Fix max_batct_size max_batch_size, etc (#62406)

co63oc · web-flow · commit aa7eaa5054ed · 2024-03-05T19:18:29.000+08:00
diff --git a/paddle/fluid/inference/tensorrt/plugin/anchor_generator_op_plugin.cu b/paddle/fluid/inference/tensorrt/plugin/anchor_generator_op_plugin.cu
@@ -279,7 +279,7 @@ void AnchorGeneratorPlugin::configurePlugin(
     const bool* input_is_broadcast,
     const bool* output_is_broadcast,
     nvinfer1::PluginFormat float_format,
-    int max_batct_size) TRT_NOEXCEPT {}
+    int max_batch_size) TRT_NOEXCEPT {}
 
 nvinfer1::IPluginV2Ext* AnchorGeneratorPlugin::clone() const TRT_NOEXCEPT {
   auto plugin = new AnchorGeneratorPlugin(data_type_,
diff --git a/paddle/fluid/inference/tensorrt/plugin/anchor_generator_op_plugin.h b/paddle/fluid/inference/tensorrt/plugin/anchor_generator_op_plugin.h
@@ -84,7 +84,7 @@ class AnchorGeneratorPlugin : public nvinfer1::IPluginV2Ext {
                        const bool* input_is_broadcast,
                        const bool* output_is_broadcast,
                        nvinfer1::PluginFormat float_format,
-                       int max_batct_size) TRT_NOEXCEPT override;
+                       int max_batch_size) TRT_NOEXCEPT override;
   nvinfer1::IPluginV2Ext* clone() const TRT_NOEXCEPT override;
 
  private:
@@ -148,10 +148,11 @@ class AnchorGeneratorPluginDynamic : public DynamicPluginTensorRT {
   AnchorGeneratorPluginDynamic(void const* data, size_t length);
   ~AnchorGeneratorPluginDynamic();
   nvinfer1::IPluginV2DynamicExt* clone() const TRT_NOEXCEPT override;
-  nvinfer1::DimsExprs getOutputDimensions(int outputIndex,
-                                          const nvinfer1::DimsExprs* inputs,
-                                          int nbInputs,
-                                          nvinfer1::IExprBuilder& exprBuilder)
+  nvinfer1::DimsExprs getOutputDimensions(
+      int outputIndex,
+      const nvinfer1::DimsExprs* inputs,
+      int nbInputs,
+      nvinfer1::IExprBuilder& exprBuilder)  // NOLINT
       TRT_NOEXCEPT override;
   bool supportsFormatCombination(int pos,
                                  const nvinfer1::PluginTensorDesc* inOut,
diff --git a/paddle/fluid/inference/tensorrt/plugin/deformable_conv_op_plugin.cu b/paddle/fluid/inference/tensorrt/plugin/deformable_conv_op_plugin.cu
@@ -829,7 +829,7 @@ void DeformableConvPlugin::configurePlugin(
     const bool* input_is_broadcast,
     const bool* output_is_broadcast,
     nvinfer1::PluginFormat float_format,
-    int max_batct_size) TRT_NOEXCEPT {
+    int max_batch_size) TRT_NOEXCEPT {
   PADDLE_ENFORCE_EQ(
       nb_inputs,
       3,
diff --git a/paddle/fluid/inference/tensorrt/plugin/deformable_conv_op_plugin.h b/paddle/fluid/inference/tensorrt/plugin/deformable_conv_op_plugin.h
@@ -108,7 +108,7 @@ class DeformableConvPlugin : public nvinfer1::IPluginV2Ext {
                        const bool* input_is_broadcast,
                        const bool* output_is_broadcast,
                        nvinfer1::PluginFormat float_format,
-                       int max_batct_size) TRT_NOEXCEPT override;
+                       int max_batch_size) TRT_NOEXCEPT override;
   nvinfer1::IPluginV2Ext* clone() const TRT_NOEXCEPT override;
 
  private:
diff --git a/paddle/fluid/inference/tensorrt/plugin/trt_plugin.cc b/paddle/fluid/inference/tensorrt/plugin/trt_plugin.cc
@@ -19,53 +19,53 @@ namespace inference {
 namespace tensorrt {
 namespace plugin {
 
-inline void Seria(void*& buffer,  // NOLINT
-                  const std::vector<nvinfer1::Dims>& input_dims,
-                  nvinfer1::DataType data_type,
-                  nvinfer1::PluginFormat data_format,
-                  bool with_fp16) {
+inline void Serialize(void*& buffer,  // NOLINT
+                      const std::vector<nvinfer1::Dims>& input_dims,
+                      nvinfer1::DataType data_type,
+                      nvinfer1::PluginFormat data_format,
+                      bool with_fp16) {
   SerializeValue(&buffer, input_dims);
   SerializeValue(&buffer, data_type);
   SerializeValue(&buffer, data_format);
   SerializeValue(&buffer, with_fp16);
 }
 
-inline void Deseria(void const*& serial_data,
-                    size_t& serial_length,  // NOLINT
-                    std::vector<nvinfer1::Dims>* input_dims,
-                    nvinfer1::DataType* data_type,
-                    nvinfer1::PluginFormat* data_format,
-                    bool* with_fp16) {
+inline void Deserialize(void const*& serial_data,  // NOLINT
+                        size_t& serial_length,     // NOLINT
+                        std::vector<nvinfer1::Dims>* input_dims,
+                        nvinfer1::DataType* data_type,
+                        nvinfer1::PluginFormat* data_format,
+                        bool* with_fp16) {
   DeserializeValue(&serial_data, &serial_length, input_dims);
   DeserializeValue(&serial_data, &serial_length, data_type);
   DeserializeValue(&serial_data, &serial_length, data_format);
   DeserializeValue(&serial_data, &serial_length, with_fp16);
 }
 
-inline size_t SeriaSize(const std::vector<nvinfer1::Dims>& input_dims,
-                        nvinfer1::DataType data_type,
-                        nvinfer1::PluginFormat data_format,
-                        bool with_fp16) {
+inline size_t SerializeSize(const std::vector<nvinfer1::Dims>& input_dims,
+                            nvinfer1::DataType data_type,
+                            nvinfer1::PluginFormat data_format,
+                            bool with_fp16) {
   return (SerializedSize(input_dims) + SerializedSize(data_type) +
           SerializedSize(data_format) + SerializedSize(with_fp16));
 }
 
 void PluginTensorRT::serializeBase(void*& buffer) const {
-  Seria(buffer, input_dims_, data_type_, data_format_, with_fp16_);
+  Serialize(buffer, input_dims_, data_type_, data_format_, with_fp16_);
 }
 
 void PluginTensorRT::deserializeBase(void const*& serial_data,
                                      size_t& serial_length) {
-  Deseria(serial_data,
-          serial_length,
-          &input_dims_,
-          &data_type_,
-          &data_format_,
-          &with_fp16_);
+  Deserialize(serial_data,
+              serial_length,
+              &input_dims_,
+              &data_type_,
+              &data_format_,
+              &with_fp16_);
 }
 
 size_t PluginTensorRT::getBaseSerializationSize() const {
-  return SeriaSize(input_dims_, data_type_, data_format_, with_fp16_);
+  return SerializeSize(input_dims_, data_type_, data_format_, with_fp16_);
 }
 
 bool PluginTensorRT::supportsFormat(
@@ -87,21 +87,21 @@ void PluginTensorRT::configureWithFormat(const nvinfer1::Dims* input_dims,
 }
 
 void PluginTensorRTV2Ext::serializeBase(void*& buffer) const {
-  Seria(buffer, input_dims_, data_type_, data_format_, with_fp16_);
+  Serialize(buffer, input_dims_, data_type_, data_format_, with_fp16_);
 }
 
 void PluginTensorRTV2Ext::deserializeBase(void const*& serial_data,
                                           size_t& serial_length) {
-  Deseria(serial_data,
-          serial_length,
-          &input_dims_,
-          &data_type_,
-          &data_format_,
-          &with_fp16_);
+  Deserialize(serial_data,
+              serial_length,
+              &input_dims_,
+              &data_type_,
+              &data_format_,
+              &with_fp16_);
 }
 
 size_t PluginTensorRTV2Ext::getBaseSerializationSize() const {
-  return SeriaSize(input_dims_, data_type_, data_format_, with_fp16_);
+  return SerializeSize(input_dims_, data_type_, data_format_, with_fp16_);
 }
 
 void PluginTensorRTV2Ext::configurePlugin(
diff --git a/paddle/fluid/inference/tensorrt/plugin/yolo_box_op_plugin.cu b/paddle/fluid/inference/tensorrt/plugin/yolo_box_op_plugin.cu
@@ -421,7 +421,7 @@ void YoloBoxPlugin::configurePlugin(const nvinfer1::Dims* input_dims,
                                     const bool* input_is_broadcast,
                                     const bool* output_is_broadcast,
                                     nvinfer1::PluginFormat float_format,
-                                    int max_batct_size) TRT_NOEXCEPT {}
+                                    int max_batch_size) TRT_NOEXCEPT {}
 
 nvinfer1::IPluginV2Ext* YoloBoxPlugin::clone() const TRT_NOEXCEPT {
   return new YoloBoxPlugin(data_type_,
diff --git a/paddle/fluid/inference/tensorrt/plugin/yolo_box_op_plugin.h b/paddle/fluid/inference/tensorrt/plugin/yolo_box_op_plugin.h
@@ -93,7 +93,7 @@ class YoloBoxPlugin : public nvinfer1::IPluginV2Ext {
                        const bool* input_is_broadcast,
                        const bool* output_is_broadcast,
                        nvinfer1::PluginFormat float_format,
-                       int max_batct_size) TRT_NOEXCEPT override;
+                       int max_batch_size) TRT_NOEXCEPT override;
   nvinfer1::IPluginV2Ext* clone() const TRT_NOEXCEPT override;
 
  private:
diff --git a/paddle/fluid/ir_adaptor/translator/op_translator.cc b/paddle/fluid/ir_adaptor/translator/op_translator.cc
@@ -334,7 +334,7 @@ pir::OpInfo OpTranscriber::LookUpOpInfo(pir::IrContext* ctx,
                paddle::framework::proto::VarType::SELECTED_ROWS) {
       need_inputs_sig.emplace_back("selected_rows");
     } else {
-      IR_THROW("Op %d only support densetensor and selected_rows, but not %d",
+      IR_THROW("Op %d only support dense tensor and selected_rows, but not %d",
                op_desc.Type(),
                var->GetType());
     }
diff --git a/paddle/fluid/memory/allocation/stream_safe_cuda_allocator.cc b/paddle/fluid/memory/allocation/stream_safe_cuda_allocator.cc
@@ -86,7 +86,7 @@ bool StreamSafeCUDAAllocation::CanBeFreed() {
     gpuError_t err = cudaEventQuery(event);
     if (err == cudaErrorNotReady) {
       VLOG(9) << "Event " << event << " for " << ptr() << " is not completed";
-      // Erase the completded event before "it"
+      // Erase the completed event before "it"
       outstanding_event_map_.erase(outstanding_event_map_.begin(), it);
       return false;
     }
@@ -96,7 +96,7 @@ bool StreamSafeCUDAAllocation::CanBeFreed() {
     gpuError_t err = hipEventQuery(event);
     if (err == hipErrorNotReady) {
       VLOG(9) << "Event " << event << " for " << ptr() << " is not completed";
-      // Erase the completded event before "it"
+      // Erase the completed event before "it"
       outstanding_event_map_.erase(outstanding_event_map_.begin(), it);
       return false;
     }
@@ -234,7 +234,7 @@ void StreamSafeCUDAAllocator::FreeImpl(phi::Allocation* allocation) {
 
 uint64_t StreamSafeCUDAAllocator::ReleaseImpl(const platform::Place& place) {
   if (UNLIKELY(in_cuda_graph_capturing_)) {
-    VLOG(7) << "Memory release forbidden in CUDA Graph Captruing";
+    VLOG(7) << "Memory release forbidden in CUDA Graph Capturing";
     return 0;
   }
 
@@ -249,8 +249,8 @@ uint64_t StreamSafeCUDAAllocator::ReleaseImpl(const platform::Place& place) {
 }
 
 void StreamSafeCUDAAllocator::ProcessUnfreedAllocations() {
-  // NOTE(Ruibiao): This condition is to reduce lock competion. It does not need
-  // to be thread-safe since here occasional misjudgments are permissible.
+  // NOTE(Ruibiao): This condition is to reduce lock completion. It does not
+  // need to be thread-safe since here occasional misjudgments are permissible.
   if (unfreed_allocations_.empty()) {
     return;
   }
diff --git a/paddle/fluid/memory/allocation/stream_safe_cuda_allocator.h b/paddle/fluid/memory/allocation/stream_safe_cuda_allocator.h
@@ -54,7 +54,7 @@ class StreamSafeCUDAAllocation : public Allocation {
   std::map<gpuStream_t, gpuEvent_t> outstanding_event_map_;
   gpuStream_t owning_stream_;
   SpinLock outstanding_event_map_lock_;
-  // To compatiable with CUDA Graph, hold the allocator shared_ptr so that
+  // To compatible with CUDA Graph, hold the allocator shared_ptr so that
   // Allocator will not deconstruct before Allocation
   std::shared_ptr<Allocator> allocator_;
 };
diff --git a/paddle/fluid/memory/allocation/stream_safe_custom_device_allocator.cc b/paddle/fluid/memory/allocation/stream_safe_custom_device_allocator.cc
@@ -215,8 +215,8 @@ uint64_t StreamSafeCustomDeviceAllocator::ReleaseImpl(
 }
 
 void StreamSafeCustomDeviceAllocator::ProcessUnfreedAllocations() {
-  // NOTE(Ruibiao): This condition is to reduce lock competion. It does not need
-  // to be thread-safe since here occasional misjudgments are permissible.
+  // NOTE(Ruibiao): This condition is to reduce lock completion. It does not
+  // need to be thread-safe since here occasional misjudgments are permissible.
   if (unfreed_allocations_.empty()) {
     return;
   }
diff --git a/paddle/fluid/operators/collective/c_allreduce_op.h b/paddle/fluid/operators/collective/c_allreduce_op.h
@@ -391,7 +391,7 @@ class CAllReduceOpCUDAKernel : public framework::OpKernel<T> {
       stream = ctx.cuda_device_context().stream();
     }
     VLOG(10) << "all reduce buffer:" << sendbuff << ", numel:" << numel
-             << ", redtype:" << static_cast<int>(red_type)
+             << ", reduce type:" << static_cast<int>(red_type)
              << ", dtype:" << dtype << ", comm:" << comm
              << ", stream:" << stream;
 
diff --git a/paddle/phi/core/kernel_factory.cc b/paddle/phi/core/kernel_factory.cc
@@ -30,7 +30,7 @@
 
 PHI_DEFINE_EXPORTED_bool(use_stride_kernel,
                          true,
-                         "Whether to use strdie kernel if op support stride.");
+                         "Whether to use stride kernel if op support stride.");
 
 COMMON_DECLARE_int32(low_precision_op_list);
 COMMON_DECLARE_bool(enable_api_kernel_fallback);
diff --git a/paddle/phi/core/os_info.h b/paddle/phi/core/os_info.h
@@ -54,7 +54,7 @@ ThreadId GetCurrentThreadId();
 
 // Return the map from StdTid to ThreadId
 // Returns current snapshot of all threads. Make sure there is no thread
-// create/destory when using it.
+// create/destroy when using it.
 std::unordered_map<uint64_t, ThreadId> GetAllThreadIds();
 
 static constexpr const char* kDefaultThreadName = "unnamed";
@@ -63,7 +63,7 @@ std::string GetCurrentThreadName();
 
 // Return the map from StdTid to ThreadName
 // Returns current snapshot of all threads. Make sure there is no thread
-// create/destory when using it.
+// create/destroy when using it.
 std::unordered_map<uint64_t, std::string> GetAllThreadNames();
 
 // Thread name is immutable, only the first call will succeed.
diff --git a/paddle/phi/core/selected_rows_impl.cc b/paddle/phi/core/selected_rows_impl.cc
@@ -188,7 +188,7 @@ void SelectedRowsImpl::Get(const phi::DenseTensor& ids,
         value->numel() / value->dims()[0],
         phi::errors::InvalidArgument(
             "Output tensor should have the same shape with table "
-            "except the first dimmension, excepted value width not counting "
+            "except the first dimension, excepted value width not counting "
             "the first dimension is %d, actual value width is %d.",
             value_width,
             value->numel() / value->dims()[0]));
diff --git a/paddle/phi/core/sparse_csr_tensor.h b/paddle/phi/core/sparse_csr_tensor.h
@@ -42,7 +42,7 @@ class SparseCsrTensor : public TensorBase,
   SparseCsrTensor(const SparseCsrTensor& other);
 
   /// \brief create the sparse csr tensor.
-  /// \param non_zero_crows The compresessed row index of non zero elements in
+  /// \param non_zero_crows The compressed row index of non zero elements in
   /// original dense tensor.
   /// \param non_zero_cols The column index of non zero elements in original
   /// dense tensor.
@@ -132,7 +132,7 @@ class SparseCsrTensor : public TensorBase,
 
   /// \brief Test whether the non_zero_elements_ storage is allocated.
   /// In special cases, when nnz=0, non_zero_elements_ will not need to be
-  /// initialized, but it is neccessary to return true here, otherwise the
+  /// initialized, but it is necessary to return true here, otherwise the
   /// gradient will be None. return Whether the non_zero_elements_ storage is
   /// allocated.
   bool initialized() const override {
@@ -145,7 +145,7 @@ class SparseCsrTensor : public TensorBase,
   void Resize(const DDim& dense_dims, const int64_t non_zero_num);
 
   /// \brief set the member of sparse csr tensor.
-  /// \param non_zero_crows The compresessed row index of non zero elements in
+  /// \param non_zero_crows The compressed row index of non zero elements in
   /// original dense tensor.
   /// \param non_zero_cols The column index of non zero elements in original
   /// dense tensor.
@@ -157,7 +157,7 @@ class SparseCsrTensor : public TensorBase,
                  const DDim& dims);
 
   /// \brief set the member of sparse csr tensor.
-  /// \param non_zero_crows The compresessed row index of non zero elements in
+  /// \param non_zero_crows The compressed row index of non zero elements in
   /// original dense tensor.
   /// \param non_zero_cols The column index of non zero elements in original
   /// dense tensor.
diff --git a/paddle/phi/core/storage_properties.h b/paddle/phi/core/storage_properties.h
@@ -63,7 +63,7 @@ struct XPUStorageProperties
 };
 #endif
 
-// Add OneDNNStorageProperties firstly for unittest covergae
+// Add OneDNNStorageProperties firstly for unittest coverage
 #ifdef PADDLE_WITH_DNNL
 struct OneDNNStorageProperties
     : public StorageProperties,
diff --git a/paddle/phi/core/stream.h b/paddle/phi/core/stream.h
@@ -26,7 +26,7 @@ class Stream final {
   StreamId id() const { return id_; }
 
  private:
-  StreamId id_{0};  // not onwed the stream
+  StreamId id_{0};  // not owned the stream
 };
 
 }  // namespace phi
diff --git a/paddle/phi/core/tensor_array.h b/paddle/phi/core/tensor_array.h
@@ -54,13 +54,13 @@ class TensorArray : public TensorBase,
   /// \return The name of the class.
   static const char* name() { return "TensorArray"; }
 
-  /// \brief This overrided function is not used in TensorArray.
+  /// \brief This overridden function is not used in TensorArray.
   TEST_API int64_t numel() const override;
 
-  /// \brief This overrided function is not used in TensorArray.
+  /// \brief This overridden function is not used in TensorArray.
   TEST_API const DDim& dims() const override;
 
-  /// \brief This overrided function is not used in TensorArray.
+  /// \brief This overridden function is not used in TensorArray.
   TEST_API const Place& place() const override;
 
   TEST_API DataType dtype() const override;
@@ -75,7 +75,7 @@ class TensorArray : public TensorBase,
   void set_layout(const DataLayout layout);
 #endif
 
-  /// \brief This overrided function is not used in TensorArray.
+  /// \brief This overridden function is not used in TensorArray.
   TEST_API bool valid() const override;
 
   /// \brief Test whether the tensor's storage in TensorArray is allocated.
diff --git a/paddle/phi/core/threadpool.cc b/paddle/phi/core/threadpool.cc
@@ -54,7 +54,7 @@ void ThreadPool::Init() {
 ThreadPool::ThreadPool(int num_threads) : running_(true) {
   threads_.resize(num_threads);
   for (auto& thread : threads_) {
-    // TODO(Yancey1989): binding the thread on the specify CPU numberw
+    // TODO(Yancey1989): binding the thread on the specify CPU number
     thread = std::make_unique<std::thread>([this] { ThreadPool::TaskLoop(); });
   }
 }
diff --git a/paddle/phi/core/threadpool.h b/paddle/phi/core/threadpool.h
@@ -80,7 +80,7 @@ class ThreadPool {
             new common::enforce::EnforceNotMet(ex));
       } catch (const std::exception& e) {
         PADDLE_THROW(phi::errors::Fatal(
-            "Unexpected exception is catched in thread pool. All "
+            "Unexpected exception is caught in thread pool. All "
             "throwable exception in Paddle should be an EnforceNotMet."
             "The exception is:\n %s.",
             e.what()));
@@ -129,7 +129,7 @@ class ThreadPoolIO : ThreadPool {
   static void InitIO();
 
  private:
-  // NOTE: threadpool in base will be inhereted here.
+  // NOTE: threadpool in base will be inherited here.
   static std::unique_ptr<ThreadPool> io_threadpool_;
   static std::once_flag io_init_flag_;
 };
diff --git a/paddle/phi/kernels/gpu/matrix_rank_tol_kernel.cu b/paddle/phi/kernels/gpu/matrix_rank_tol_kernel.cu
diff --git a/paddle/phi/kernels/gpu/shuffle_batch_utils.h b/paddle/phi/kernels/gpu/shuffle_batch_utils.h

Original file line number	Diff line number	Diff line change
`@@ -334,7 +334,7 @@ pir::OpInfo OpTranscriber::LookUpOpInfo(pir::IrContext* ctx,`
`334`	`334`	`paddle::framework::proto::VarType::SELECTED_ROWS) {`
`335`	`335`	`need_inputs_sig.emplace_back("selected_rows");`
`336`	`336`	`} else {`
`337`		`- IR_THROW("Op %d only support densetensor and selected_rows, but not %d",`
	`337`	`+ IR_THROW("Op %d only support dense tensor and selected_rows, but not %d",`
`338`	`338`	`op_desc.Type(),`
`339`	`339`	`var->GetType());`
`340`	`340`	`}`