pub fn upload_tensor(rt: &CudaRuntime, tensor: &Tensor) -> Result<Tensor>Expand description
Upload a host tensor into a CubeCL-managed GPU allocation.
§Examples
use tenferro_gpu::{cuda::upload_tensor, cuda::CudaRuntime};
use tenferro_tensor::{Result, Tensor};
let _upload: fn(&CudaRuntime, &Tensor) -> Result<Tensor> = upload_tensor;§Errors
Returns crate::Error::RuntimeState when the source is backend-resident
or belongs to another placement, crate::Error::Unsupported for a dtype
unavailable in CubeCL, or crate::Error::BackendSource on allocation.