[[["容易理解","easyToUnderstand","thumb-up"],["確實解決了我的問題","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["缺少我需要的資訊","missingTheInformationINeed","thumb-down"],["過於複雜/步驟過多","tooComplicatedTooManySteps","thumb-down"],["過時","outOfDate","thumb-down"],["翻譯問題","translationIssue","thumb-down"],["示例/程式碼問題","samplesCodeIssue","thumb-down"],["其他","otherDown","thumb-down"]],["上次更新時間:2024-10-01 (世界標準時間)。"],[[["To use custom-trained models in Earth Engine, configure them for compatibility with Vertex AI for seamless inference requests."],["Earth Engine supports various payload formats, including gRPC and HTTP, for sending inference requests to your model, with gRPC being preferred for TensorFlow models due to its lower latency."],["When using gRPC, select the appropriate format (`GRPC_TF_TENSORS`, `GRPC_SERIALIZED_TF_TENSORS`, or `GRPC_SERIALIZED_TF_EXAMPLES`) based on your model's requirements and whether it needs a gRPC prediction handler."],["For HTTP payloads, Earth Engine offers options like `SERIALIZED_TF_TENSORS` (default and efficient for TensorFlow), `RAW_JSON` (flexible for PyTorch and AutoML), and `ND_ARRAYS` (for PyTorch with numerical data)."],["Carefully choose the payload format that best aligns with your model framework and data structure for optimal performance and compatibility."]]],["Earth Engine (EE) interacts with custom models on Vertex AI via inference requests. Users configure the payload format using `payloadFormat` in `ee.Model.fromVertexAi`. gRPC is preferred for lower latency and higher reliability, using formats like `GRPC_TF_TENSORS`, `GRPC_SERIALIZED_TF_TENSORS`, or `GRPC_SERIALIZED_TF_EXAMPLES`. HTTP API payloads are also supported, including `SERIALIZED_TF_TENSORS` (default for TensorFlow), `RAW_JSON` (for PyTorch/AutoML, converting numbers to strings), and `ND_ARRAYS` (similar to `RAW_JSON`, but without keys). Each format determines how data is sent.\n"]]