diff --git a/crates/bunsen-firehose-image/src/colortype_support.rs b/crates/bunsen-firehose-image/src/colortype_support.rs
index 17c08d8..d3cddba 100644
--- a/crates/bunsen-firehose-image/src/colortype_support.rs
+++ b/crates/bunsen-firehose-image/src/colortype_support.rs
@@ -260,43 +260,43 @@ mod tests {
 
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::L8),
-            DynamicImage::from(source.clone().to_luma8())
+            DynamicImage::from(source.to_luma8())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::La8),
-            DynamicImage::from(source.clone().to_luma_alpha8())
+            DynamicImage::from(source.to_luma_alpha8())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::Rgb8),
-            DynamicImage::from(source.clone().to_rgb8())
+            DynamicImage::from(source.to_rgb8())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::Rgba8),
-            DynamicImage::from(source.clone().to_rgba8())
+            DynamicImage::from(source.to_rgba8())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::L16),
-            DynamicImage::from(source.clone().to_luma16())
+            DynamicImage::from(source.to_luma16())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::La16),
-            DynamicImage::from(source.clone().to_luma_alpha16())
+            DynamicImage::from(source.to_luma_alpha16())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::Rgb16),
-            DynamicImage::from(source.clone().to_rgb16())
+            DynamicImage::from(source.to_rgb16())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::Rgba16),
-            DynamicImage::from(source.clone().to_rgba16())
+            DynamicImage::from(source.to_rgba16())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::Rgb32F),
-            DynamicImage::from(source.clone().to_rgb32f())
+            DynamicImage::from(source.to_rgb32f())
         );
         assert_eq!(
             convert_to_colortype(source.clone(), ColorType::Rgba32F),
-            DynamicImage::from(source.clone().to_rgba32f())
+            DynamicImage::from(source.to_rgba32f())
         );
     }
 }
diff --git a/crates/bunsen/src/blocks/images/drop/drop_path.rs b/crates/bunsen/src/blocks/images/drop/drop_path.rs
index 73e5da0..d5d3cbc 100644
--- a/crates/bunsen/src/blocks/images/drop/drop_path.rs
+++ b/crates/bunsen/src/blocks/images/drop/drop_path.rs
@@ -282,7 +282,7 @@ mod tests {
                 Tensor::<B, 3>::from_data([[[1.0]], [[0.0]], [[1.0]]], device)
             },
         );
-        res.to_data().assert_eq(&x.clone().to_data(), true);
+        res.to_data().assert_eq(&x.to_data(), true);
 
         // No-op case: training, but drop_prob = 0.0
         let training = true;
@@ -299,7 +299,7 @@ mod tests {
                 Tensor::<B, 3>::from_data([[[1.0]], [[0.0]], [[1.0]]], device)
             },
         );
-        res.to_data().assert_eq(&x.clone().to_data(), true);
+        res.to_data().assert_eq(&x.to_data(), true);
 
         // Training, but no scaling
         let training = true;
diff --git a/crates/bunsen/src/blocks/rnn/lstm/ext_lstm_state.rs b/crates/bunsen/src/blocks/rnn/lstm/ext_lstm_state.rs
index 9497297..5469787 100644
--- a/crates/bunsen/src/blocks/rnn/lstm/ext_lstm_state.rs
+++ b/crates/bunsen/src/blocks/rnn/lstm/ext_lstm_state.rs
@@ -267,8 +267,8 @@ mod tests {
         let device = Default::default();
         let state: ExtLstmState<B, 3> = random_state([2, 3, 4], &device);
 
-        let expected_cell = state.cell.clone().to_data();
-        let expected_hidden = state.hidden.clone().to_data();
+        let expected_cell = state.cell.to_data();
+        let expected_hidden = state.hidden.to_data();
 
         let (cell, hidden) = state.unpack();
 
@@ -354,7 +354,7 @@ mod tests {
         let device = Default::default();
         let state: ExtLstmState<B, 2> = random_state([2, 3], &device);
 
-        let expected_cell = state.cell.clone().to_data();
+        let expected_cell = state.cell.to_data();
 
         let roundtrip: ExtLstmState<B, 2> = state.unsqueeze_dim::<3>(1).squeeze_dim(1);
 
@@ -384,8 +384,8 @@ mod tests {
         let device = Default::default();
         let state: ExtLstmState<B, 2> = random_state([2, 3], &device);
 
-        let expected_cell = state.cell.clone().to_data();
-        let expected_hidden = state.hidden.clone().to_data();
+        let expected_cell = state.cell.to_data();
+        let expected_hidden = state.hidden.to_data();
 
         // The shape passed here is intentionally different; it must be ignored.
         let unwrapped = Some(state).unwrap_or_initial([9, 9], &device);
diff --git a/crates/bunsen/src/burner/tensor/tensor_op_ext.rs b/crates/bunsen/src/burner/tensor/tensor_op_ext.rs
index e6ab0b2..6294e33 100644
--- a/crates/bunsen/src/burner/tensor/tensor_op_ext.rs
+++ b/crates/bunsen/src/burner/tensor/tensor_op_ext.rs
@@ -12,6 +12,7 @@ use burn::{
         BasicOps,
         Bool,
         DType,
+        DataError,
         Element,
         Float,
         Int,
@@ -80,13 +81,19 @@ where
     K: BasicOps<B>,
     K::Elem: Element,
 {
-    /// Copies the current `Tensor` into `TensorData`; converts the dtype.
+    /// Copies the current `Tensor` into a `TensorData`; converts the dtype.
+    ///
+    /// By contract, this will yield the same result as
+    /// `tensor.to_data().convert::<E>(dtype)`.
     ///
     /// The conversion is a no-op if the dtype is the same as the current dtype.
-    fn to_data_convert<E: Element>(&self) -> TensorData;
+    fn to_data_as<E: Element>(&self) -> TensorData;
 
     /// Copies the current `Tensor` into `TensorData`; converts the dtype.
     ///
+    /// By contract, this will yield the same result as
+    /// `tensor.to_data().convert_dtype(dtype)`.
+    ///
     /// The conversion is a no-op if the dtype is the same as the current dtype.
     fn to_data_cast(
         &self,
@@ -95,11 +102,17 @@ where
 
     /// Converts the current `Tensor` into `TensorData`; converts the dtype.
     ///
+    /// By contract, this will yield the same result as
+    /// `tensor.into_data().convert::<E>(dtype)`.
+    ///
     /// The conversion is a no-op if the dtype is the same as the current dtype.
-    fn into_data_convert<E: Element>(self) -> TensorData;
+    fn into_data_as<E: Element>(self) -> TensorData;
 
     /// Converts the current `Tensor` into `TensorData`; converts the dtype.
     ///
+    /// By contract, this will yield the same result as
+    /// `tensor.into_data().convert_dtype(dtype)`.
+    ///
     /// The conversion is a no-op if the dtype is the same as the current dtype.
     fn into_data_cast(
         self,
@@ -113,8 +126,8 @@ where
     K: BasicOps<B>,
     K::Elem: Element,
 {
-    fn to_data_convert<E: Element>(&self) -> TensorData {
-        self.to_data().convert::<E>()
+    fn to_data_as<E: Element>(&self) -> TensorData {
+        self.to_data_cast(E::dtype())
     }
 
     fn to_data_cast(
@@ -124,8 +137,8 @@ where
         self.to_data().convert_dtype(dtype)
     }
 
-    fn into_data_convert<E: Element>(self) -> TensorData {
-        self.into_data().convert::<E>()
+    fn into_data_as<E: Element>(self) -> TensorData {
+        self.into_data_cast(E::dtype())
     }
 
     fn into_data_cast(
@@ -278,6 +291,74 @@ where
     }
 }
 
+/// Extension trait for `TensorData` that provides additional methods.
+pub trait TensorDataToVecAsExt {
+    /// Cast the data to a new dtype.
+    ///
+    /// TODO: Implement proper error handling in `TensorData`.
+    ///
+    /// # Returns
+    /// Ok(data) on success, (Currently) panics on failure.
+    fn try_cast(
+        self,
+        dtype: DType,
+    ) -> Result<TensorData, DataError>;
+
+    /// Convert the data to a new dtype.
+    ///
+    /// TODO: Implement proper error handling in `TensorData`.
+    ///
+    /// By contract, this is equivalent to:
+    /// `data.try_cast(E::dtype())`
+    ///
+    /// # Returns
+    /// Ok(data) on success, (Currently) panics on failure.
+    fn try_convert<E: Element>(self) -> Result<TensorData, DataError>;
+
+    /// Copy and convert the data to a [`Vec<E>`].
+    ///
+    /// By contract, this is equivalent to:
+    /// `data.clone().into_vec_as::<E>()`
+    ///
+    /// Particular conversions may provide more efficient implementations.
+    ///
+    /// # Returns
+    /// `Ok(vec)` on success, or an error if the conversion fails.
+    fn to_vec_as<E: Element>(&self) -> Result<Vec<E>, DataError>;
+
+    /// Convert the data to [`Vec<E>`].
+    ///
+    /// By contract, this is equivalent to:
+    /// `data.try_convert::<E>()?.to_vec::<E>()`
+    ///
+    /// Particular conversions may provide more efficient implementations.
+    ///
+    /// # Returns
+    /// `Ok(vec)` on success, or an error if the conversion fails.
+    fn into_vec_as<E: Element>(self) -> Result<Vec<E>, DataError>;
+}
+
+impl TensorDataToVecAsExt for TensorData {
+    fn try_cast(
+        self,
+        dtype: DType,
+    ) -> Result<TensorData, DataError> {
+        Ok(self.convert_dtype(dtype))
+    }
+
+    fn try_convert<E: Element>(self) -> Result<TensorData, DataError> {
+        self.try_cast(E::dtype())
+    }
+
+    fn to_vec_as<E: Element>(&self) -> Result<Vec<E>, DataError> {
+        self.clone().into_vec_as::<E>()
+    }
+
+    fn into_vec_as<E: Element>(self) -> Result<Vec<E>, DataError> {
+        self.try_convert::<E>()?.to_vec::<E>()
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use burn::tensor::{
@@ -381,4 +462,41 @@ mod tests {
             .to_data()
             .assert_eq(&TensorData::from([3]), false);
     }
+
+    #[test]
+    fn test_to_vec_as() {
+        let data = TensorData::from([0.0f32, 1.0, 2.5]);
+
+        // Same-dtype copy.
+        assert_eq!(data.to_vec_as::<f32>().unwrap(), vec![0.0f32, 1.0, 2.5]);
+
+        // Widening cast (different element size).
+        assert_eq!(data.to_vec_as::<f64>().unwrap(), vec![0.0f64, 1.0, 2.5]);
+
+        // Float to int cast (same element size) truncates.
+        assert_eq!(data.to_vec_as::<i32>().unwrap(), vec![0i32, 1, 2]);
+
+        // The source data is borrowed, not consumed.
+        data.assert_eq(&TensorData::from([0.0f32, 1.0, 2.5]), true);
+    }
+
+    #[test]
+    fn test_into_vec_as() {
+        let data = TensorData::from([0i32, 1, 2, 3]);
+
+        // Same-dtype conversion.
+        assert_eq!(
+            data.clone().into_vec_as::<i32>().unwrap(),
+            vec![0i32, 1, 2, 3]
+        );
+
+        // Int to float cast.
+        assert_eq!(
+            data.clone().into_vec_as::<f32>().unwrap(),
+            vec![0.0f32, 1.0, 2.0, 3.0]
+        );
+
+        // Narrowing int cast.
+        assert_eq!(data.into_vec_as::<u8>().unwrap(), vec![0u8, 1, 2, 3]);
+    }
 }
diff --git a/crates/bunsen/src/kits/bimm/resnet/blocks/layer_block.rs b/crates/bunsen/src/kits/bimm/resnet/blocks/layer_block.rs
index 2d8bf7d..8df5ff5 100644
--- a/crates/bunsen/src/kits/bimm/resnet/blocks/layer_block.rs
+++ b/crates/bunsen/src/kits/bimm/resnet/blocks/layer_block.rs
@@ -467,6 +467,7 @@ mod tests {
     use crate::{
         contracts::assert_shape_contract,
         kits::bimm::resnet::blocks::BasicBlockConfig,
+        prelude::*,
         support::testing::PerformanceBackend,
     };
 
@@ -556,7 +557,7 @@ mod tests {
             expected = block.forward(expected);
         }
         output
-            .to_data()
-            .assert_approx_eq::<F>(&expected.to_data(), Tolerance::default());
+            .to_data_as::<F>()
+            .assert_approx_eq::<F>(&expected.to_data_as::<F>(), Tolerance::default());
     }
 }
diff --git a/crates/bunsen/src/kits/bimm/swin/v2/blocks/window_attention/pos_grid.rs b/crates/bunsen/src/kits/bimm/swin/v2/blocks/window_attention/pos_grid.rs
index 5335758..dc886fe 100644
--- a/crates/bunsen/src/kits/bimm/swin/v2/blocks/window_attention/pos_grid.rs
+++ b/crates/bunsen/src/kits/bimm/swin/v2/blocks/window_attention/pos_grid.rs
@@ -284,7 +284,7 @@ mod tests {
 
         let device = Default::default();
         let rel = window_attention_relative_position_index::<B>(window_shape, &device);
-        rel.clone().to_data().assert_eq(
+        rel.to_data().assert_eq(
             &TensorData::from([
                 [7, 6, 5, 2, 1, 0],
                 [8, 7, 6, 3, 2, 1],
diff --git a/crates/bunsen/src/kits/sims/conway/life2d.rs b/crates/bunsen/src/kits/sims/conway/life2d.rs
index 38ca63d..581d3af 100644
--- a/crates/bunsen/src/kits/sims/conway/life2d.rs
+++ b/crates/bunsen/src/kits/sims/conway/life2d.rs
@@ -88,7 +88,7 @@ where
 
     state
         .slice(slices)
-        .to_data_convert::<bool>()
+        .to_data_as::<bool>()
         .to_vec()
         .unwrap()
         .chunks(w)
diff --git a/crates/bunsen/src/kits/sims/lbm/d2q9/collision.rs b/crates/bunsen/src/kits/sims/lbm/d2q9/collision.rs
index 6861c1a..9e273a3 100644
--- a/crates/bunsen/src/kits/sims/lbm/d2q9/collision.rs
+++ b/crates/bunsen/src/kits/sims/lbm/d2q9/collision.rs
@@ -119,7 +119,7 @@ mod tests {
         // Invariant: density(collision(dist, param)) == density(dist)
         density(col_dist.clone())
             .to_data()
-            .assert_approx_eq::<f32>(&rho.clone().to_data(), Tolerance::default());
+            .assert_approx_eq::<f32>(&rho.to_data(), Tolerance::default());
 
         // With Correction
         {
diff --git a/crates/bunsen/src/kits/sims/lbm/d2q9/space.rs b/crates/bunsen/src/kits/sims/lbm/d2q9/space.rs
index 83da824..b8332e3 100644
--- a/crates/bunsen/src/kits/sims/lbm/d2q9/space.rs
+++ b/crates/bunsen/src/kits/sims/lbm/d2q9/space.rs
@@ -395,7 +395,7 @@ mod tests {
 
         let momentum = macroscopic_momentum(dist.clone(), lbm_tables.e_vec());
 
-        momentum.clone().to_data().assert_approx_eq::<f32>(
+        momentum.to_data().assert_approx_eq::<f32>(
             &Tensor::<B, 3>::from_data([[[-1., -1.], [15., 5.]]], &device).to_data(),
             Tolerance::default(),
         );
@@ -404,7 +404,7 @@ mod tests {
 
         let rho_data = rho.to_data().to_vec::<f32>().unwrap();
 
-        u.clone().to_data().assert_approx_eq::<f32>(
+        u.to_data().assert_approx_eq::<f32>(
             &Tensor::<B, 3>::from_data(
                 [[
                     [-1. / rho_data[0], -1. / rho_data[0]],
@@ -418,7 +418,7 @@ mod tests {
 
         let v_sq = velocity_squared(u.clone());
 
-        v_sq.clone().to_data().assert_approx_eq::<f32>(
+        v_sq.to_data().assert_approx_eq::<f32>(
             &Tensor::<B, 2>::from_data(
                 [[
                     (1. + 1.) / rho_data[0].powi(2),
diff --git a/crates/bunsen/src/kits/sims/lbm/d2q9/thermal.rs b/crates/bunsen/src/kits/sims/lbm/d2q9/thermal.rs
index ff86693..82b94dc 100644
--- a/crates/bunsen/src/kits/sims/lbm/d2q9/thermal.rs
+++ b/crates/bunsen/src/kits/sims/lbm/d2q9/thermal.rs
@@ -165,7 +165,7 @@ mod tests {
 
         let parts = ldv_projection(e.clone(), u.clone());
 
-        parts.clone().to_data().assert_approx_eq::<f32>(
+        parts.to_data().assert_approx_eq::<f32>(
             &Tensor::<B, 5>::from_data(
                 [[
                     [
@@ -187,7 +187,7 @@ mod tests {
 
         let e_u = lattice_dot_velocity(u.clone(), e.clone());
 
-        e_u.clone().to_data().assert_approx_eq::<f32>(
+        e_u.to_data().assert_approx_eq::<f32>(
             &parts.sum_dim(4).squeeze_dims::<4>(&[4]).to_data(),
             Tolerance::default(),
         );
diff --git a/crates/bunsen/src/kits/speech/silero_vad/blocks/module.rs b/crates/bunsen/src/kits/speech/silero_vad/blocks/module.rs
index df68689..a2e3c76 100644
--- a/crates/bunsen/src/kits/speech/silero_vad/blocks/module.rs
+++ b/crates/bunsen/src/kits/speech/silero_vad/blocks/module.rs
@@ -858,7 +858,10 @@ mod tests {
     };
 
     use super::*;
-    use crate::support::testing::PerformanceBackend;
+    use crate::{
+        prelude::*,
+        support::testing::PerformanceBackend,
+    };
 
     type B = PerformanceBackend;
 
@@ -1060,11 +1063,11 @@ mod tests {
 
         let tol = Tolerance::<F>::default();
         seq_probs
-            .into_data()
-            .assert_approx_eq::<F>(&step_probs.into_data(), tol);
+            .into_data_as::<F>()
+            .assert_approx_eq::<F>(&step_probs.into_data_as::<F>(), tol);
         seq_state
-            .into_data()
-            .assert_approx_eq::<F>(&state.into_data(), tol);
+            .into_data_as::<F>()
+            .assert_approx_eq::<F>(&state.into_data_as::<F>(), tol);
     }
 
     #[test]
diff --git a/crates/bunsen/src/kits/speech/silero_vad/cross_test.rs b/crates/bunsen/src/kits/speech/silero_vad/cross_test.rs
index 13eb963..3bd2262 100644
--- a/crates/bunsen/src/kits/speech/silero_vad/cross_test.rs
+++ b/crates/bunsen/src/kits/speech/silero_vad/cross_test.rs
@@ -26,6 +26,7 @@ mod tests {
             SileroVadMeta,
             reference::ReferenceModel,
         },
+        prelude::*,
         support::testing::PerformanceBackend,
     };
 
@@ -65,12 +66,12 @@ mod tests {
 
             s_out
                 .reshape([batch, 1])
-                .to_data()
-                .assert_approx_eq::<F>(&r_out.to_data(), Tolerance::permissive());
+                .to_data_as::<F>()
+                .assert_approx_eq::<F>(&r_out.to_data_as::<F>(), Tolerance::permissive());
 
             s_state
-                .to_data()
-                .assert_approx_eq::<F>(&r_state.to_data(), Tolerance::permissive());
+                .to_data_as::<F>()
+                .assert_approx_eq::<F>(&r_state.to_data_as::<F>(), Tolerance::permissive());
         }
     }
 
diff --git a/crates/bunsen/src/kits/speech/ten_vad/cross_test.rs b/crates/bunsen/src/kits/speech/ten_vad/cross_test.rs
index d36a504..42c4a2b 100644
--- a/crates/bunsen/src/kits/speech/ten_vad/cross_test.rs
+++ b/crates/bunsen/src/kits/speech/ten_vad/cross_test.rs
@@ -13,6 +13,7 @@ mod test {
             TenVad,
             reference::ReferenceModel,
         },
+        prelude::*,
         support::testing::PerformanceBackend,
     };
 
@@ -50,31 +51,27 @@ mod test {
 
         mod_prob
             .unsqueeze_dim::<3>(2)
-            .to_data()
-            .assert_approx_eq::<F>(&ref_prob.to_data(), Tolerance::permissive());
+            .to_data_as::<F>()
+            .assert_approx_eq::<F>(&ref_prob.to_data_as::<F>(), Tolerance::permissive());
 
         mod_lstm1_state
             .hidden
-            .clone()
-            .to_data()
-            .assert_approx_eq::<F>(&ref_lstm1_hidden.to_data(), Tolerance::permissive());
+            .to_data_as::<F>()
+            .assert_approx_eq::<F>(&ref_lstm1_hidden.to_data_as::<F>(), Tolerance::permissive());
 
         mod_lstm1_state
             .cell
-            .clone()
-            .to_data()
-            .assert_approx_eq::<F>(&ref_lstm1_cell.to_data(), Tolerance::permissive());
+            .to_data_as::<F>()
+            .assert_approx_eq::<F>(&ref_lstm1_cell.to_data_as::<F>(), Tolerance::permissive());
 
         mod_lstm2_state
             .hidden
-            .clone()
-            .to_data()
-            .assert_approx_eq::<F>(&ref_lstm2_hidden.to_data(), Tolerance::permissive());
+            .to_data_as::<F>()
+            .assert_approx_eq::<F>(&ref_lstm2_hidden.to_data_as::<F>(), Tolerance::permissive());
 
         mod_lstm2_state
             .cell
-            .clone()
-            .to_data()
-            .assert_approx_eq::<F>(&ref_lstm2_cell.to_data(), Tolerance::permissive());
+            .to_data_as::<F>()
+            .assert_approx_eq::<F>(&ref_lstm2_cell.to_data_as::<F>(), Tolerance::permissive());
     }
 }
diff --git a/crates/bunsen/src/kits/speech/whisper/blocks/decoder_block.rs b/crates/bunsen/src/kits/speech/whisper/blocks/decoder_block.rs
index aab7a1b..ca4e5f5 100644
--- a/crates/bunsen/src/kits/speech/whisper/blocks/decoder_block.rs
+++ b/crates/bunsen/src/kits/speech/whisper/blocks/decoder_block.rs
@@ -281,12 +281,12 @@ mod tests {
             .output
             .clone()
             .into_data()
-            .assert_approx_eq::<f64>(&expected.output.clone().into_data(), Default::default());
+            .assert_approx_eq::<f64>(&expected.output.to_data(), Default::default());
         result
             .ca_weights
             .clone()
             .into_data()
-            .assert_approx_eq::<f64>(&expected.ca_weights.clone().into_data(), Default::default());
+            .assert_approx_eq::<f64>(&expected.ca_weights.to_data(), Default::default());
 
         assert_shape_contract!(
             ["batch", "seq_len", "d_model"],
diff --git a/crates/bunsen/src/ops/arange.rs b/crates/bunsen/src/ops/arange.rs
index 2eb6188..dc1efb4 100644
--- a/crates/bunsen/src/ops/arange.rs
+++ b/crates/bunsen/src/ops/arange.rs
@@ -134,9 +134,12 @@ mod tests {
     };
 
     use super::*;
-    use crate::support::testing::{
-        CpuBackend,
-        assert_close_to_vec,
+    use crate::{
+        prelude::*,
+        support::testing::{
+            CpuBackend,
+            assert_close_to_vec,
+        },
     };
     type B = CpuBackend;
     type F = <B as BackendTypes>::FloatElem;
@@ -248,7 +251,7 @@ mod tests {
 
         let actual = tensor_linspace::<B>(start, end, num, &device);
 
-        actual.to_data().assert_approx_eq::<F>(
+        actual.to_data_as::<F>().assert_approx_eq::<F>(
             &TensorData::from([0.0, 0.25, 0.5, 0.75, 1.0]),
             Tolerance::default(),
         );
@@ -264,7 +267,7 @@ mod tests {
 
         let actual = tensor_linspace::<B>(start, end, num, &device);
 
-        actual.to_data().assert_approx_eq::<F>(
+        actual.to_data_as::<F>().assert_approx_eq::<F>(
             &TensorData::from([1.0, 0.7, 0.4, 0.1, -0.2]),
             Tolerance::default(),
         );
@@ -281,7 +284,7 @@ mod tests {
         let actual = tensor_linspace::<B>(start, end, num, &device);
 
         actual
-            .to_data()
+            .to_data_as::<F>()
             .assert_approx_eq::<F>(&TensorData::from([0.0]), Tolerance::default());
     }
 }
diff --git a/crates/bunsen/src/ops/signal/cosine_window.rs b/crates/bunsen/src/ops/signal/cosine_window.rs
index dc4932c..5477c25 100644
--- a/crates/bunsen/src/ops/signal/cosine_window.rs
+++ b/crates/bunsen/src/ops/signal/cosine_window.rs
@@ -235,6 +235,7 @@ mod tests {
     use super::*;
     use crate::{
         ops::signal::testing::assert_builder_impls_match,
+        prelude::*,
         support::testing::CpuBackend,
     };
 
@@ -255,8 +256,11 @@ mod tests {
 
         info!("checking hann_window reference implementation");
         hann_window::<B>(size, periodic, options.clone())
-            .to_data()
-            .assert_approx_eq::<F>(&TensorData::from(expected), Tolerance::default());
+            .to_data_as::<F>()
+            .assert_approx_eq::<F>(
+                &TensorData::from(expected).convert::<F>(),
+                Tolerance::default(),
+            );
 
         info!("cross-checking vec/tensor impls");
         assert_builder_impls_match::<B>(&cfg, expected, options.clone());
@@ -312,8 +316,11 @@ mod tests {
 
         info!("checking blackman_window reference implementation");
         blackman_window::<B>(size, periodic, options.clone())
-            .to_data()
-            .assert_approx_eq::<F>(&TensorData::from(expected), Tolerance::default());
+            .to_data_as::<F>()
+            .assert_approx_eq::<F>(
+                &TensorData::from(expected).convert::<F>(),
+                Tolerance::default(),
+            );
 
         info!("cross-checking vec/tensor impls");
         assert_builder_impls_match::<B>(&cfg, expected, options.clone());
diff --git a/crates/bunsen/src/ops/signal/sliding_stft.rs b/crates/bunsen/src/ops/signal/sliding_stft.rs
index c40e283..7dd9753 100644
--- a/crates/bunsen/src/ops/signal/sliding_stft.rs
+++ b/crates/bunsen/src/ops/signal/sliding_stft.rs
@@ -474,7 +474,10 @@ mod tests {
     };
 
     use super::*;
-    use crate::support::testing::CpuBackend;
+    use crate::{
+        prelude::*,
+        support::testing::CpuBackend,
+    };
 
     type B = CpuBackend;
     type F = <B as BackendTypes>::FloatElem;
@@ -662,10 +665,12 @@ mod tests {
                 .flat_map(|(host, row)| host.push(row))
                 .collect();
 
-            out.cast(DType::F64).to_data().assert_approx_eq::<F>(
-                &TensorData::new(expected, [batch, n_bins, 2]),
-                Tolerance::permissive(),
-            );
+            out.cast(DType::F64)
+                .to_data_as::<F>()
+                .assert_approx_eq::<F>(
+                    &TensorData::new(expected, [batch, n_bins, 2]).convert::<F>(),
+                    Tolerance::permissive(),
+                );
         }
     }
 
@@ -707,12 +712,12 @@ mod tests {
 
             let tol = Tolerance::<F>::permissive();
             seq_out
-                .to_data()
-                .assert_approx_eq::<F>(&step_out.to_data(), tol);
+                .to_data_as::<F>()
+                .assert_approx_eq::<F>(&step_out.to_data_as::<F>(), tol);
             seq_stft
                 .queue
-                .to_data()
-                .assert_approx_eq::<F>(&step_stft.queue.to_data(), tol);
+                .to_data_as::<F>()
+                .assert_approx_eq::<F>(&step_stft.queue.to_data_as::<F>(), tol);
         }
     }
 
@@ -729,8 +734,9 @@ mod tests {
         stft.forward(hop);
 
         stft.reset();
-        stft.queue
-            .to_data()
-            .assert_eq(&Tensor::<B, 2>::zeros([1, 48], &device).to_data(), true);
+        stft.queue.to_data_as::<F>().assert_eq(
+            &Tensor::<B, 2>::zeros([1, 48], &device).to_data_as::<F>(),
+            true,
+        );
     }
 }
diff --git a/examples/conway_vis/src/main.rs b/examples/conway_vis/src/main.rs
index 77cdcee..c218f67 100644
--- a/examples/conway_vis/src/main.rs
+++ b/examples/conway_vis/src/main.rs
@@ -246,7 +246,7 @@ impl Simulation {
         export_duration: Duration,
     ) -> Self {
         let shutdown = Arc::new(AtomicBool::new(false));
-        let frame_handle_1 = Arc::new(Mutex::new(conway.state.clone().into_data()));
+        let frame_handle_1 = Arc::new(Mutex::new(conway.state.to_data()));
         let frame_handle_2 = frame_handle_1.clone();
 
         let shutdown_clone = shutdown.clone();
@@ -284,7 +284,7 @@ impl Simulation {
                 if t1 - last_export > export_duration {
                     last_export = t1;
 
-                    let frame = conway.state.clone().into_data_convert::<bool>();
+                    let frame = conway.state.to_data_as::<bool>();
                     *frame_handle_1.lock().unwrap() = frame;
 
                     t1 = std::time::Instant::now();
diff --git a/examples/lbm2d_vis/src/main.rs b/examples/lbm2d_vis/src/main.rs
index 2cac0d9..b379b5d 100644
--- a/examples/lbm2d_vis/src/main.rs
+++ b/examples/lbm2d_vis/src/main.rs
@@ -190,7 +190,7 @@ fn run<B: Backend>(
             let cells = ((cells / scale) + 1.0) / 2.0;
             // let cells = cells.mul_scalar(std::f64::consts::PI / 2.0).sin();
 
-            *vis_cells_publish.lock().unwrap() = cells.to_data_convert::<f32>();
+            *vis_cells_publish.lock().unwrap() = cells.to_data_as::<f32>();
 
             last_export = std::time::Instant::now();
         }
@@ -212,7 +212,7 @@ fn run<B: Backend>(
     let mut app = FlowVisApp {
         gl: GlGraphics::new(opengl),
         cell_data: vis_cells,
-        solid_mask: solid_mask.to_data_convert::<bool>(),
+        solid_mask: solid_mask.to_data_as::<bool>(),
         opacity: args.opacity,
     };