xref: /drivers/interface/nnrt/v2_0/IPreparedModel.idl

/*
 * Copyright (c) 2023 Huawei Device Co., Ltd.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

/**
 * @addtogroup NNRt
 * @{
 *
 * @brief Provides a unified interface for AI chip drivers to access OpenHarmony.\n
 * Neural Network Runtime (NNRt) is a cross-chip inference computing runtime environment oriented to the AI field.
 *
 * @since 3.2
 * @version 2.0
 */

/**
 * @file IPreparedModel.idl
 *
 * @brief Defines the APIs for AI model inference, obtaining the input tensor dimension range,\n
 * and exporting the AI model built.
 *
 * @since 3.2
 * @version 2.0
 */

/**
 * @brief Defines the package path of the NNRt module.
 *
 * @since 3.2
 * @version 2.0
 */
package ohos.hdi.nnrt.v2_0;

import ohos.hdi.nnrt.v2_0.NnrtTypes;

/**
 * @brief Provides the APIs for exporting AI models and performing AI model inference.
 *
 * @since 3.2
 * @version 2.0
 */
interface IPreparedModel {
    /**
     * @brief Exports an AI model from the cache.
     *
     * @param modelCache Array of the model files, which are in the same sequence as they exported.\n
     *For details, see {@link SharedBuffer}.
     *
     * @return Returns <b>0</b> if the operation is successful.
     * @return Returns a non-0 value if the operation fails. A negative value is an HDF standard error code,\n
     *and a positive value is a dedicated error code defined by NNRt. For details, see {@link NNRT_ReturnCode}.
     */
    ExportModelCache([out] struct SharedBuffer[] modelCache);

    /**
     * @brief Obtains the tensor dimension range supported by AI model. If a fixed dimension is used,\n
     *the maximum dimension value is the same as the minimum dimension value.
     *
     * @param minInputDims Two-dimensional array that stores the minimum dimension of the model input data.\n
     *The first dimension of the array indicates the number of tensors, and the second dimension indicates the number of dimensions of the tensors.
     * @param maxInputDims Two-dimensional array that stores the maximum dimension of the model input data.\n
     *The first dimension of the array indicates the number of tensors, and the second dimension indicates the number of dimensions of the tensors.
     *
     * @return Returns <b>0</b> if the operation is successful.
     * @return Returns a non-0 value if the operation fails. A negative value is an HDF standard error code,\n
     *and a positive value is a dedicated error code defined by NNRt. For details, see {@link NNRT_ReturnCode}.
     */
    GetInputDimRanges([out] unsigned int[][] minInputDims, [out] unsigned int[][] maxInputDims);

    /**
     * @brief Performs AI model inference.
     *
     * @param inputs Input data for AI model inference. The data is input in the sequence defined by the model.\n
     *For details about the input data type, see {@link IOTensor}.
     * @param outputs Output data of AI model inference. After inference, the output data is written to the\n
     *shared buffer. For details about the output data type, see {@link IOTensor}.
     * @param outputDims Dimensions of the output data. The output sequence is the same as that of <b>outputs</b>.
     * @param isOutputBufferEnough Whether the shared buffer space is sufficient for the output data.\n
     *The value <b>true</b> means the shared buffer space is sufficient; the value <b>false</b> means the opposite.
     *
     * @return Returns <b>0</b> if the operation is successful.
     * @return Returns a non-0 value if the operation fails. A negative value is an HDF standard error code,\n
     *and a positive value is a dedicated error code defined by NNRt. For details, see {@link NNRT_ReturnCode}.
     */
    Run([in] struct IOTensor[] inputs, [in] struct IOTensor[] outputs, [out] int[][] outputDims);
}

/** @} */