如何 把VarBase 转换为LodTensor

[zzz2010]

你好， 我想直接在GPU上把VarBase转换为LodTensor，而不是先转cpu numpy 再create_lod_tensor. 请问有方法吗？ 谢谢

[paddle-bot-old]

您好，我们已经收到了您的问题，会安排技术人员尽快解答您的问题，请耐心等待。请您再次检查是否提供了清晰的问题描述、复现代码、环境&版本、报错信息等。同时，您也可以通过查看官网API文档、常见问题、历史Issue、AI社区来寻求解答。祝您生活愉快～

Hi! We've received your issue and please be patient to get responded. We will arrange technicians to answer your questions as soon as possible. Please make sure that you have posted enough message to demo your request. You may also check out the API，FAQ，Github Issue and AI community to get the answer.Have a nice day!

[danleifeng]

您好，可以试试这种方法，得到的是一个lodtensor
a.value().get_tensor()

[zzz2010]

@danleifeng 谢谢。 想追加一个问题， 如何 把 paddle.fluid.core_avx.Tensor 转回Varbase？
我用fluid.core.from_dlpack(dlpack) 得到 paddle.fluid.core_avx.Tensor ，希望变回Varbase

[danleifeng]

可以使用paddle.to_variable(a)接口

[zzz2010]

可以使用paddle.to_variable(a)接口

@danleifeng 我尝试了 fluid.dygraph.to_variable() , 和paddle.to_tensor() 都不支持paddle.fluid.core_avx.Tensor

测试代码

import paddle
tensor=paddle.randn((3,4))
dlpack=tensor.value().get_tensor()._to_dlpack()
tensor_from_dlpack = paddle.fluid.core.from_dlpack(dlpack)
paddle.fluid.dygraph.to_variable(tensor_from_dlpack)

错误：

TypeError: __init__(): incompatible constructor arguments. The following argument types are supported:
   1. paddle.fluid.core_avx.VarBase()
   2. paddle.fluid.core_avx.VarBase(arg0: paddle::framework::proto::VarType_Type, arg1: List[int], arg2: handle, arg3: paddle::framework::proto::VarType_Type, arg4: bool)
   3. paddle.fluid.core_avx.VarBase(value: array, place: paddle::platform::CPUPlace, persistable: bool = False, zero_copy: bool = False, name: str = '', stop_gradient: int = -1)
   4. paddle.fluid.core_avx.VarBase(value: array, place: paddle::platform::XPUPlace, persistable: bool = False, zero_copy: bool = False, name: str = '', stop_gradient: int = -1)
   5. paddle.fluid.core_avx.VarBase(value: array, place: paddle::platform::CUDAPlace, persistable: bool = False, zero_copy: bool = False, name: str = '', stop_gradient: int = -1)
   6. paddle.fluid.core_avx.VarBase(value: array, place: paddle::platform::CUDAPinnedPlace, persistable: bool = False, zero_copy: bool = False, name: str = '', stop_gradient: int = -1)
   7. paddle.fluid.core_avx.VarBase(value: array)
   8. paddle.fluid.core_avx.VarBase(tensor: paddle::framework::LoDTensor)
   9. paddle.fluid.core_avx.VarBase(**kwargs)
Invoked with: <paddle.fluid.core_avx.Tensor object at 0x000001FA6AD11CB0>

[zhwesky2010]

@zzz2010 构造这里有个问题，不支持core.Tensor，只支持core.LoDTensor，后面我们修复一下

[zhwesky2010]

@zzz2010 构造这里有个问题，不支持core.Tensor，只支持core.LoDTensor，后面我们修复一下

#33430 修复了这个问题

[paddle-bot-old]

Since you haven't replied for more than a year, we have closed this issue/pr.
If the problem is not solved or there is a follow-up one, please reopen it at any time and we will continue to follow up.
由于您超过一年未回复，我们将关闭这个issue/pr。
若问题未解决或有后续问题，请随时重新打开，我们会继续跟进。