- 894174abfe add some warnings for useless function
frelam commented on pull request OpenI/MSAdapter#895
update autograd.Function and distribute api是的。 根据影响添加了warning或者报错。
2 days ago
frelam pushed to master0319 at frelam/MSAdapter
2 days ago
frelam commented on issue OpenI/MSAdapter#915
Cannot find reference 'mstorch_enable' in '__init__.py'是的, 如果是这样的话, 那需要等待mindspore 2.3的发布。 在您当前使用的版本的情况下, 可以用方式二进行修改, 可以参考下面链接中的“方式二”。 https://mindtorch.readthedocs.io/zh-cn/latest/docs/User_Guide_Import.html
6 days ago
frelam commented on issue OpenI/MSAdapter#915
Cannot find reference 'mstorch_enable' in '__init__.py'> 无法引入替换导入模块路径。 > Windows11+Python3.9+conda > mindspore 2.2.13 > mindtorch 0.2.1 > ![image](/attachments/bb17fcdf-0512-4274-ac05-fbc797e83b66) 您好。 该“mstorch_enable”功能, 在mindtorch 0.2.1 上未支持, 所以会报以上错误。 当前有两种方式支持该功能: 1. 更新到mindspore 2.3最新版本, 下载mindtorch ‘master’分支, 然后通过源码安装。 2. 仍然采用mindspore 2.2版本, 下载mindtorch ‘r0.3_import’分支,然后通过源码安装。
1 week ago
frelam commented on pull request OpenI/MSAdapter#911
[WIP]fix tensor.typecuda 标记的传递性暂无法解决。
1 week ago
frelam commented on pull request OpenI/MSAdapter#895
update autograd.Function and distribute api无法提前做 。 因为grad每次都是新的tensor对象, 需要每次都重新获取。
2 weeks ago
frelam commented on pull request OpenI/MSAdapter#895
update autograd.Function and distribute api已添加。 已知两点限制: 1. 当返回梯度shape与参数shape不一致时, 自动规约不支持。 2. Function.apply 图模式下不支持。
2 weeks ago
frelam commented on pull request OpenI/MSAdapter#895
update autograd.Function and distribute api在Ascend上, 如果输入的dtype, 不在通信算子支持的 (int8, int32, float16, float32, bfloat16) 类型里面, 会触发类型转换, 其余情况相当于透传。
2 weeks ago
frelam commented on pull request OpenI/MSAdapter#895
update autograd.Function and distribute api用动态生成函数的方式, 在__init__阶段自动生成了bprop, 当前用法可以与pytorch相同了。
2 weeks ago
frelam pushed to master0319 at frelam/MSAdapter
2 weeks ago
frelam pushed to master0319 at frelam/MSAdapter
2 weeks ago
frelam pushed to master0319 at frelam/MSAdapter
2 weeks ago
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》