Skip to content

Releases: Jasonsey/Fern

1.1.1.dev4(2021-07-20)

Pre-release

Choose a tag to compare

@Jasonsey Jasonsey released this 20 Jul 14:21
  • Fix Bugs:
    • 默认logging的名字改为Fern

1.1.1.dev3(2021-07-20)

Pre-release

Choose a tag to compare

@Jasonsey Jasonsey released this 20 Jul 13:55
  • Fix Bugs:
    • 修复数据集分割时,大于1浮点数问题

1.1.1.dev2(2021-07-19)

Pre-release

Choose a tag to compare

@Jasonsey Jasonsey released this 19 Jul 05:34
  • Fix Bugs:
    • 修复logger继承 root logger带来的负面影响

1.1.1.dev1

1.1.1.dev1 Pre-release
Pre-release

Choose a tag to compare

@Jasonsey Jasonsey released this 18 Jul 10:36
  • Fix Bug:
    • 修复Optional的错误

1.1.0 (2021-07-14)

Choose a tag to compare

@Jasonsey Jasonsey released this 14 Jul 11:11
  • New Feature:

    • 添加Bert token方法
    • 添加yaml文件读取工具
    • 添加keras模型快速编译函数
    • 添加keras模型基于dataset的训练脚本
  • Fix Bug:

    • 修复logging模块重复输出日志的bug

1.0.0 (2021-04-27)

Choose a tag to compare

@Jasonsey Jasonsey released this 27 Apr 15:48
  • New Feature:
    • 按照模块功能, 重新调整模型的结构, 调整后的模块为:
    • data: 数据预处理的各种功能函数
    • models: 自定义层, 常用模型
    • utils: 全局通用函数
    • metrics: 自定义衡量函数
    • pipeline: 流程控制函数
    • train: 自定义训练函数

0.9.0 (2020-12-13)

Choose a tag to compare

@Jasonsey Jasonsey released this 13 Dec 04:27
  • New Feature:
    • Add setting GPU function
    • Add FernSimpleTrainer
  • Style:
    • Update trainer's annotation

0.8.1 (2020-11-16)

Choose a tag to compare

@Jasonsey Jasonsey released this 16 Nov 13:12
  • Fix Bug:
    • Map function error without send a function
  • New Feature:
    • Support parallel_apply for FernDataFrame
    • Update FernCleaner to process data with parallel process
  • Test:
    • Add test_parallel_apply method to test parallel_apply

0.8.0 (2020-11-16)

Choose a tag to compare

@Jasonsey Jasonsey released this 16 Nov 07:53
  • New Feature:
    • Support parallel_map for FernSeries
  • Style:
    • Use Google annotation style in data.py

0.7.0 (2020-11-11)

Choose a tag to compare

@Jasonsey Jasonsey released this 11 Nov 13:01
  • New Feature:
    • Support multi inputs and outputs with dictionary format data
    • Label weight is no longer recommended, but dataset balancing is
  • Style:
    • Use a stub file to annotate the parameter types of the file train.py