[skip ci] Auto-format code with clang-format and black

Author: github-actions[bot]

csrc/include/pmpp/types/torch_types.hpp

@@ -15,8 +15,8 @@ static_assert(sizeof(ScalarType) == 1);
 /**
  * @brief Combine DeviceT and ScalarT to a uint16_t.
  */
-constexpr auto combineDeviceTandScalarT(DeviceType deviceT,
-                                        ScalarType scalarT) -> uint16_t
+constexpr auto combineDeviceTandScalarT(DeviceType deviceT, ScalarType scalarT)
+    -> uint16_t
 {
     return (uint16_t(deviceT) << 8) | uint16_t(scalarT);
 }

csrc/lib/ops/torch_impl.hpp

@@ -4,14 +4,14 @@
 
 namespace pmpp::ops::cpu::torch_impl
 {
-auto vectorAddImpl(const torch::Tensor& A,
-                   const torch::Tensor& B) -> torch::Tensor;
+auto vectorAddImpl(const torch::Tensor& A, const torch::Tensor& B)
+    -> torch::Tensor;
 auto cvtRGBtoGrayImpl(const torch::Tensor& img) -> torch::Tensor;
 }  // namespace pmpp::ops::cpu::torch_impl
 
 namespace pmpp::ops::cuda::torch_impl
 {
-auto vectorAddImpl(const torch::Tensor& A,
-                   const torch::Tensor& B) -> torch::Tensor;
+auto vectorAddImpl(const torch::Tensor& A, const torch::Tensor& B)
+    -> torch::Tensor;
 auto cvtRGBtoGrayImpl(const torch::Tensor& img) -> torch::Tensor;
 }  // namespace pmpp::ops::cuda::torch_impl

csrc/lib/ops/vecAdd/torch_impl.cpp

@@ -27,8 +27,8 @@ extern void launchVecAdd(const fp32_t* a, const fp32_t* b, fp32_t* c,
 
 namespace torch_impl
 {
-auto vectorAddImpl(const torch::Tensor& A,
-                   const torch::Tensor& B) -> torch::Tensor
+auto vectorAddImpl(const torch::Tensor& A, const torch::Tensor& B)
+    -> torch::Tensor
 {
     VECTOR_ADD_CHECK(A, B, "CPU");
 
@@ -57,8 +57,8 @@ extern void launchVecAdd(const fp32_t* d_A, const fp32_t* d_B, fp32_t* d_C,
 
 namespace torch_impl
 {
-auto vectorAddImpl(const torch::Tensor& A,
-                   const torch::Tensor& B) -> torch::Tensor
+auto vectorAddImpl(const torch::Tensor& A, const torch::Tensor& B)
+    -> torch::Tensor
 {
     VECTOR_ADD_CHECK(A, B, "CUDA");
 

src/pmpp/models/attention.py

@@ -215,9 +215,7 @@ def forward(self, prompt: torch.Tensor, is_prefilling: bool = True):
             # out: (seq_len, embed_dim)
             # k: (kv_len, embed_dim)
             # v: (kv_len, embed_dim)
-            out, k, v = self.cached_mha(
-                embedded_prompt, self.k_cache, self.v_cache
-            )
+            out, k, v = self.cached_mha(embedded_prompt, self.k_cache, self.v_cache)
 
         # Update k cache and v cache
         # [NOTE]

test/test.py

@@ -18,5 +18,3 @@
 pic_in = pic_in.cuda()
 pic_out_cuda = torch.ops.pmpp.cvt_rgb_to_gray(pic_in)
 print(pic_out_cuda.cpu())
-
-