Currently the forward function of useExecutorchModule accepts the following:
forward: ( input: ETInput[], shape: number[][] | number[] )
This is pretty confusing and hard to work with for the developer using it and for us. This task aims to change the API to resemble how ExecuTorch tensors work, meaning that we would pass an array of Tensors to the forward function. Each Tensor would have its own dataPtr, scalar type, and shape.
Currently the forward function of useExecutorchModule accepts the following:
This is pretty confusing and hard to work with for the developer using it and for us. This task aims to change the API to resemble how ExecuTorch tensors work, meaning that we would pass an array of Tensors to the forward function. Each Tensor would have its own dataPtr, scalar type, and shape.