Bert4torch

Latest version: v0.5.6

Safety actively analyzes 723607 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 6 of 7

0.2.6

build_transformer_model需显式指定add_trainer才从BaseModel继承, 增加guwenbert, macbert,text2vec-bert-chinese, wobert预训练模型,允许position_ids从padding开始, transformer.configs支持点操作,可以使用torch4keras的Trainer(net)来初始化, 修复tokenizer的切分subtoken的bug, 允许embedding_size!=hidden_size

0.2.5

对抗训练从compile转为使用Callback来实现,修复1.7.1版本兼容bug, uie模型内置

0.2.4

增加deberta_v2模型

0.2.3

虚拟对抗VAT在多个ouput时支持指定,把Trainer抽象到[torch4keras](https://github.com/Tongjilibo/torch4keras)中,修复DP和DDP出现resume_epoch不存在的bug, tokenizer的never_split去除None, transformer_xl的bug, 增加gradient_checkpoint

0.2.2

修复t5的norm_mode问题,允许hidden_size不整除num_attention_heads, 支持多个schedule(如同时ema+warmup)

0.2.1

兼容torch<=1.7.1的torch.div无rounding_mode, 增加自定义metrics,支持断点续训,增加默认Logger和Tensorboard日志

Page 6 of 7

© 2025 Safety CLI Cybersecurity Inc. All Rights Reserved.