kevinthesun commented on issue #5243: [Frontend][TensorFlow]Improve TensorFlow
Static Shape Tensor Array
URL: https://github.com/apache/incubator-tvm/pull/5243#issuecomment-612260354
@masahi Sure. Please go ahead and review. I think a lot of logics can be
reused in pytorch.
kevinthesun commented on issue #5243: [Frontend][TensorFlow]Improve TensorFlow
Static Shape Tensor Array
URL: https://github.com/apache/incubator-tvm/pull/5243#issuecomment-612251085
@masahi @wweic PTAL
This is an automated
kevinthesun commented on issue #5243: [Frontend][TensorFlow]Improve TensorFlow
Static Shape Tensor Array
URL: https://github.com/apache/incubator-tvm/pull/5243#issuecomment-612176243
> @kevinthesun @wweic Is it reasonable to add "axis" parameter to tensor
array concat? I encountered a
kevinthesun commented on issue #5243: [Frontend][TensorFlow]Improve TensorFlow
Static Shape Tensor Array
URL: https://github.com/apache/incubator-tvm/pull/5243#issuecomment-610039127
https://github.com/apache/incubator-tvm/pull/5243/files#diff-eae8ecf976e0031823eeae454466f964R903
Take
kevinthesun commented on issue #5243: [Frontend][TensorFlow]Improve TensorFlow
Static Shape Tensor Array
URL: https://github.com/apache/incubator-tvm/pull/5243#issuecomment-610029111
@masahi The shape passed to ```get_var_static``` is for identification. For
tensor_get_data, it is just
kevinthesun commented on issue #5243: [Frontend][TensorFlow]Improve TensorFlow
Static Shape Tensor Array
URL: https://github.com/apache/incubator-tvm/pull/5243#issuecomment-609926012
> 1. When I append the tensor to tensor array (by concat), I can do infer
shape on the input tensor to get
kevinthesun commented on issue #5243: [Frontend][TensorFlow]Improve TensorFlow
Static Shape Tensor Array
URL: https://github.com/apache/incubator-tvm/pull/5243#issuecomment-609924079
> 2\. The output type of stack is currently
`static_tensor_float32_?_2_4_t[]` in my test. Is there a way