模型训练成功,本地infer失败
Created by: JCJTrying
问题描述: embedding 矩阵使用is_distributed=True后,训练产出的模型在本地infer失败。is_distributed=False时,训练产出的模型在本地能正常infer
- 版本、环境信息: 1)PaddlePaddle版本:paddle-fluid-v1.2.0
-错误信息 Traceback (most recent call last): File "infer.py", line 236, in main() File "infer.py", line 232, in main infer(use_cuda) File "infer.py", line 211, in infer fetch_list=fetch_targets) File "/home/zhangshaochun/jiangchangjun/deepfm/python27-gcc482/lib/python2.7/site-packages/paddle/fluid/executor.py", line 472, in run self.executor.run(program.desc, scope, 0, True, True) paddle.fluid.core.EnforceNotMet: holder_ should not be null Tensor holds no memory. Call Tensor::mutable_data first. at [/script/paddle/paddle/fluid/framework/tensor.cc:22]
--PaddlePaddle Call Stacks: 0 0x7fb056adb3c6p paddle::platform::EnforceNotMet::EnforceNotMet(std::__exception_ptr::exception_ptr, char const*, int) + 486