diff options
Diffstat (limited to 'arm_compute/runtime/CL/functions/CLSoftmaxLayer.h')
-rw-r--r-- | arm_compute/runtime/CL/functions/CLSoftmaxLayer.h | 58 |
1 files changed, 23 insertions, 35 deletions
diff --git a/arm_compute/runtime/CL/functions/CLSoftmaxLayer.h b/arm_compute/runtime/CL/functions/CLSoftmaxLayer.h index ab10a64de4..68541e35c5 100644 --- a/arm_compute/runtime/CL/functions/CLSoftmaxLayer.h +++ b/arm_compute/runtime/CL/functions/CLSoftmaxLayer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2020 Arm Limited. + * Copyright (c) 2017-2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -21,11 +21,9 @@ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -#ifndef ARM_COMPUTE_CLSOFTMAXLAYER_H -#define ARM_COMPUTE_CLSOFTMAXLAYER_H +#ifndef ACL_ARM_COMPUTE_RUNTIME_CL_FUNCTIONS_CLSOFTMAXLAYER_H +#define ACL_ARM_COMPUTE_RUNTIME_CL_FUNCTIONS_CLSOFTMAXLAYER_H -#include "arm_compute/runtime/CL/CLTensor.h" -#include "arm_compute/runtime/CL/functions/CLPermute.h" #include "arm_compute/runtime/IFunction.h" #include "arm_compute/runtime/IMemoryManager.h" #include "arm_compute/runtime/MemoryGroup.h" @@ -34,11 +32,9 @@ namespace arm_compute { -class CLCompileContext; -class CLLogits1DMaxShiftExpSumKernel; -class CLLogits1DNormKernel; class ICLTensor; class ITensorInfo; +class CLCompileContext; /** Basic function to compute a SoftmaxLayer. * @@ -47,12 +43,6 @@ class ITensorInfo; * * Log Softmax is calculated by : * @f[ out = (x - max(x) * beta) - log(\sum{e^{x - max(x) * beta}}) @f] - * - * This function runs the following kernels: - * -# If axis is not 0: - * -# @ref CLPermute - * -# @ref CLLogits1DNormKernel - * -# @ref CLLogits1DMaxShiftExpSumKernel */ template <bool IS_LOG = false> class CLSoftmaxLayerGeneric : public IFunction @@ -60,18 +50,21 @@ class CLSoftmaxLayerGeneric : public IFunction public: /** Constructor */ CLSoftmaxLayerGeneric(std::shared_ptr<IMemoryManager> memory_manager = nullptr); - /** Prevent instances of this class from being copied */ - CLSoftmaxLayerGeneric(const CLSoftmaxLayerGeneric &) = delete; - /** Prevent instances of this class from being copied */ - CLSoftmaxLayerGeneric &operator=(const CLSoftmaxLayerGeneric &) = delete; - /** Prevent instances of this class to be moved */ - CLSoftmaxLayerGeneric(CLSoftmaxLayerGeneric &&) = delete; - /** Prevent instances of this class to be moved */ - CLSoftmaxLayerGeneric &operator=(CLSoftmaxLayerGeneric &&) = delete; /** Default destructor */ ~CLSoftmaxLayerGeneric(); /** Set the input and output tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src |dst | + * |:--------------|:--------------| + * |QASYMM8 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED | + * |F16 |F16 | + * |F32 |F32 | + * * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32 for Softmax and F16/F32 for Log Softmax * @param[out] output Destination tensor. Data types supported: same as @p input * @param[in] beta (Optional) A scaling factor for the exponent. Defaults to 1.f @@ -88,7 +81,11 @@ public: * @param[in] axis (Optional) The dimension in which to apply the function. E.g. for input of shape 4x5x6 and * axis=1, softmax will be applied to 4x6=24 vectors of size 5. Defaults to 0 */ - void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, float beta = 1.0f, int32_t axis = 0); + void configure(const CLCompileContext &compile_context, + const ICLTensor *input, + ICLTensor *output, + float beta = 1.0f, + int32_t axis = 0); /** Static function to check if given info will lead to a valid configuration of @ref CLSoftmaxLayer * * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32 for Softmax and F16/F32 for Log Softmax @@ -105,20 +102,11 @@ public: void run() override; private: - MemoryGroup _memory_group; - CLPermute _permute_input; - CLPermute _permute_output; - std::unique_ptr<CLLogits1DMaxShiftExpSumKernel> _max_shift_exp_sum_kernel; - std::unique_ptr<CLLogits1DNormKernel> _norm_kernel; - CLTensor _max; - CLTensor _sum; - CLTensor _tmp; - CLTensor _input_permuted; - CLTensor _output_permuted; - bool _needs_permute; + struct Impl; + std::unique_ptr<Impl> _impl; }; using CLSoftmaxLayer = CLSoftmaxLayerGeneric<false>; using CLLogSoftmaxLayer = CLSoftmaxLayerGeneric<true>; } // namespace arm_compute -#endif /* ARM_COMPUTE_CLSOFTMAXLAYER_H */ +#endif // ACL_ARM_COMPUTE_RUNTIME_CL_FUNCTIONS_CLSOFTMAXLAYER_H |