You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
i-robot 8a48a3b66a
!3011 fix issue I9NQQW
2 days ago
..
core polynomialwithwarmuplr bugfix 1 week ago
dataset add wikipedia dataset preprocess and fix dataset issue 1 week ago
experimental !2846 fix ring attention bug 3 weeks ago
generation !3011 fix issue I9NQQW 2 days ago
models !3011 fix issue I9NQQW 2 days ago
modules !2941 【Bugfix】修复moe中dtype硬编码问题 1 week ago
pet 修复lora推理参数对不齐问题 2 weeks ago
pipeline 修复glm32k多batch推理长度不一致的问题,readme补充推理说明 3 weeks ago
tools 多副本并行加cell共享校验 1 week ago
trainer fixed 3df0904 from https://gitee.com/Lin-Bert/transformer/pulls/2952 1 week ago
utils 权重数据类型转换设置为在cpu上处理 1 week ago
wrapper fixed 2cf3fab from https://gitee.com/huanglei_Sorry/mindformers/pulls/2422 2 months ago
__init__.py add adamwzero2 optimizer 3 weeks ago
auto_class.py 接口统一 1 month ago
mindformer_book.py !3031 【check】codellama配置改名 3 days ago
model_runner.py fix dosample 5 days ago
version_control.py 【dev】删除gpt2以及multiheadattention中对ifa的调用 3 weeks ago

MindSpore Transformers套件的目标是构建一个大模型训练、微调、评估、推理、部署的全流程开发套件: 提供业内主流的Transformer类预训练模型和SOTA下游任务应用,涵盖丰富的并行特性。期望帮助用户轻松的实现大模型训练和创新研发。

Jupyter Notebook Python Markdown Shell