oreoscale.blogg.se - Tensorflow permute dimensions

Into non-overlapping blocks of size blockSize x blockSize The attr blockSize indicates the input block size and how the data isĬhunks of data of size blockSize * blockSize from depth are rearranged This op outputs a copy of the input tensor where values from the depthĭimension are moved in spatial blocks to the height and width dimensions. Rearranges data from depth into blocks of spatial data. , x.shape * blockShape - crops - crops,x.shape. , x.shape]Ĭrop the start and end of dimensions of reshapedPermutedĪccording to crops to produce the output of shape: * blockShape - crops - crops. Reshape permuted to produce reshapedPermuted of shape * blockShape. Permute dimensions of reshapedto produce permuted of shape, blockShape. , blockShape, batch / prod(blockShape), x.shape. This operation is equivalent to the following steps: That cropStart + cropEnd <= blockShape * inputShape Must have shape, all values must be >= 0.Ĭrops = specifies the amount to crop from inputĭimension i + 1, which corresponds to spatial dimension i. N-D with x.shape = + spatialShape + remainingShape, where spatialShape has M dimensions.

x ( tf.Tensor| TypedArray|Array) A tf.Tensor.

In any case, if I find a better solution I'll let you know. However, it didn't solve my slow performance issue. Thanks a lot! I managed to do it the same way you did and it worked. If you find another solution for the problem, would you please share it to me? Thank you so much. Therefore, I have to fix the xml file manually. Therefore, I manually edited the xml file as following:Īctually, I can't find the python file that custom layers and add the permute operation it in model_optimizer folder. I will post here when I find out more info.Īs the Shubha's advise in the previous comment, we need to add 2 reshape layer before and after the Permute layer. I will send the IR xml file to experts within the development team. Can you attach the xml file here which has the TensorIterator error ? I will file a bug for you.

Now I need to file another about the CPU documentation being inaccurate. So first, I have already filed a bug about the original problem for you. At least the Myriad document tells the truth about the "TensorIterator" primitive not being supported. When you get that error Unsupported primitive of type: it means that the operation does not exist in the model optimizer (and likely it doesn't exist at the Inference Engine too). I'm terribly sorry that you're having such problems even on CPU. It's entirely a Model Optimizer-to-Inference Engine function. You are right - you will not find anything related to this in Tensorflow.

is a function which Model Optimizer uses to define rules for layout conversation to make Inference Engine happy. Is it related to bug of Inferences Engine?Ĭould you please tell me what did you do exactly to solve the Permute error? I am still trying to solve this so I would really appreciate your feedback.ĭear quyet, PermuteAttrs.create_permute_attrs(node, attrs=). But, i got the error Unsupported primitive of type: TensorIterator name: bidirectional_1/while/LoopCond/TensorIteratorCondition_/TensorIterator. In addition, in the document said that the TensorIterator was supported by Inferences Engine in CPU. Can I do something to figure the error or wait the new release of Openvino toolkit?. I guess that the TensorIterator layer was unsupported by Inferences Engine in Neural Compute Stick as the document. Cannot convert layer "bidirectional_1/while/LoopCond/TensorIteratorCondition_/TensorIterator" due to unsupported layer type "TensorIterator" I manually edited the xml file and figure out the problems with Permute operation in Neural compute stick2.

Thank you so much for your close support. I searched about PermuteAttrs.create_permute_attrs(node, attrs=) in google but i can not find any documents about it. In my tensorflow graph, therer is not any Permute operation after the Squeeze operation. And I do not understand where you can add the Permute operation after the Squeeze operation. But i do not find the command add Permute layer in the python file except the command PermuteAttrs.create_permute_attrs(node, attrs=). Therefore, I try to edit by rewrite the squeeze operation (/deployment_tools/model_optimizer/mo/front/common/partial_infer/squeeze.py). However, If i manually edit the IR file, it will affect to the id of layer. As your advise, i will add reshape layer before and after Permute layer. I have a question about the command PermuteAttrs.create_permute_attrs(node, attrs=).