by a 300-latitude word vector.
400000 300

训练数据

python train.py --preprocess=true --aclimdb_path=your_imdb_path --glove_path=your_glove_path  > out.train.log 2>&1 & 
#注意替换路径

结果

epoch: 20 step: 386, loss is 0.14661261
epoch: 20 step: 387, loss is 0.1384366
epoch: 20 step: 388, loss is 0.10829709
epoch: 20 step: 389, loss is 0.2376634
epoch: 20 step: 390, loss is 0.15992649
epoch time: 27890.857696533203, per step time: 71.51501973470052

• 遇到的问题以及解决方案

  https://gitee.com/mindspore/docs/blob/master/tutorials/source_zh_cn/advanced_use/nlp_application.md
  https://gitee.com/mindspore/mindspore/issues/I1IYSD?from=project-issue
  https://github.com/jroakes/glove-to-word2vec/blob/master/convert.py
  

评估

python eval.py --ckpt_path=./lstm-20-390.ckpt > out.eval.log 2>&1 & 

结果

============== Starting Testing ==============
============== Accuracy:{'acc': 0.9599358974358975} ==============

ok,结束炼丹