masahi edited a comment on issue #5243: [Frontend][TensorFlow]Improve 
TensorFlow Static Shape Tensor Array
URL: https://github.com/apache/incubator-tvm/pull/5243#issuecomment-612270589
 
 
   > @masahi To support different axis we need to change both 
`define_tensor_concatenate` and `define_tensor_array_concat` to support axis 
arguments. The main thing we need to take care is the output shape.
   
   Ok for now I went an easy route of just defining concat_last op. It seems to 
work, but I'm getting the following typing error:
   
   ```
   ...
     %101 = @map(tensor_constructor_float32_?_2_4(Tensor[(?, 2, 4), float32]), 
%100);
     %102 = @tensor_array_concat_last_float32_?_2_?(%101) unable to unify: 
`static_tensor_float32_?_2_4_t` and `static_tensor_float32_?_2_?_t`; ;
     %103 = @tensor_get_data_float32_?_2_?(%102);
   ...
   ```
   
   The first axis is already Any by tensor array stack. Now I'm trying to 
concat (?, 2, 4) tensors along -1 axis to get (?, 2, ?) tensor. Is this 
possible? They typing error suggests no.
   
   UPDATE: Solved by mapping tensor_constructor with concat-ed shape (?, 2, ?):
   ```
     %101 = @map(tensor_constructor_float32_?_2_?(Tensor[(?, 2, ?), float32]), 
%100) /* ty=List[static_tensor_float32_?_2_?_t[]] */;
     %102 = @tensor_array_concat_last_float32_?_2_?(%101) /* 
ty=static_tensor_float32_?_2_?_t[] */;
     %103 = @tensor_get_data_float32_?_2_?(%102) /* ty=Tensor[(?, 2, ?), 
float32] */;
   ```

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
[email protected]


With regards,
Apache Git Services

Reply via email to