yuyijie1995 / 2018-ccf-bdci-alibaba-3dpoint-cloud-segmentation Goto Github PK

我在代码复现时，运行到eval.py 第76行 boxes, probs, labels, *masks = model_func(resized_img)
时，输出一句check failed就退出了，没有抛出异常，请问有人知道是什么原因么？ 日志输出如下：
F d:\build\tensorflow\tensorflow-r1.9\tensorflow\core\kernels\mkl_avgpooling_op.cc:146] Check failed: dnnExecute_F32(mkl_context.prim_pooling_fwd, mkl_context.pooling_res) == E_SUCCESS (-1 vs. 0)

为什么下载好的权重不能匹配train_11_5.py啊，还是使用方法有什么问题吗，一直在报错：

InvalidArgumentError: Restoring from checkpoint failed. This is most likely due to a mismatch between the current graph and the graph from the checkpoint. Please ensure that you have not altered the graph expected based on the checkpoint. Original error:

Assign requires shapes of both tensors to match. lhs shape= [8] rhs shape= [2]
[[node 2267349901744/Assign_12 (defined at E:\Anaconda3\envs\my_Tensorflow-gpu\lib\site-packages\tensorpack\tfutils\sessinit.py:111) ]]
[[node 2267349901744/RestoreV2 (defined at E:\Anaconda3\envs\my_Tensorflow-gpu\lib\site-packages\tensorpack\tfutils\sessinit.py:111) ]]

emmmmm所以既然不能用权重，那怎么训练（就是运行命令，数据存储位置之类的）能说明一下吗