Tabular Workflow untuk Perkiraan adalah pipeline lengkap untuk
tugas perkiraan. Hal ini mirip dengan
AutoML API,
tetapi memungkinkan Anda memilih apa yang akan dikontrol dan diotomatisasi. Alih-alih memiliki kontrol untuk seluruh pipeline, Anda memiliki kontrol untuk setiap langkah di pipeline. Kontrol pipeline ini mencakup:
Pemisahan data
Rekayasa fitur
Penelusuran arsitektur
Pelatihan model
Ansambel model
Manfaat
Berikut adalah beberapa manfaat
Tabular Workflow untuk Perkiraan
:
Mendukung set data besar yang berukuran hingga 1 TB dan memiliki maksimal 200 kolom.
Memungkinkan Anda meningkatkan stabilitas dan menurunkan waktu pelatihan dengan membatasi ruang penelusuran jenis arsitektur atau melewati penelusuran arsitektur.
Memungkinkan Anda meningkatkan kecepatan pelatihan dengan memilih secara manual hardware yang digunakan untuk penelusuran arsitektur dan pelatihan.
Memungkinkan Anda mengurangi ukuran model dan meningkatkan latensi dengan mengubah ukuran ansambel.
Setiap komponen AutoML dapat diperiksa dengan antarmuka grafik pipeline andal yang memungkinkan Anda melihat tabel data yang ditransformasi, arsitektur model yang dievaluasi, dan banyak detail lainnya.
Setiap komponen mendapatkan fleksibilitas dan transparansi yang lebih luas, seperti kemampuan untuk menyesuaikan parameter, hardware, status proses tampilan, log, dan lain-lain.
Perkiraan di Vertex AI Pipelines
Tabular Workflow untuk Perkiraan
adalah instance terkelola dari Vertex AI Pipelines.
Vertex AI Pipelines adalah layanan
tanpa server yang menjalankan pipeline Kubeflow. Anda dapat menggunakan pipeline untuk mengotomatisasi
dan memantau machine learning serta tugas penyiapan data Anda. Setiap langkah di
pipeline menjalankan bagian dari alur kerja pipeline. Misalnya,
pipeline dapat mencakup langkah-langkah untuk memisahkan data, mengubah jenis data, dan melatih model. Karena langkah tersebut
adalah instance komponen pipeline, langkah memiliki input, output, dan
image container. Input langkah dapat ditetapkan dari input pipeline atau dapat
bergantung pada output langkah lain dalam pipeline ini. Dependensi ini
menentukan alur kerja pipeline sebagai directed acyclic graph.
Ringkasan pipeline dan komponen
Diagram berikut menunjukkan pipeline pemodelan untuk
Tabular Workflow for Forecasting
:
Komponen pipeline adalah:
feature-transform-engine: Melakukan rekayasa fitur. Lihat Feature Transform Engine untuk mengetahui detailnya.
training-configurator-and-validator: Memvalidasi konfigurasi pelatihan dan membuat metadata pelatihan.
Input:
instance_schema: Skema instance dalam spesifikasi OpenAPI, yang menjelaskan jenis data dari data inferensi.
dataset_stats: Statistik yang mendeskripsikan set data mentah. Misalnya, dataset_stats memberikan jumlah baris dalam set data.
training_schema: Skema data pelatihan dalam spesifikasi OpenAPI, yang
menjelaskan jenis data dari data pelatihan.
split-materialized-data: Memisahkan data terwujud ke dalam set pelatihan, set evaluasi, dan set pengujian.
calculate-training-parameters-2: Menghitung durasi runtime yang diharapkan untuk automl-forecasting-stage-1-tuner.
get-hyperparameter-tuning-results - Opsional: Jika Anda mengonfigurasi pipeline untuk melewati penelusuran arsitektur, muat hasil penyesuaian hyperparameter dari pipeline sebelumnya.
Melakukan penelusuran arsitektur model dan menyesuaikan hyperparameter (automl-forecasting-stage-1-tuner) atau menggunakan hasil penyesuaian hyperparameter dari operasi pipeline sebelumnya (automl-forecasting-stage-2 -tuner).
Arsitektur ditentukan oleh sekumpulan hyperparameter.
Hyperparameter mencakup jenis model dan parameter model.
Jenis model yang dipertimbangkan adalah jaringan neural dan hierarki yang ditingkatkan.
Sebuah model dilatih untuk setiap arsitektur yang dipertimbangkan.
artifact - Hasil penyesuaian hyperparameter dari operasi pipeline sebelumnya.
Artefak ini adalah input hanya jika Anda mengonfigurasi pipeline untuk melewati
penelusuran arsitektur.
Output:
tuning_result_output: Menyesuaikan output.
get-prediction-image-uri-2: Membuat URI gambar inferensi yang benar berdasarkan jenis model.
automl-forecasting-ensemble-2: Menggabungkan arsitektur terbaik untuk menghasilkan model akhir.
Input:
tuning_result_output: Menyesuaikan output.
Output:
unmanaged_container_model: Model output.
model-upload-2 - Mengupload model.
Input:
unmanaged_container_model: Model output.
Output:
model: Model Vertex AI.
should_run_model_Evaluation - Opsional: Menggunakan set pengujian untuk menghitung metrik evaluasi.
[[["Mudah dipahami","easyToUnderstand","thumb-up"],["Memecahkan masalah saya","solvedMyProblem","thumb-up"],["Lainnya","otherUp","thumb-up"]],[["Sulit dipahami","hardToUnderstand","thumb-down"],["Informasi atau kode contoh salah","incorrectInformationOrSampleCode","thumb-down"],["Informasi/contoh yang saya butuhkan tidak ada","missingTheInformationSamplesINeed","thumb-down"],["Masalah terjemahan","translationIssue","thumb-down"],["Lainnya","otherDown","thumb-down"]],["Terakhir diperbarui pada 2025-09-02 UTC."],[],[],null,["# Tabular Workflow for Forecasting\n\n| **Preview**\n|\n|\n| This feature is subject to the \"Pre-GA Offerings Terms\" in the General Service Terms section\n| of the [Service Specific Terms](/terms/service-terms#1).\n|\n| Pre-GA features are available \"as is\" and might have limited support.\n|\n| For more information, see the\n| [launch stage descriptions](/products#product-launch-stages).\n\nThis document provides an overview of\nTabular Workflow for Forecasting\n\n[pipeline and components](#components). To learn how to train a model, see\n[Train a model with\nTabular Workflow for Forecasting](/vertex-ai/docs/tabular-data/tabular-workflows/forecasting-train).\n\n\nTabular Workflow for Forecasting is the complete pipeline for\nforecasting tasks. It is similar to the\n[AutoML API](/vertex-ai/docs/tabular-data/forecasting/overview),\nbut lets you to choose what to control and what to automate. Instead of having\ncontrols for the *whole* pipeline, you have controls for *every step* in the\npipeline. These pipeline controls include:\n\n- Data splitting\n- Feature engineering\n- Architecture search\n- Model training\n- Model ensembling\n\n\u003cbr /\u003e\n\nBenefits\n--------\n\nThe following are some of the benefits of\nTabular Workflow for Forecasting\n:\n\n\n- Supports **large datasets** that are up to 1TB in size and have up to 200 columns.\n- Lets you **improve stability and lower training time** by limiting the search space of architecture types or skipping architecture search.\n- Lets you **improve training speed** by manually selecting the hardware used for training and architecture search.\n- Lets you **reduce model size and improve latency** by changing the ensemble size.\n- Each component can be inspected in a powerful pipelines graph interface that lets you see the transformed data tables, evaluated model architectures and many more details.\n- Each component gets extended flexibility and transparency, such as being able to customize parameters, hardware, view process status, logs and more.\n\n\u003cbr /\u003e\n\nForecasting on Vertex AI Pipelines\n----------------------------------\n\n\nTabular Workflow for Forecasting\nis a managed instance of Vertex AI Pipelines.\n\n\n[Vertex AI Pipelines](/vertex-ai/docs/pipelines/introduction) is a serverless\nservice that runs Kubeflow pipelines. You can use pipelines to automate\nand monitor your machine learning and data preparation tasks. Each step in a\npipeline performs part of the pipeline's workflow. For example,\na pipeline can include steps to split data, transform data types, and train a model. Since steps\nare instances of pipeline components, steps have inputs, outputs, and a\ncontainer image. Step inputs can be set from the pipeline's inputs or they can\ndepend on the output of other steps within this pipeline. These dependencies\ndefine the pipeline's workflow as a directed acyclic graph.\n\nOverview of pipeline and components\n-----------------------------------\n\nThe following diagram shows the modeling pipeline for\nTabular Workflow for Forecasting\n:\n\n\u003cbr /\u003e\n\nThe pipeline components are:\n\n1. **feature-transform-engine** : Performs feature engineering. See [Feature Transform Engine](/vertex-ai/docs/tabular-data/tabular-workflows/feature-engineering) for details.\n2. **training-configurator-and-validator**: Validates the training configuration and generates the training metadata.\n\n Input:\n - `instance_schema`: Instance schema in OpenAPI specification, which describes the data types of the inference data.\n - `dataset_stats`: Statistics that describe the raw dataset. For example, `dataset_stats` gives the number of rows in the dataset.\n - `training_schema`: Training data schema in OpenAPI specification, which describes the data types of the training data.\n3. **split-materialized-data**: Splits the materialized data into a training set, an evaluation set, and a test set.\n\n Input:\n - `materialized_data`: Materialized data.\n\n Output:\n - `materialized_train_split`: Materialized training split.\n - `materialized_eval_split`: Materialized evaluation split.\n - `materialized_test_split`: Materialized test set.\n4. **calculate-training-parameters-2** : Calculates the expected runtime duration\n for **automl-forecasting-stage-1-tuner**.\n\n5. **get-hyperparameter-tuning-results** - **Optional**: If you configure the\n pipeline to skip the architecture search, load the hyperparameter tuning\n results from a previous pipeline run.\n\n6. Perform model architecture search and tune hyperparameters (**automl-forecasting-stage-1-tuner** ) or use the hyperparameter tuning results\n from a previous pipeline run (**automl-forecasting-stage-2-tuner**).\n\n - An architecture is defined by a set of hyperparameters.\n - Hyperparameters include the model type and the model parameters.\n - Model types considered are neural networks and boosted trees.\n - A model is trained for each architecture considered.\n\n Input:\n - `materialized_train_split`: Materialized training split.\n - `materialized_eval_split`: Materialized evaluation split.\n - `artifact` - Hyperparameter tuning results from a previous pipeline run. This artifact is an input only if you configure the pipeline to skip the architecture search.\n\n Output:\n - `tuning_result_output`: Tuning output.\n7. **get-prediction-image-uri-2** : Produces the correct inference image URI based on the [model type](/vertex-ai/docs/tabular-data/forecasting/train-model#training-methods).\n\n8. **automl-forecasting-ensemble-2**: Ensembles the best architectures to produce a final model.\n\n Input:\n - `tuning_result_output`: Tuning output.\n\n Output:\n - `unmanaged_container_model`: Output model.\n9. **model-upload-2** - Uploads the model.\n\n Input:\n - `unmanaged_container_model`: Output model.\n\n Output:\n - `model`: Vertex AI model.\n10. **should_run_model_evaluation** - **Optional**: Use the test set to calculate evaluation metrics.\n\nWhat's next\n-----------\n\n- [Train a model using Tabular Workflow for Forecasting](/vertex-ai/docs/tabular-data/tabular-workflows/forecasting-train)."]]
