diff options
Diffstat (limited to 'src/gpu/cl/operators/ClFullyConnected.cpp')
| -rw-r--r-- | src/gpu/cl/operators/ClFullyConnected.cpp | 282 |
1 files changed, 167 insertions, 115 deletions
diff --git a/src/gpu/cl/operators/ClFullyConnected.cpp b/src/gpu/cl/operators/ClFullyConnected.cpp index 5845bbc69e..6969ac8ab3 100644 --- a/src/gpu/cl/operators/ClFullyConnected.cpp +++ b/src/gpu/cl/operators/ClFullyConnected.cpp @@ -24,12 +24,13 @@ #include "src/gpu/cl/operators/ClFullyConnected.h" #include "arm_compute/core/Size2D.h" -#include "arm_compute/core/Validate.h" #include "arm_compute/core/utils/misc/ShapeCalculator.h" #include "arm_compute/core/utils/quantization/AsymmHelpers.h" +#include "arm_compute/core/Validate.h" #include "arm_compute/runtime/CL/CLScheduler.h" -#include "src/core/CL/kernels/CLFillBorderKernel.h" +#include "src/common/utils/Log.h" +#include "src/core/CL/kernels/CLFillBorderKernel.h" #include "src/core/helpers/MemoryHelpers.h" #include "src/gpu/cl/operators/ClConvertFullyConnectedWeights.h" #include "src/gpu/cl/operators/ClFlatten.h" @@ -38,11 +39,8 @@ #include "src/gpu/cl/operators/ClMatMul.h" #include "src/gpu/cl/operators/ClTranspose.h" #include "src/gpu/cl/utils/ClAuxTensorHandler.h" - #include "src/runtime/heuristics/matmul_native/ClMatMulNativeKernelConfig.h" #include "src/runtime/heuristics/matmul_native/IClMatMulNativeKernelConfig.h" - -#include "src/common/utils/Log.h" #include "support/Cast.h" #include <algorithm> @@ -62,8 +60,11 @@ inline TensorShape get_reshaped_matmul_tensor(const TensorShape &src) return TensorShape(src.x(), 1, src.y(), src.collapsed_from(2).z()); // Return value optimisation } -Status construct_gemmlowp_output_stage(const ITensorInfo &src, const ITensorInfo &weights, const ITensorInfo &dst, - GEMMLowpOutputStageInfo &gemmlowp_output_stage, ActivationLayerInfo activation_info) +Status construct_gemmlowp_output_stage(const ITensorInfo &src, + const ITensorInfo &weights, + const ITensorInfo &dst, + GEMMLowpOutputStageInfo &gemmlowp_output_stage, + ActivationLayerInfo activation_info) { gemmlowp_output_stage.type = GEMMLowpOutputStageType::QUANTIZE_DOWN_FIXEDPOINT; gemmlowp_output_stage.gemmlowp_offset = 0; @@ -73,7 +74,7 @@ Status construct_gemmlowp_output_stage(const ITensorInfo &src, const ITensorInfo const auto data_type = src.data_type(); // Configure output stage for quantized case - if(is_data_type_quantized_asymmetric(data_type)) + if (is_data_type_quantized_asymmetric(data_type)) { const QuantizationInfo oq_info = dst.quantization_info(); const UniformQuantizationInfo iq_unif = src.quantization_info().uniform(); @@ -85,15 +86,17 @@ Status construct_gemmlowp_output_stage(const ITensorInfo &src, const ITensorInfo const float multiplier = (iq_unif.scale * wq_unif.scale) / output_quant_info.scale; int output_multiplier = 0; int output_shift = 0; - ARM_COMPUTE_RETURN_ON_ERROR(quantization::calculate_quantized_multiplier(multiplier, &output_multiplier, &output_shift)); + ARM_COMPUTE_RETURN_ON_ERROR( + quantization::calculate_quantized_multiplier(multiplier, &output_multiplier, &output_shift)); PixelValue type_min{}; PixelValue type_max{}; std::tie(type_min, type_max) = get_min_max(data_type); - if(activation_info.enabled()) + if (activation_info.enabled()) { - std::tie(type_min, type_max) = get_quantized_activation_min_max(activation_info, data_type, output_quant_info); + std::tie(type_min, type_max) = + get_quantized_activation_min_max(activation_info, data_type, output_quant_info); } // Set the GEMMLowp output stage info @@ -109,31 +112,41 @@ Status construct_gemmlowp_output_stage(const ITensorInfo &src, const ITensorInfo return Status{}; } -Status validate_mm(const ITensorInfo &src, const ITensorInfo &weights, const ITensorInfo *bias, const ITensorInfo &dst, const FullyConnectedLayerInfo &fc_info, bool use_matmul) +Status validate_mm(const ITensorInfo &src, + const ITensorInfo &weights, + const ITensorInfo *bias, + const ITensorInfo &dst, + const FullyConnectedLayerInfo &fc_info, + bool use_matmul) { // Note : If input is dynamic and data is not batched, use matmul, else use gemm const bool transpose_weights = fc_info.transpose_weights ? !fc_info.are_weights_reshaped : false; - const bool use_dynamic_gemm = !use_matmul && !weights.are_values_constant() && transpose_weights; // use dynamic gemm as fallback for matmul - const bool is_quantized = is_data_type_quantized_asymmetric(src.data_type()); + const bool use_dynamic_gemm = + !use_matmul && !weights.are_values_constant() && transpose_weights; // use dynamic gemm as fallback for matmul + const bool is_quantized = is_data_type_quantized_asymmetric(src.data_type()); - if(use_matmul) + if (use_matmul) { const MatMulInfo m_info = MatMulInfo().adj_rhs(transpose_weights); // Note: LHS is reshaped here to match ClMatMul expectations of batch index - From [M, B0, B1] to [M, 1, B0, B1] TensorInfo lhs_to_use = src.clone()->set_tensor_shape(get_reshaped_matmul_tensor(src.tensor_shape())); - const GPUTarget gpu_target = CLScheduler::get().target(); - std::unique_ptr<cl_matmul::IClMatMulNativeKernelConfig> t = cl_matmul::ClMatMulNativeKernelConfigurationFactory::create(gpu_target); - const MatMulKernelInfo kernel_info = t->configure(&lhs_to_use, &weights, m_info); + const GPUTarget gpu_target = CLScheduler::get().target(); + std::unique_ptr<cl_matmul::IClMatMulNativeKernelConfig> t = + cl_matmul::ClMatMulNativeKernelConfigurationFactory::create(gpu_target); + const MatMulKernelInfo kernel_info = t->configure(&lhs_to_use, &weights, m_info); - return is_quantized ? kernels::ClMatMulLowpNativeKernel::validate(&lhs_to_use, &weights, bias, &dst, kernel_info, fc_info.activation_info) : - kernels::ClMatMulNativeKernel::validate(&lhs_to_use, &weights, bias, &dst, kernel_info, fc_info.activation_info); + return is_quantized ? kernels::ClMatMulLowpNativeKernel::validate(&lhs_to_use, &weights, bias, &dst, + kernel_info, fc_info.activation_info) + : kernels::ClMatMulNativeKernel::validate(&lhs_to_use, &weights, bias, &dst, kernel_info, + fc_info.activation_info); } else { GEMMLowpOutputStageInfo gemmlowp_output_stage; - ARM_COMPUTE_RETURN_ON_ERROR(construct_gemmlowp_output_stage(src, weights, dst, gemmlowp_output_stage, fc_info.activation_info)); + ARM_COMPUTE_RETURN_ON_ERROR( + construct_gemmlowp_output_stage(src, weights, dst, gemmlowp_output_stage, fc_info.activation_info)); const GEMMInfo &gemm_info = GEMMInfo(false, // is_a_reshaped false, // is_b_reshaped @@ -147,7 +160,7 @@ Status validate_mm(const ITensorInfo &src, const ITensorInfo &weights, const ITe true, // broadcast_bias ActivationLayerInfo()); // activation_info - if(is_quantized) + if (is_quantized) { const UniformQuantizationInfo iq_info = src.quantization_info().uniform(); const UniformQuantizationInfo wq_info = weights.quantization_info().uniform(); @@ -158,11 +171,9 @@ Status validate_mm(const ITensorInfo &src, const ITensorInfo &weights, const ITe const QuantizationInfo weights_quantization_info(wq_info.scale, -wq_info.offset); // Validate gemmlowp function - ARM_COMPUTE_RETURN_ON_ERROR(ClGemmLowpMatrixMultiplyCore::validate(&src.clone()->set_quantization_info(src_quantization_info), - &weights.clone()->set_quantization_info(weights_quantization_info), - bias, - &dst, - gemm_info)); + ARM_COMPUTE_RETURN_ON_ERROR(ClGemmLowpMatrixMultiplyCore::validate( + &src.clone()->set_quantization_info(src_quantization_info), + &weights.clone()->set_quantization_info(weights_quantization_info), bias, &dst, gemm_info)); } else { @@ -188,11 +199,15 @@ ClFullyConnected::ClFullyConnected() ClFullyConnected::~ClFullyConnected() = default; -void ClFullyConnected::configure_mm(const CLCompileContext &compile_context, ITensorInfo *src, ITensorInfo *weights, ITensorInfo *bias, ITensorInfo *dst, +void ClFullyConnected::configure_mm(const CLCompileContext &compile_context, + ITensorInfo *src, + ITensorInfo *weights, + ITensorInfo *bias, + ITensorInfo *dst, const FullyConnectedLayerInfo &fc_info) { // If weights are dynamic and matmul is supported use matmul, else use gemm - if(_use_matmul) + if (_use_matmul) { // Specify whether transpose weights is necessary in matmul info const MatMulInfo mat_info = MatMulInfo().adj_rhs(_transpose_weights); @@ -202,22 +217,25 @@ void ClFullyConnected::configure_mm(const CLCompileContext &compile_context, ITe _lhs_to_use = src->clone()->set_tensor_shape(get_reshaped_matmul_tensor(_lhs_to_use.tensor_shape())); // 2. Use heuristics to get kernel info object - const GPUTarget gpu_target = CLScheduler::get().target(); - std::unique_ptr<cl_matmul::IClMatMulNativeKernelConfig> kernel_config = cl_matmul::ClMatMulNativeKernelConfigurationFactory::create(gpu_target); - MatMulKernelInfo kernel_info = kernel_config->configure(src, weights, mat_info); + const GPUTarget gpu_target = CLScheduler::get().target(); + std::unique_ptr<cl_matmul::IClMatMulNativeKernelConfig> kernel_config = + cl_matmul::ClMatMulNativeKernelConfigurationFactory::create(gpu_target); + MatMulKernelInfo kernel_info = kernel_config->configure(src, weights, mat_info); // 3. Configure relevant matmul kernel - if(_is_quantized) + if (_is_quantized) { _matmul_lowp_native_kernel = std::make_unique<kernels::ClMatMulLowpNativeKernel>(); _matmul_lowp_native_kernel->set_target(gpu_target); - _matmul_lowp_native_kernel->configure(compile_context, src, weights, bias, dst, kernel_info, fc_info.activation_info); + _matmul_lowp_native_kernel->configure(compile_context, src, weights, bias, dst, kernel_info, + fc_info.activation_info); } else { _matmul_native_kernel = std::make_unique<kernels::ClMatMulNativeKernel>(); _matmul_native_kernel->set_target(gpu_target); - _matmul_native_kernel->configure(compile_context, src, weights, bias, dst, kernel_info, fc_info.activation_info); + _matmul_native_kernel->configure(compile_context, src, weights, bias, dst, kernel_info, + fc_info.activation_info); } } else @@ -238,7 +256,7 @@ void ClFullyConnected::configure_mm(const CLCompileContext &compile_context, ITe true, // broadcast_bias fc_info.activation_info); // activation_info - if(_is_quantized) + if (_is_quantized) { // Since we need negative offsets for computing convolution, we need to change QuantizationInfo() // Extract and negate input and weights offset @@ -248,8 +266,10 @@ void ClFullyConnected::configure_mm(const CLCompileContext &compile_context, ITe TensorInfo src_info = src->clone()->set_quantization_info(src_quantization_info); TensorInfo weights_info = weights->clone()->set_quantization_info(weights_quantization_info); - src_info.set_quantization_info(QuantizationInfo(src_quantization_info.uniform().scale, -src_quantization_info.uniform().offset)); - weights_info.set_quantization_info(QuantizationInfo(weights_quantization_info.uniform().scale, -weights_quantization_info.uniform().offset)); + src_info.set_quantization_info( + QuantizationInfo(src_quantization_info.uniform().scale, -src_quantization_info.uniform().offset)); + weights_info.set_quantization_info(QuantizationInfo(weights_quantization_info.uniform().scale, + -weights_quantization_info.uniform().offset)); // Configure gemmlowp function _mm_gemmlowp = std::make_unique<ClGemmLowpMatrixMultiplyCore>(); @@ -264,16 +284,25 @@ void ClFullyConnected::configure_mm(const CLCompileContext &compile_context, ITe } } -void ClFullyConnected::configure_conv_fc(const CLCompileContext &compile_context, ITensorInfo *src, ITensorInfo *weights, ITensorInfo *bias, ITensorInfo *dst, +void ClFullyConnected::configure_conv_fc(const CLCompileContext &compile_context, + ITensorInfo *src, + ITensorInfo *weights, + ITensorInfo *bias, + ITensorInfo *dst, const FullyConnectedLayerInfo &fc_info) { // MatMul fuses transpose operation, so we use the first dimension for comparison where appropriate. - ARM_COMPUTE_ERROR_ON((weights->dimension((_use_matmul && _transpose_weights) ? 0 : 1) != (src->dimension(0) * src->dimension(1) * src->dimension(2)))); + ARM_COMPUTE_ERROR_ON((weights->dimension((_use_matmul && _transpose_weights) ? 0 : 1) != + (src->dimension(0) * src->dimension(1) * src->dimension(2)))); // If the fully connected layer is called after a convolution layer, the input tensor must be linearized // Initialize output tensor for flatten - _flattened_src = src->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_flatten_shape(src)).set_data_layout(DataLayout::NCHW); + _flattened_src = src->clone() + ->set_is_resizable(true) + .reset_padding() + .set_tensor_shape(compute_flatten_shape(src)) + .set_data_layout(DataLayout::NCHW); // Configure flatten kernel _flatten = std::make_unique<ClFlatten>(); @@ -284,7 +313,11 @@ void ClFullyConnected::configure_conv_fc(const CLCompileContext &compile_context configure_mm(compile_context, &_flattened_src, weights, bias, dst, fc_info); } -void ClFullyConnected::configure_fc_fc(const CLCompileContext &compile_context, ITensorInfo *src, ITensorInfo *weights, ITensorInfo *bias, ITensorInfo *dst, +void ClFullyConnected::configure_fc_fc(const CLCompileContext &compile_context, + ITensorInfo *src, + ITensorInfo *weights, + ITensorInfo *bias, + ITensorInfo *dst, const FullyConnectedLayerInfo &fc_info) { // MatMul fuses transpose operation, so we use the first dimension for comparison where appropriate. @@ -294,7 +327,11 @@ void ClFullyConnected::configure_fc_fc(const CLCompileContext &compile_context, configure_mm(compile_context, src, weights, bias, dst, fc_info); } -void ClFullyConnected::configure(const CLCompileContext &compile_context, ITensorInfo *src, ITensorInfo *weights, ITensorInfo *biases, ITensorInfo *dst, +void ClFullyConnected::configure(const CLCompileContext &compile_context, + ITensorInfo *src, + ITensorInfo *weights, + ITensorInfo *biases, + ITensorInfo *dst, FullyConnectedLayerInfo fc_info) { ARM_COMPUTE_ERROR_ON_NULLPTR(src, weights, dst); @@ -317,8 +354,9 @@ void ClFullyConnected::configure(const CLCompileContext &compile_context, ITenso // 2. MatMul does not support broadcasting batch dimension, and therefore is disabled if fc is batched. // 3. When FC is after convolution and src tensor data layout does not match weights trained data layout (weights conversion kernel is required) const bool is_batched_fc_layer = dst->dimension(1) > 1; - _use_matmul = gpu_target != GPUTarget::MIDGARD && !weights->are_values_constant() && !is_batched_fc_layer && !(src->num_dimensions() > 1 && (src->data_layout() != fc_info.weights_trained_layout)); - _dynamic_gemm = !weights->are_values_constant() && _transpose_weights && !_use_matmul; + _use_matmul = gpu_target != GPUTarget::MIDGARD && !weights->are_values_constant() && !is_batched_fc_layer && + !(src->num_dimensions() > 1 && (src->data_layout() != fc_info.weights_trained_layout)); + _dynamic_gemm = !weights->are_values_constant() && _transpose_weights && !_use_matmul; // With the Fully Connected layer we can have 4 different cases: // 1) Convolution layer -> Fully Connected layer without batches @@ -327,11 +365,11 @@ void ClFullyConnected::configure(const CLCompileContext &compile_context, ITenso // 4) Fully Connected layer -> Fully Connected layer with batches // Check if we have a fully connected layer with batches - if(is_batched_fc_layer) + if (is_batched_fc_layer) { - _is_fc_after_conv = (TensorShape::num_max_dimensions >= 4) && (std::equal(src->tensor_shape().cbegin() + 3, - src->tensor_shape().cend(), - dst->tensor_shape().cbegin() + 1)); + _is_fc_after_conv = (TensorShape::num_max_dimensions >= 4) && + (std::equal(src->tensor_shape().cbegin() + 3, src->tensor_shape().cend(), + dst->tensor_shape().cbegin() + 1)); } else { @@ -341,7 +379,7 @@ void ClFullyConnected::configure(const CLCompileContext &compile_context, ITenso ITensorInfo *weights_used = weights; // Reshape weights if needed - Not needed when matmul is in use as matmul fuses transpose op. - if(_transpose_weights && !_use_matmul) + if (_transpose_weights && !_use_matmul) { // Reshape the weights _reshape_weights = std::make_unique<ClTranspose>(); @@ -351,14 +389,11 @@ void ClFullyConnected::configure(const CLCompileContext &compile_context, ITenso } // Convert weights if needed - if(_is_fc_after_conv && (src->data_layout() != fc_info.weights_trained_layout)) + if (_is_fc_after_conv && (src->data_layout() != fc_info.weights_trained_layout)) { // Convert weights _convert_weights = std::make_unique<ClConvertFullyConnectedWeights>(); - _convert_weights->configure(compile_context, - weights_used, - &_converted_weights, - src->tensor_shape(), + _convert_weights->configure(compile_context, weights_used, &_converted_weights, src->tensor_shape(), fc_info.weights_trained_layout); weights_used = &_converted_weights; @@ -366,7 +401,7 @@ void ClFullyConnected::configure(const CLCompileContext &compile_context, ITenso _run_convert_weights = true; } - if(_is_fc_after_conv) + if (_is_fc_after_conv) { // Fully Connected layer after a Convolution Layer without batches configure_conv_fc(compile_context, src, weights_used, biases, dst, fc_info); @@ -379,60 +414,69 @@ void ClFullyConnected::configure(const CLCompileContext &compile_context, ITenso // Update TensorInfo of final weights used (Need to be done in the end due to padding expansion) _weights_to_use = *weights_used; - if(_use_matmul) + if (_use_matmul) { // Note : MatMul does not use transpose and does not need auxillary memory, so only converted weights are added to aux_mem - _aux_mem[ConvertedWeights] = MemoryInfo(offset_int_vec(ConvertedWeights), MemoryLifetime::Temporary, _converted_weights.total_size()); + _aux_mem[ConvertedWeights] = + MemoryInfo(offset_int_vec(ConvertedWeights), MemoryLifetime::Temporary, _converted_weights.total_size()); } else { // Set auxiliary memory requirements for gemm operators auto gemm_mem_req = (_is_quantized) ? _mm_gemmlowp->workspace() : _mm_gemm->workspace(); - for(unsigned int i = 0; i < gemm_mem_req.size(); ++i) + for (unsigned int i = 0; i < gemm_mem_req.size(); ++i) { _aux_mem[i] = gemm_mem_req[i]; } - if(_aux_mem[1].size > 0 || _aux_mem[2].size > 0) // Persistent weights memory on GEMMs + if (_aux_mem[1].size > 0 || _aux_mem[2].size > 0) // Persistent weights memory on GEMMs { // Release permuted weights at the of prepare as they are further transposed by the assembly dispatch // Keep all the auxiliary tensors in case of dynamic weights as they are recalculated every time _aux_mem[TransposedWeights] = MemoryInfo( - offset_int_vec(TransposedWeights), - _dynamic_gemm ? MemoryLifetime::Temporary : MemoryLifetime::Prepare, - _reshaped_weights.total_size()); - _aux_mem[ConvertedWeights] = MemoryInfo( - offset_int_vec(ConvertedWeights), - _dynamic_gemm ? MemoryLifetime::Temporary : MemoryLifetime::Prepare, - _converted_weights.total_size()); + offset_int_vec(TransposedWeights), _dynamic_gemm ? MemoryLifetime::Temporary : MemoryLifetime::Prepare, + _reshaped_weights.total_size()); + _aux_mem[ConvertedWeights] = MemoryInfo(offset_int_vec(ConvertedWeights), + _dynamic_gemm ? MemoryLifetime::Temporary : MemoryLifetime::Prepare, + _converted_weights.total_size()); } else { // Release permuted weights at the of prepare as they are further transposed by the assembly dispatch - const auto transposed_wei_lft = (_weights_to_use_idx == offset_int_vec(TransposedWeights)) ? MemoryLifetime::Persistent : MemoryLifetime::Prepare; - const auto converted_wei_lft = (_weights_to_use_idx == offset_int_vec(ConvertedWeights)) ? MemoryLifetime::Persistent : MemoryLifetime::Prepare; - - _aux_mem[TransposedWeights] = MemoryInfo( - offset_int_vec(TransposedWeights), - _dynamic_gemm ? MemoryLifetime::Temporary : transposed_wei_lft, - _reshaped_weights.total_size()); - _aux_mem[ConvertedWeights] = MemoryInfo( - offset_int_vec(ConvertedWeights), - _dynamic_gemm ? MemoryLifetime::Temporary : converted_wei_lft, - _converted_weights.total_size()); + const auto transposed_wei_lft = (_weights_to_use_idx == offset_int_vec(TransposedWeights)) + ? MemoryLifetime::Persistent + : MemoryLifetime::Prepare; + const auto converted_wei_lft = (_weights_to_use_idx == offset_int_vec(ConvertedWeights)) + ? MemoryLifetime::Persistent + : MemoryLifetime::Prepare; + + _aux_mem[TransposedWeights] = MemoryInfo(offset_int_vec(TransposedWeights), + _dynamic_gemm ? MemoryLifetime::Temporary : transposed_wei_lft, + _reshaped_weights.total_size()); + _aux_mem[ConvertedWeights] = MemoryInfo(offset_int_vec(ConvertedWeights), + _dynamic_gemm ? MemoryLifetime::Temporary : converted_wei_lft, + _converted_weights.total_size()); } } - _aux_mem[FlattenedSrc] = MemoryInfo(offset_int_vec(FlattenedSrc), MemoryLifetime::Temporary, _flattened_src.total_size()); + _aux_mem[FlattenedSrc] = + MemoryInfo(offset_int_vec(FlattenedSrc), MemoryLifetime::Temporary, _flattened_src.total_size()); } -Status ClFullyConnected::validate(const ITensorInfo *src, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *dst, +Status ClFullyConnected::validate(const ITensorInfo *src, + const ITensorInfo *weights, + const ITensorInfo *biases, + const ITensorInfo *dst, FullyConnectedLayerInfo fc_info) { ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src, weights, dst); - ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(src, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::F16, DataType::F32); + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(src, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, + DataType::F16, DataType::F32); ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, weights, dst); ARM_COMPUTE_RETURN_ERROR_ON(weights->num_dimensions() > 2); - ARM_COMPUTE_RETURN_ERROR_ON(fc_info.activation_info.enabled() && is_data_type_quantized(src->data_type()) && fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::RELU - && fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::BOUNDED_RELU && fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU); + ARM_COMPUTE_RETURN_ERROR_ON( + fc_info.activation_info.enabled() && is_data_type_quantized(src->data_type()) && + fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::RELU && + fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::BOUNDED_RELU && + fc_info.activation_info.activation() != ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU); const GPUTarget gpu_target = get_arch_from_target(CLScheduler::get().target()); const bool transpose_weights = fc_info.transpose_weights ? !fc_info.are_weights_reshaped : false; @@ -441,11 +485,20 @@ Status ClFullyConnected::validate(const ITensorInfo *src, const ITensorInfo *wei // When using dynamic weights - use matmul kernels. // Note: MatMul does not support broadcasting so fallback with batched cases. const bool is_batched_fc_layer = dst->dimension(1) > 1; - const bool use_matmul = gpu_target != GPUTarget::MIDGARD && !weights->are_values_constant() && !is_batched_fc_layer && !(src->num_dimensions() > 1 && (src->data_layout() != fc_info.weights_trained_layout)); - - const ITensorInfo &flatten_src = TensorInfo(src->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_flatten_shape(src)).set_data_layout(DataLayout::NCHW)); - const ITensorInfo &reshaped_weights = TensorInfo(weights->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_transposed_shape(*weights))); - const ITensorInfo &converted_weights = (transpose_weights && !use_matmul) ? TensorInfo(*reshaped_weights.clone()) : TensorInfo(weights->clone()->set_is_resizable(true).reset_padding()); + const bool use_matmul = gpu_target != GPUTarget::MIDGARD && !weights->are_values_constant() && + !is_batched_fc_layer && + !(src->num_dimensions() > 1 && (src->data_layout() != fc_info.weights_trained_layout)); + + const ITensorInfo &flatten_src = TensorInfo(src->clone() + ->set_is_resizable(true) + .reset_padding() + .set_tensor_shape(compute_flatten_shape(src)) + .set_data_layout(DataLayout::NCHW)); + const ITensorInfo &reshaped_weights = TensorInfo( + weights->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(compute_transposed_shape(*weights))); + const ITensorInfo &converted_weights = (transpose_weights && !use_matmul) + ? TensorInfo(*reshaped_weights.clone()) + : TensorInfo(weights->clone()->set_is_resizable(true).reset_padding()); // With the Fully Connected layer we can have 4 different cases: // 1) Convolution layer -> Fully Connected layer without batches @@ -456,10 +509,10 @@ Status ClFullyConnected::validate(const ITensorInfo *src, const ITensorInfo *wei const ITensorInfo *src_to_use = src; const ITensorInfo *weights_to_use = weights; - if(biases != nullptr) + if (biases != nullptr) { ARM_COMPUTE_RETURN_ERROR_ON(biases->num_dimensions() > 1); - if(is_data_type_quantized(src->data_type())) + if (is_data_type_quantized(src->data_type())) { ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(biases, 1, DataType::S32); } @@ -470,11 +523,11 @@ Status ClFullyConnected::validate(const ITensorInfo *src, const ITensorInfo *wei } // Check if FC is after conv (flatten kernel is run in case where FC is after conv.) - if(is_batched_fc_layer) + if (is_batched_fc_layer) { - is_fc_after_conv = (TensorShape::num_max_dimensions >= 4) && (std::equal(src->tensor_shape().cbegin() + 3, - src->tensor_shape().cend(), - dst->tensor_shape().cbegin() + 1)); + is_fc_after_conv = (TensorShape::num_max_dimensions >= 4) && + (std::equal(src->tensor_shape().cbegin() + 3, src->tensor_shape().cend(), + dst->tensor_shape().cbegin() + 1)); } else { @@ -482,29 +535,28 @@ Status ClFullyConnected::validate(const ITensorInfo *src, const ITensorInfo *wei } // Transpose kernel does not run when matmul is supported as matmul fuses transpose op. - if(transpose_weights && !use_matmul) + if (transpose_weights && !use_matmul) { // Validate reshape weights kernel ARM_COMPUTE_RETURN_ON_ERROR(ClTranspose::validate(weights, &reshaped_weights)); weights_to_use = &reshaped_weights; } - if(is_fc_after_conv && (src->data_layout() != fc_info.weights_trained_layout)) + if (is_fc_after_conv && (src->data_layout() != fc_info.weights_trained_layout)) { // Validate convert weights kernel - ARM_COMPUTE_RETURN_ON_ERROR(ClConvertFullyConnectedWeights::validate(weights_to_use, - &converted_weights, - src->tensor_shape(), - fc_info.weights_trained_layout)); + ARM_COMPUTE_RETURN_ON_ERROR(ClConvertFullyConnectedWeights::validate( + weights_to_use, &converted_weights, src->tensor_shape(), fc_info.weights_trained_layout)); weights_to_use = &converted_weights; } - if(is_fc_after_conv) + if (is_fc_after_conv) { // Fully Connected layer after a Convolution Layer without batches // K Index of matrix multiplication. MatMul performs transpose in kernel, so index is 0 when matmul and transpose enabled const int weight_idx = (use_matmul && transpose_weights) ? 0 : 1; - ARM_COMPUTE_RETURN_ERROR_ON((weights_to_use->dimension(weight_idx) != (src->dimension(0) * src->dimension(1) * src->dimension(2)))); + ARM_COMPUTE_RETURN_ERROR_ON( + (weights_to_use->dimension(weight_idx) != (src->dimension(0) * src->dimension(1) * src->dimension(2)))); // Validate flatten kernel ARM_COMPUTE_RETURN_ON_ERROR(ClFlatten::validate(src, &flatten_src)); @@ -539,24 +591,24 @@ void ClFullyConnected::run(ITensorPack &tensors) CLAuxTensorHandler weights(_weights_to_use_idx, _weights_to_use, tensors, false); // Linearize input if it comes from a convolutional layer - if(_is_fc_after_conv) + if (_is_fc_after_conv) { - ITensorPack flatten_pack{ { ACL_SRC, src }, { ACL_DST, flattened_src.get() } }; + ITensorPack flatten_pack{{ACL_SRC, src}, {ACL_DST, flattened_src.get()}}; _flatten->run(flatten_pack); } ITensorPack gemm_pack = tensors; gemm_pack.add_const_tensor(ACL_SRC_0, (_is_fc_after_conv) ? flattened_src.get() : src); - if(_weights_to_use_idx != ACL_SRC_1) + if (_weights_to_use_idx != ACL_SRC_1) { gemm_pack.add_const_tensor(ACL_SRC_1, weights.get()); } // Run MatMul Op - if(_use_matmul) + if (_use_matmul) { // Run matmul kernels for matrix multiplication - if(_is_quantized) + if (_is_quantized) { CLScheduler::get().enqueue_op(*_matmul_lowp_native_kernel, gemm_pack, true); } @@ -568,7 +620,7 @@ void ClFullyConnected::run(ITensorPack &tensors) else { // Run matrix multiply - if(_is_quantized) + if (_is_quantized) { _mm_gemmlowp->run(gemm_pack); } @@ -582,7 +634,7 @@ void ClFullyConnected::run(ITensorPack &tensors) void ClFullyConnected::prepare(ITensorPack &tensors) { // Note : Running prepare() each run when _use_matmul is true is unnecessary unless weights conversion is needed. - if(!_is_prepared || _dynamic_gemm) + if (!_is_prepared || _dynamic_gemm) { #ifdef ARM_COMPUTE_ASSERTS_ENABLED ++_asrt_prepare_count; @@ -598,10 +650,10 @@ void ClFullyConnected::prepare(ITensorPack &tensors) const ITensor *cur_weights = weights; // Reshape weights if needed. Disabled when matmul kernels are enabled as matmul fuses transpose. - if(_transpose_weights && !_use_matmul) + if (_transpose_weights && !_use_matmul) { // Run reshape weights kernel and mark weights as unused - ITensorPack transpose_pack{ { ACL_SRC, weights }, { ACL_DST, reshaped_weights.get() } }; + ITensorPack transpose_pack{{ACL_SRC, weights}, {ACL_DST, reshaped_weights.get()}}; _reshape_weights->run(transpose_pack); cur_weights->mark_as_unused(); @@ -609,9 +661,9 @@ void ClFullyConnected::prepare(ITensorPack &tensors) } // Convert weights if needed - if(_run_convert_weights) + if (_run_convert_weights) { - ITensorPack convert_pack{ { ACL_SRC, cur_weights }, { ACL_DST, converted_weights.get() } }; + ITensorPack convert_pack{{ACL_SRC, cur_weights}, {ACL_DST, converted_weights.get()}}; _convert_weights->run(convert_pack); cur_weights->mark_as_unused(); @@ -622,9 +674,9 @@ void ClFullyConnected::prepare(ITensorPack &tensors) gemm_pack.add_const_tensor(ACL_SRC_1, cur_weights); // Prepare GEMM prepare and release unused weights - if(_dynamic_gemm || !_use_matmul) + if (_dynamic_gemm || !_use_matmul) { - if(!_is_quantized) + if (!_is_quantized) { _mm_gemm->prepare(gemm_pack); } |