Fix register_float_tensor to use the correct dtype (#3774)

* Fix register_float_tensor to use the correct dtype

* Fix clippy

---------

Co-authored-by: Guillaume Lagrange <[email protected]>

Author: A2va

crates/burn-remote/src/client/runner.rs

@@ -3,7 +3,7 @@ use burn_communication::{Address, ProtocolClient, data_service::TensorTransferId
 use burn_ir::TensorIr;
 use burn_router::{MultiBackendBridge, RouterTensor, RunnerClient, get_client};
 use burn_tensor::{
-    DType, TensorData,
+    TensorData,
     backend::{DeviceId, DeviceOps},
 };
 use std::{
@@ -65,9 +65,9 @@ impl RunnerClient for RemoteClient {
     fn register_float_tensor(
         &self,
         shape: Vec<usize>,
-        _dtype: burn_tensor::FloatDType,
+        dtype: burn_tensor::FloatDType,
     ) -> RouterTensor<Self> {
-        self.register_empty_tensor(shape, DType::F32)
+        self.register_empty_tensor(shape, dtype.into())
     }
 
     fn device(&self) -> Self::Device {