[Layer] Remove Tensor setDataType() usage

[djeong20]

In several layers, there are attempts to change the data type of a Tensor object after initializing it.
This is currently possible but can cause issues down the line (e.g., treat FloatTensor object as HalfTensor).
As such, the setDataType() method will be removed and considered not to be used in future updates.
Instead, users will need to provide the desired data type when creating a new tensor.

Self-evaluation:

1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

[taos-ci]

📝 TAOS-CI Version: 1.5.20200925. Thank you for submitting PR #2498. Please a submit 1commit/1PR (one commit per one PR) policy to get comments quickly from reviewers. Your PR must pass all verificiation processes of cibot before starting a review process from reviewers. If you are new member to join this project, please read manuals in documentation folder and wiki page. In order to monitor a progress status of your PR in more detail, visit http://ci.nnstreamer.ai/.

[taos-ci]

@djeong20, 💯 All CI checkers are successfully verified. Thanks.

[taos-ci]

@djeong20, 💯 All CI checkers are successfully verified. Thanks.

[skykongkong8]

Think this is quite important point.
I will keep this in mind in the future development of myself as well :)

[myungjoo]

Question:

In a network, is the Tensor name required to be unique?

  Tensor tensor =
    Tensor("some_name", value.getFormat(), value.getDataType());

(I hope not..)

[jijoongmoon]

We can set the format and data_type with weight_tensor_type. No need to set separately. Also, we do not need to set the tensor name always. I think when we create tensor like here, Tensor empty, then it does not allocate the buffer. So.. I think it is ok to set the Tensor type here. The reason we set like this.. we do not want to allocate a buffer at this time.

[djeong20]

As we discussed offline, Tensor empty; would create FloatTensor as a default.

nntrainer::Tensor t;
if (t.getDataType() == nntrainer::Tdatatype::FP32) {
  // This would be executed
  std::cout << "FP32" << std::endl;
} else {
  std::cout << "FP16" << std::endl;
}

I will make an additional Tensor constructor that only takes data type without allocating a buffer (e.g., Tensor empty = Tensor(weight_tensor_type.data_type);)

[djeong20]

Question:

In a network, is the Tensor name required to be unique?

  Tensor tensor =
    Tensor("some_name", value.getFormat(), value.getDataType());

(I hope not..)

I don't think the name has to be unique.

[myungjoo]

Question:
In a network, is the Tensor name required to be unique?

  Tensor tensor =
    Tensor("some_name", value.getFormat(), value.getDataType());

(I hope not..)

I don't think the name has to be unique.

If this is for debugging and developer referencing, it's ok.

[jijoongmoon]

LGTM