Fix pre-commit

Signed-off-by: Jiayu Chang <jiayuc@nvidia.com>

Author: JyChang012

cpp/tensorrt_llm/kernels/cuda_graph_grouped_gemm.cu

@@ -368,4 +368,5 @@ void cudaGraphSplitKGroupedGemm(cutlass::gemm::GemmCoord* problemSizesPtr, int p
 }
 
 } // namespace kernels
+
 TRTLLM_NAMESPACE_END

cpp/tensorrt_llm/kernels/cuda_graph_grouped_gemm.h

@@ -59,4 +59,5 @@ void cudaGraphSplitKGroupedGemm(cutlass::gemm::GemmCoord* problemSizesPtr, int p
     cutlass::gemm::GemmCoord* hostMaxProblemSizesPtr, int64_t* splitKOffsetsGpu, cudaStream_t stream);
 
 } // namespace kernels
+
 TRTLLM_NAMESPACE_END

cpp/tensorrt_llm/kernels/lora/loraGroupGEMMParamFillRowReorderFusion.cu

@@ -307,8 +307,8 @@ void launchKernelWithModuleCount(int32_t* in_sizes, int32_t* out_sizes, int64_t*
     int const itemsPerRow = ITEMS_PER_BLOCK * gridDimX;
     gridDimY = std::max(gridDimY,
         common::ceilDiv(static_cast<int>(module_count * batch_size * max_lora_rank * dtype_element_size), itemsPerRow));
-    gridDimY = std::max(
-        gridDimY, common::ceilDiv(static_cast<int>(batch_size * sum_output_hidden_size * dtype_element_size), itemsPerRow));
+    gridDimY = std::max(gridDimY,
+        common::ceilDiv(static_cast<int>(batch_size * sum_output_hidden_size * dtype_element_size), itemsPerRow));
 
     dim3 grid(gridDimX, gridDimY);
     dim3 block(BlockDim);

cpp/tensorrt_llm/thop/loraOp.cpp

@@ -419,5 +419,6 @@ TORCH_LIBRARY_IMPL(trtllm, CUDA, m)
 {
     m.impl("lora_grouped_gemm", &tensorrt_llm::torch_ext::lora_grouped_gemm);
     m.impl("lora_grouped_gemm_cuda_graph", &tensorrt_llm::torch_ext::lora_grouped_gemm_cuda_graph);
-    m.impl("lora_group_gemm_param_fill_row_reorder_fusion", &tensorrt_llm::torch_ext::lora_group_gemm_param_fill_row_reorder_fusion);
+    m.impl("lora_group_gemm_param_fill_row_reorder_fusion",
+        &tensorrt_llm::torch_ext::lora_group_gemm_param_fill_row_reorder_fusion);
 }

tensorrt_llm/_torch/pyexecutor/model_engine.py

@@ -3288,15 +3288,11 @@ def _prepare_inputs(
                 raise NotImplementedError(
                     f"Unsupported cp_type {getattr(cp_type, 'name', cp_type)}.")
 
-        return self._prepare_tp_inputs(scheduled_requests,
-                                       kv_cache_manager,
-                                       attn_metadata,
-                                       spec_metadata,
-                                       new_tensors_device,
-                                       cache_indirection_buffer,
-                                       num_accepted_tokens_device,
-                                       req_id_to_old_request, resource_manager,
-                                       maybe_graph)
+        return self._prepare_tp_inputs(
+            scheduled_requests, kv_cache_manager, attn_metadata, spec_metadata,
+            new_tensors_device, cache_indirection_buffer,
+            num_accepted_tokens_device, req_id_to_old_request, resource_manager,
+            maybe_graph)
 
     @torch.inference_mode()
     @with_model_extra_attrs(lambda self: self.model.extra_attrs)