Skip to content

why the absolute fk does recurrent.LSTM() output a string instead of an array with array based training data??? #947

Open
@Jy738

Description

@Jy738

What is wrong?

recurrent.LSTM() outputs a string instead of an array with array based training data, this just seems like a major design flaw to me

Where does it happen?

running a basic instance of recurrent.lstm()

How do we replicate the issue?

  • Step 1 create a basic recurrent.lstm() instance
  • Step 2 trained it with incredibly basic data cus i thought i was going insane: net.train([{input:[1,2,3], output: [4,5,6]},{input:[4,5,6], output: [7,8,9]}], {
    iterations:100,
    log:true,
    errorThresh:0.0001
    });
  • Step 3 i logged the typeof net.run([1,2,3]), and got a string. dafuq?????????????????????????????????????????????????????

Expected behavior (i.e. solution)

get an array that i can run methods on and easily get the values of

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions