feat: update workspace paths and enhance gitignore

- Updated stablediffusion crate path from "../stable-diffusion-burn" to "./crates/stable-diffusion-burn" for proper workspace resolution
- Enhanced .gitignore to include generated model files (.mpk, .pt, .bin, .safetensors, .ckpt) and user_data directory
- Added Cargo.lock to gitignore with appropriate comment
- Reorganized IDE files section in gitignore for better clarity
- Added newline at end of file for proper formatting

crates/stable-diffusion-burn/burn-crates/burn-backend-tests/tests/common/autodiff.rs

@@ -0,0 +1,35 @@
+// Burn autodiff tests, reusable with element types.
+
+pub use super::*;
+
+#[path = "../autodiff/mod.rs"]
+mod base;
+
+mod checkpointing {
+    pub use super::*;
+    use burn_autodiff::checkpoint::strategy::BalancedCheckpointing;
+
+    // Override type def
+    pub type TestAutodiffBackend = Autodiff<TestBackend, BalancedCheckpointing>;
+    pub type TestAutodiffTensor<const D: usize> = Tensor<TestAutodiffBackend, D>;
+
+    include!("../autodiff/mod.rs");
+}
+
+use burn_backend_tests::test_float_elem_variant;
+
+// NOTE: this currently doesn't test checkpointing with different dtypes
+test_float_elem_variant!(
+    f16,
+    burn_tensor::f16,
+    "../autodiff/mod.rs",
+    ["vulkan", "cuda", "rocm", "metal"]
+);
+
+// TODO: bf16 not yet supported on any backend for full test suite
+// test_float_elem_variant!(
+//     bf16,
+//     burn_tensor::bf16,
+//     "../autodiff/mod.rs",
+//     [] // ["cuda", "rocm"] TODO, ["vulkan"] only supports bf16 for matmul, metal/wgpu doesn't support bf16
+// );

crates/stable-diffusion-burn/burn-crates/burn-backend-tests/tests/common/backend.rs

@@ -0,0 +1,47 @@
+// Re-export
+use super::FloatElemType;
+
+// Default
+#[cfg(feature = "ndarray")]
+pub type TestBackend = burn_ndarray::NdArray<FloatElemType>;
+
+#[cfg(feature = "tch")]
+pub type TestBackend = burn_tch::LibTorch<FloatElemType>;
+
+#[cfg(feature = "cuda")]
+pub type TestBackend = burn_cuda::Cuda<FloatElemType, super::IntElemType>;
+
+#[cfg(feature = "rocm")]
+pub type TestBackend = burn_rocm::Rocm<FloatElemType, super::IntElemType>;
+
+#[cfg(feature = "wgpu")]
+pub type TestBackend = burn_wgpu::Wgpu<FloatElemType, super::IntElemType>;
+
+#[cfg(feature = "cpu")]
+pub type TestBackend = burn_cpu::Cpu<FloatElemType, super::IntElemType>;
+
+#[cfg(feature = "router")]
+pub type TestBackend = burn_router::BackendRouter<
+    burn_router::DirectByteChannel<(burn_ndarray::NdArray, burn_wgpu::Wgpu)>,
+>;
+
+/// Collection of types used across tests
+#[allow(unused)]
+pub mod prelude {
+    pub use burn_autodiff::Autodiff;
+    pub use burn_tensor::Tensor;
+
+    use super::*;
+    pub type TestTensor<const D: usize> = Tensor<TestBackend, D>;
+    pub type TestTensorInt<const D: usize> = Tensor<TestBackend, D, burn_tensor::Int>;
+    pub type TestTensorBool<const D: usize> = Tensor<TestBackend, D, burn_tensor::Bool>;
+
+    pub type FloatElem = burn_tensor::ops::FloatElem<TestBackend>;
+    pub type IntElem = burn_tensor::ops::IntElem<TestBackend>;
+
+    pub type TestAutodiffBackend = Autodiff<TestBackend>;
+    pub type TestAutodiffTensor<const D: usize> = Tensor<TestAutodiffBackend, D>;
+}
+
+#[allow(unused)]
+pub use prelude::*;

crates/stable-diffusion-burn/burn-crates/burn-backend-tests/tests/common/tensor.rs

@@ -0,0 +1,39 @@
+// Burn backend tensor tests, reusable with element types.
+
+pub use super::*;
+
+#[path = "../tensor/clone_invariance.rs"]
+mod clone_invariance;
+
+#[cfg(feature = "std")]
+#[path = "../tensor/multi_threads.rs"]
+mod multi_threads;
+
+// Default float dtype
+#[path = "../tensor/float/mod.rs"]
+mod float;
+
+// Default integer dtype
+#[path = "../tensor/int/mod.rs"]
+mod int;
+
+// Default bool dtype
+#[path = "../tensor/bool/mod.rs"]
+mod bool;
+
+use burn_backend_tests::test_float_elem_variant;
+
+test_float_elem_variant!(
+    f16,
+    burn_tensor::f16,
+    "../tensor/float/mod.rs",
+    ["vulkan", "cuda", "rocm", "metal"]
+);
+
+// TODO: bf16 not yet supported on any backend for full test suite
+// test_float_elem_variant!(
+//     bf16,
+//     burn_tensor::bf16,
+//     "../tensor/float/mod.rs",
+//     [] // ["cuda", "rocm"] TODO, ["vulkan"] only supports bf16 for matmul, metal/wgpu doesn't support bf16
+// );